d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/md5pickle md5pickle was documented, but not declared. it didn't work either. (didn't archive state->len, missed out the terminating null for enc64) d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/hget-tcpdir tcpdir should be NETPATHLEN bytes long, not 64. It ensures that overflow error is consistent with the system limit. d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/ap2 ap fixed for real this time -Steve d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/acmechord i've noticed that when snarfing some text in an acme window by chording cut & paste that the window often ends up in "modified" state. after some experimentation, i've found that this is because i'm generating the following sequence of button events (button 1 least significant): 0 0 1 0 1 1 1 1 1 1 0 1 1 0 0 0 0 0 i.e. button 1 is being released before button 3. acme interprets the last event as an extra "paste", which sets the window to modified. this patch is one way of fixing the problem: it makes b1-b2-b3-b3... equivalent to: b1-b2-b3 the fix relies on the observation that once you've done one operation (cut or paste) you'll never do a different operation (other than undo) for that chording sequence. d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/trampoline there wasn't a manual page for it d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/exportfs i noticed that i couldn't import /net.alt from a server. on the outside network it was in the name space from /lib/namespace.ext. exportfs's newns used the default. fair enough but surprisingly a compensating bind /lib/namespace.ext /lib/namespace in /bin/service.alt/il17007 didn't work, because exportfs does two newns! auth_chuid does one itself, which uses the rebound /lib/namespace, but leaves a name space in which the bind no longer was there, and then on return exportfs does another, using the `original' /lib/namespace, which doesn't set up (the right) /net.alt. two changes: one, remove the second newns, because it didn't seem justified. two, add a -n option to allow a name space file to be specified explicitly (which is perhaps more straightforward in some cases than the bind). i added an explanatory remark about name space creation to the manual page, and in that part i documented the new -n option. d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/face.entry add myself to faces --npe d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/mondef.cyberhack a case statement added to let the cyber driver to recognize TGUI 9320's(need to turn off linear to get it to work though) Samsung SyncMaster 753DF monitor sync definition -npe d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/mail.bmp bmp support for nedmail and marshal (missing from previous post), also, nedmail now uses file(1) to guess the extension to add to unknown mimetypes, (application/octet-stream attachemnts get sensible extensions) Not entirely happy with this implementation, file(1) has some intelligence, so does marshal, and now nedmail, page also knows some things, and the plumber knows others. Shouldn't this magic (sic) be in one place? -Steve d-rwxrwxr-x M 104260 rsc sys 0 Apr 6 2004 /n/sources/patch/applied/hgetnoseek as discussed on 9fans, a version of hget that doesn't seek on output. it also takes a bit more care to make sure that it doesn't mix up the output from several different documents. it requires the md5pickle patch. although i've done some testing of the output routing independently, i don't know how to provoke webservers into giving the required responses so it needs some more testing. d-rwxrwxr-x M 104260 rsc sys 0 Apr 18 2004 /n/sources/patch/applied/kfsbit delete several unused fields from Wpath d-rwxrwxr-x M 104260 rsc sys 0 Apr 18 2004 /n/sources/patch/applied/ether-multicast-shutdown new ethernet drivers with multicast and shutdown entry points; not all have been tested. multicast is needed for ipv6 and shutdown is needed for /dev/reboot. d-rwxrwxr-x M 104260 rsc sys 0 Apr 18 2004 /n/sources/patch/applied/vnclabel Make vncv set /dev/label (if the file is there) to the serveraddr. Otherwise we may have several clients that show up as "vncv" in window lists. Not a big deal, but convenient. d-rwxrwxr-x M 104260 rsc sys 0 Apr 18 2004 /n/sources/patch/applied/fixedmap2 This is the fixedmap sent again, as you asked for. d-rwxrwxr-x M 104260 rsc sys 0 Apr 19 2004 /n/sources/patch/applied/vnc.utf-keysym This adds support in vncv and vncs for sending/receiving utf-8 text. The problem comes from the RFB protocol which requires special keys to pass through as their keysym encodings (). vncv and vncs do not convert between utf-8 and keysym, therefore they're unusable when one wants to write in foreign characters that don't match keysyms. the two header files (ksym2utf.h and utf2ksym.h) declare arrays with the same names mapping keysym definitions to unicode runes. kbdv.c and kbds.c get the corresponding character when the rune read from /dev/cons or from a client. inadvertantly this fixes a problem with vncv which would cause xterminals not to scroll on ctrl+[PgUp|PgDn]. also changed is the mkfile to add dependency between kbdv.c→utf2ksym.h and kbds.c→ksym2utf.h the arrays add a total of ~32K to the run-time memory image of vncv and vncs. andrey d-rwxrwxr-x M 104260 rsc sys 0 Apr 19 2004 /n/sources/patch/applied/vnc.ksym-hfiles added about 300 more ksym2utf and utf2ksym conversions. this is believed to be complete now, because it was auto-generated from 4.3.0's keysymdef.h (the older version was from, i believe 1996... there's more work to be done -- unicode → keysym is done by: keysym = unicode | 0x1000000; conversely keysym → unicode is: unicode = keysym & 0xffffff; but there appears to be a big problem with some servers accepting such conversions and others not. this affects unicode glyphs > 0x20ac... not sure if anything should be done immediately or it's better to have someone complain first and then see whether the unix/windows servers have started accepting this behaviour. for now we won't be able to type smileys in vnc ☺ d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/fossilcons-man describe the 'who' command in a paragraph of its own. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/beatles adds: I am he as you are he as you are me and we are all together. - forsyth after Lennon on authentication d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/page-man i should've just put the man page in the original patch, sorry for generating noise. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/page-zerox add a zerox-like functionality accessible via the middle mouse button. creates a new window with the current image only, useful for selecting important pages of large documents. note that it doesn't implement _full_ Zerox functionality because modifications to the current image aren't reflected in the zerox-ed one. it is doubtful that such functionality is needed. man page will be modified as soon as the name 'zerox' is agreed upon. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/vga-ituner support for vga card found in iTuner MiniBox d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/ipv6-addressing correct address-test macros to check addresses in conformance with rfc 3513, thus detecting more addresses in each class. lightly tested; nothing broke. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/statsfixes This does a couple of things: 1. Honors the comment by presotto, and removes the access for ether stats in the old place. 2. Fixes the index for what's etherout and ethererr the stats contain In Link Out, but stats thinks it has In Out Err... 3. does not put dead in the machine name when it does not respond for a while hth d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/ipv6-through-gw This corrects ipv6.h to recognise global addresses as starting with the byte 0x20 instead of 0x02. I've tested it; it works. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/ipv6-rev-dns-domain change ip6.int to ip6.arpa, per the rfcs. of course, nothing in ndb mangles ipv6 addresses into the form needed for reverse lookups yet. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/apeldr01 Fixes instigated by attempting to compile Tcl-7.6. - gettimeofday() used wrong device for time sampling - vfprintf.c ignored left justification in favour of zero left padding - wait.c was rewritten to reap processes by PID - _buf.c miscalculated the number of open FDs in select() - added wait3() and wait4() prototypes to "sys/wait.h" - unprotected hypot() prototype in "math.h". d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/apeldr02 Fixes instigated by attempting to compile Tcl-7.6. - gettimeofday() used wrong device for time sampling - vfprintf.c ignored left justification in favour of zero left padding - wait.c was rewritten to reap processes by PID - _buf.c miscalculated the number of open FDs in select() - added wait3() and wait4() prototypes to "sys/wait.h" - unprotected hypot() prototype in "math.h". d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/apeldr03 Fixes instigated by attempting to compile Tcl-7.6. - gettimeofday() used wrong device for time sampling - vfprintf.c ignored left justification in favour of zero left padding - wait.c was rewritten to reap processes by PID - _buf.c miscalculated the number of open FDs in select() - added wait3() and wait4() prototypes to "sys/wait.h" - unprotected hypot() prototype in "math.h". d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/lib9p-leak Make sure that closefid/closereq are called with a refcount of 1 when fid or request are not returned. (non-return should not be counted as return in refcount). Without the patch, the fid/req structure will leak in the error case. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/sys-lib-newuser create users' tmp/ directories with the -t flag (i.e. temporary files) so they won't be stored to venti. d-rwxrwxr-x M 104260 rsc sys 0 Jun 17 2004 /n/sources/patch/applied/smtpd-net-not-exempt remove my bogus test to exclude /net from /mail/lib/senders validation. just bind the appropriate file to /mail/lib/senders in /rc/bin/service*/tcp25. d-rwxrwxr-x M 104260 rsc sys 0 Jun 21 2004 /n/sources/patch/applied/page-typos fix page(1) typos d-rwxrwxr-x M 104260 jmk sys 0 Jul 1 2004 /n/sources/patch/applied/ape-wait-bug Bugfix for the new PID cache in wait(2). The simple case of process successfull terminated was missed causing make and sh to error and suicide respectively. -Steve d-rwxrwxr-x M 104260 jmk sys 0 Jul 1 2004 /n/sources/patch/applied/nvidia-fix Some fixes for nvidia. (forgot to zero some variables) d-rwxrwxr-x M 104260 jmk sys 0 Jul 3 2004 /n/sources/patch/applied/_good_libcontrolfix This patch fixes a silly bug in libcontrol. It would complaint about column not being a control type. There was an extra Ctltextbutton3 or similar that was set as nil in the list of controls. So, _ctllookup would stop looking up controls. I changed _ctllookup not to stop on the first nil entry, to avoid this kind of bug in the future. I am sorry for the ugly name for the patch. I hit ESC while sending the "libcontrolfix" patch but patch/create did send the patch, so I couldnt recreate it with a non nil comment. hth d-rwxrwxr-x M 104260 jmk sys 0 Jul 6 2004 /n/sources/patch/applied/9p-man-typo change .PP to .IP in description of Open. d-rwxrwxr-x M 104260 jmk sys 0 Jul 11 2004 /n/sources/patch/applied/vgadb-rage adding ATI Technologies Inc 3D Rage II+ 215GTB [Mach64 GTB] (rev 9a) vid=0x1002, did=4755 d-rwxrwxr-x M 104260 jmk sys 0 Jul 11 2004 /n/sources/patch/applied/vgadbmodes This adds more entries for some modes that we use at urjc.es. They are for vmware and virtual pc at various environments. d-rwxrwxr-x M 104260 jmk sys 0 Jul 15 2004 /n/sources/patch/applied/wait4 lpdaemon was crashing. this seems to be the reason why. d-rwxrwxr-x M 104260 jmk sys 0 Jul 30 2004 /n/sources/patch/applied/fqdn-in-greylist tiny change to greylist processing to add fqdn as a comment when adding name to whitelist, makes admin easier. d-rwxrwxr-x M 104260 jmk sys 0 Jul 30 2004 /n/sources/patch/applied/beauty-fortune added: “Beauty is more important in computing than anywhere else in technology because software is so complicated. Beauty is the ultimate defence against complexity.” -- David Gelernter -Steve d-rwxrwxr-x M 104260 jmk sys 0 Jul 30 2004 /n/sources/patch/applied/smtp-man1 added note that /mail/lib/whitelist must be append only -Steve d-rwxrwxr-x M 104260 jmk sys 0 Jul 30 2004 /n/sources/patch/applied/devlm78 while adding support for a winbond hardware monitor i found the lm78 driver. it is in a bad state of disarray, so i updated it to compile for the 4th ed.... the code inside is ugly and there's no man page. d-rwxrwxr-x M 104260 jmk sys 0 Jul 30 2004 /n/sources/patch/applied/hget-basic-auth Added basic authentication to hget via factotum. didn't do challange/response (sorry), but its enough to query my firewall and cable modem's status in the absence of snmpfs. -Steve d-rwxrwxr-x M 104260 jmk sys 0 Jul 31 2004 /n/sources/patch/applied/nbns-portname added udp=nbns port=137 -Steve d-rwxrwxr-x M 104260 jmk sys 0 Aug 2 2004 /n/sources/patch/applied/du-char-overflow this fixes the maximum filename length restriction of the previous du command. back in the days of kfs and ken's fs it was ok (max length of 28 characters per file name) but now it segfaults badly for big directories: plan9% mkdir `{for (i in `{seq 1 1000}) echo -n $i} plan9% du . du 99198: suicide: sys: trap: fault write addr=0x7ffff000 pc=0x000059b1 plan9% ls 1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000 plan9% NOTE: the Strings are created a bit wastefully (each sprint is a malloc now) so please modify the source if you think there's a more elegant way of doing it. d-rwxrwxr-x M 104260 jmk sys 0 Aug 8 2004 /n/sources/patch/applied/tiny-grep-silly Arg parsing didn't check for duff args (written for an early version of ARGBEGIN ARGEND ?), thus "grep -f" suicide'ed. fixed with EARGF(). -Steve d-rwxrwxr-x M 104260 jmk sys 0 Aug 8 2004 /n/sources/patch/applied/mkext-quoted-filenames mkfs -a understands how to quote filenames with %q but mkext doesn't know how to unpack them. My fix doesn't allow any of the fields in the header to be zero length, however looking at mkfs it uses %q for the filename and uid and gid so I don't beleive this is a problem. Perhaps there some old mkfs archives which have missing uid or gid that this change might break? -Steve d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/man-proc-typo wire should be wired. d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/ftpfsremove ftpfs didn't clunk the fid when a remove failed, leading to "fid in use" errors. d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/listen-accept fix listen and listen1 to use the right fds for accept and reject. d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/tftpd-accept-syslog make accept use the right fd; add error checks and improve logging to assist when things go wrong. d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/acdlongnames when a track has a long title, it comes in several parts. this patch allows acd to display such long titles, rather than the last section of the name only. d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/tftpd-none tftpd has been running as bootes rather than none. this fix makes it switch to none correctly. d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/glenda-pull old pull script assumed kfs... d-rwxrwxr-x M 104260 rsc sys 0 Aug 26 2004 /n/sources/patch/applied/upas-help-typoo silly typoo in help text for nedmail -Steve d-rwxrwxr-x M 104260 jmk sys 0 Sep 1 2004 /n/sources/patch/applied/syncarena checkarenas -f didn't mend an inconsistency that was reparable, or rather it did, but forgot that. in my case, clumps had been written but where the writes to the clump info had not completed. syncarenas kept a set of summary errors that noted a fault before deciding whether it could repair it. this change helped me recover my venti without fuss. there might be more along these lines. d-rwxrwxr-x M 104260 jmk sys 0 Sep 1 2004 /n/sources/patch/applied/commfix test script: cat >/tmp/f1 <<'!' a c ! cat >/tmp/f2 <<'!' b c ! fn r { echo $* $* } r comm /tmp/f1 /tmp/f2 r comm -1 /tmp/f1 /tmp/f2 r comm -2 /tmp/f1 /tmp/f2 r comm -3 /tmp/f1 /tmp/f2 r comm -23 /tmp/f1 /tmp/f2 r comm -12 /tmp/f1 /tmp/f2 before: term% r comm /tmp/f1 /tmp/f2 comm /tmp/f1 /tmp/f2 a b c term% r comm -1 /tmp/f1 /tmp/f2 comm -1 /tmp/f1 /tmp/f2 b c term% r comm -2 /tmp/f1 /tmp/f2 comm -2 /tmp/f1 /tmp/f2 a c term% r comm -3 /tmp/f1 /tmp/f2 comm -3 /tmp/f1 /tmp/f2 a b term% r comm -23 /tmp/f1 /tmp/f2 comm -23 /tmp/f1 /tmp/f2 a term% r comm -12 /tmp/f1 /tmp/f2 comm -12 /tmp/f1 /tmp/f2 c note the empty columns in comm -1, comm -2, and comm -12: there is one tab too many. it's an easy fix. i noticed only later that the plan 9 command was essentially the ancient unix one, and a diff against that shows that indeed the l = 1 -> l = 2 was a change in going to plan 9, which dented it. after: doppio% r comm /tmp/f1 /tmp/f2 comm /tmp/f1 /tmp/f2 a b c doppio% r comm -1 /tmp/f1 /tmp/f2 comm -1 /tmp/f1 /tmp/f2 b c doppio% r comm -2 /tmp/f1 /tmp/f2 comm -2 /tmp/f1 /tmp/f2 a c doppio% r comm -3 /tmp/f1 /tmp/f2 comm -3 /tmp/f1 /tmp/f2 a b doppio% r comm -23 /tmp/f1 /tmp/f2 comm -23 /tmp/f1 /tmp/f2 a doppio% r comm -12 /tmp/f1 /tmp/f2 comm -12 /tmp/f1 /tmp/f2 c it now correctly suppresses columns. d-rwxrwxr-x M 104260 jmk sys 0 Sep 1 2004 /n/sources/patch/applied/apewait fixed many things, including a memory leak. it doesn't forget about previously seen pids if WNOHANG set. it doesn't hang in some cases when WNOHANG set. it returns status correctly for things it has put on the list. it handles pid 0 consistently (posix says same process group; here it has always been equivalent to -1,except in the wait4 version one part did and another didn't) it now fills in a struct rusage if provided (though most of it is zero). it should be easier to follow: see if there is a wait record already; if not and WNOHANG set, see if wait will not block (if it will, return); otherwise wait for the pid requested and enlist all wait records that do not match; set optional status and rusage; free the record and return pid. its source format is more plan 9-ish. i've added struct rusage to sys/resource.h, and in sys/wait.h made the declarations of wait4 and rusage conditional on _BSD_EXTENSION, and wait4 says struct rusage* not void* Unix has indeed got very smelly. according to linux lists, posix is going to add another wait variant, so watch this space ... i can't wait. d-rwxrwxr-x M 104260 jmk sys 0 Sep 2 2004 /n/sources/patch/applied/nedmail-singleton The senario: upas/fs -n echo open /apop/pop.sally.com/steve sally.johnston > /mail/fs/ctl echo open /n/fred/mail/box/steve/mbox fred > /mail/fs/ctl faces -i -m /mail/fs/sally.johnston -m /mail/fs/fred B3 clicking on the icons in faces doesn't work. The problem is the logic in nedmail WRT the singleton mail file, which I have added another clause to, however, I'am not sure how presotto expected this to work - I may have worked around a much simpler bug. This code fixes the problem anyway. -Steve d-rwxrwxr-x M 104260 jmk sys 0 Sep 2 2004 /n/sources/patch/applied/fortune-add found on a new zealand linux user group's discussion board in a topic mentioning (in the same sentence) Qt, KDE, GNU Lisp and Plan 9. d-rwxrwxr-x M 104260 jmk sys 0 Sep 8 2004 /n/sources/patch/applied/tareof readme must not be 0-length d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/stats-suicide stats would suicide if envoked with a non-existant host name arg, now checks at least one valid host is specified. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/boot-pc-p4 recognise new southbridge and new pentium IV xeon cpu. d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/9-p4 recognise new southbridge and new pentium IV xeon cpu in the kernel. d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/plot.flushimage flushes the image onto the display after it's done drawing. previous behaviour was to flush after a mouse event. d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/newuser.mail-cron Trivial change to get newuser to run mail -c and cron -c, I never understood why they didn't, perhaps there is a good reason I missed? d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/cpp-buffer there's a buffer size problem that is very rarely hit by any normal programs but appears when compiling some of the nuttier lunix code. solution: if a line of code is longer than the default buffer size (30k) then we realloc the buffer. this isn't currently triggered by any plan9 code and will hopefully never be. for an example of what could happen in loonix-land see this gcc-parsed code: http://pages.cpsc.ucalgary.ca/~mirtchov/screenshots/ffmpeg.gif andrey d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/mach64xx-fix Fix bios freq table offset Enlarge bios size to capture panel info d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/dict-gmail added gmail.com/unknown → google.com enough emails are arriving from gmail nowadays to warrant it. d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/sokoban-levels bug loading the hard level set fixed... andrey d-rwxrwxr-x M 104260 rsc sys 0 Oct 16 2004 /n/sources/patch/applied/libplot-subr-dbgprint removed (debugging?) print statement to avoid print RED for each 'co r' command. d-rwxrwxr-x M 104260 rsc sys 0 Oct 17 2004 /n/sources/patch/applied/vwhois-typo d-rwxrwxr-x M 104260 jmk sys 0 Oct 19 2004 /n/sources/patch/applied/lex lex would suicide under some strange circumstances (not my lex code) Turned out to be multiple Bterm()s of stdin. -Steve d-rwxrwxr-x M 104260 jmk sys 0 Oct 19 2004 /n/sources/patch/applied/scroll-mouse basic scroll-button support for plan9 andrey d-rwxrwxr-x M 104260 jmk sys 0 Oct 19 2004 /n/sources/patch/applied/stub-rootfile-usage handle case where argument is of form /file. e.g. aux/stub /test. also sync usage with man page. also add missing exits. d-rwxrwxr-x M 104260 jmk sys 0 Oct 27 2004 /n/sources/patch/applied/ether82557-pro100ve pci device id for my intel pro/100 ve, an on-board lan nic. btw, judging from pci device id lists in e.g. bsd drivers, there are many more intel nics that work with this driver. should these be added to the drivers, even though they cannot be tested? it could alleviate the plan9 hardware-support problem a bit... also, i didn't modify /sys/src/fs/pc/ether82557.c since i just saw it existed and cannot test it not anyway. mjl (mechiel@xs4all.nl) d-rwxrwxr-x M 104260 jmk sys 0 Oct 29 2004 /n/sources/patch/applied/dst-fixed It appears the EU directed the UK to change when the summer ends (thanks Foryth) http://europa.eu.int/eur-lex/pri/en/oj/dat/2000/ce337/ce33720001128en01360137.pdf I have munged the Unix zic(1) tables to plan9 format by hand, however I have done only /adm/timezone/BG-Eire. The tables all expire in 2038 of course... PS. Forsyth hinted dhog might have had some code to do all this in a dusty corner of the labs venti. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Oct 31 2004 /n/sources/patch/applied/ape-ranlib nop ranlib to simplify porting - easier than hacking dozens of makefiles. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Oct 31 2004 /n/sources/patch/applied/chap-client Added chap client support (factotum had server only). I include chaptest.c which exercises the code. I use it in the latest cifs client daily, but havent ported ip/ppp as I no longer use that. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Oct 31 2004 /n/sources/patch/applied/devmouse This does two things (well, three, but I'd delete the third one). First, it adds "scrollswap" as a ctl for mousectl. The reason is that some mouses we found send the wrong buttons (5/4) instead of the usual ones (4/5) for the wheel. Second (the reason for this patch) it changes buf in mouseread to hold one extra space after the 4th number and puts an space after the 4th number in the read buffer. We noticed (the hard way) that /dev/mouse was not supplying 49 bytes as the man page said. This has nasty effects when sending mouse events through tcp (that does not have message delimiters). I'd either fix this, or change the man page. In any case, I think /dev/mouse is supplying the extra \0 of the string instead of the extra " ", which is weird. The third thing (that I'd probably not apply), is to support "M ..." messages instead of "m ...", to let the user send absolute mouse positions to the mouse. We use this for the usb touch screen, but we're going to convert it to use #m/mousein instead. Hope it's not hard for you to just extract the useful part of this patch. d-rwxrwxr-x M 104260 jmk sys 0 Nov 3 2004 /n/sources/patch/applied/ide5513 it works for me d-rwxrwxr-x M 104260 rsc sys 0 Nov 4 2004 /n/sources/patch/applied/acme-fix acme commits suicide after typing left or right arrows within window-free tag lines d-rwxrwxr-x M 104260 rsc sys 0 Nov 4 2004 /n/sources/patch/applied/rx rx(1) choses 9p, ssh, and bsd protocols in that order, this patch adds support for -l remoteuser and -r (strip CR's) options to the latter two. d-rwxrwxr-x M 104260 rsc sys 0 Nov 4 2004 /n/sources/patch/applied/cpu-manpage About to add a way to restrict the exported namespace in cpu(1), discovered it has been implemented already, so I documented it. d-rwxrwxr-x M 104260 rsc sys 0 Nov 7 2004 /n/sources/patch/applied/diff-rec remove temp files even from recursive call through diffdir (see the not much sense diff /boot /boot) d-rwxrwxr-x M 104260 jmk sys 0 Nov 15 2004 /n/sources/patch/applied/exportfs-man Delete the description of srvfs -S, since there is no such option. d-rwxrwxr-x M 104260 jmk sys 0 Nov 15 2004 /n/sources/patch/applied/fshalt-drain Ensure no trailing input will screw things up d-rwxrwxr-x M 104260 jmk sys 0 Nov 15 2004 /n/sources/patch/applied/nntpfs-groups Some newsgroups seems to have a ' in their name, thus tokenize(..) is not appropriate. d-rwxrwxr-x M 104260 jmk sys 0 Nov 15 2004 /n/sources/patch/applied/faces-dup merge duplicate faces(1) man pages: /sys/man/1/face and /sys/man/1/faces d-rwxrwxr-x M 104260 rsc sys 0 Nov 16 2004 /n/sources/patch/applied/iwhois-more-domains make iwhois cope with more domains d-rwxrwxr-x M 104260 rsc sys 0 Nov 16 2004 /n/sources/patch/applied/acme-delcol Enable to delete the last column consistently in acme since errorwin1 seems to handle the trick. d-rwxrwxr-x M 104260 jmk sys 0 Nov 16 2004 /n/sources/patch/applied/auth-convkeys2 convkeys was mentioned twice. changed the second one to convkeys2 andrey d-rwxrwxr-x M 104260 rsc sys 0 Nov 19 2004 /n/sources/patch/applied/davros I am Davros, my Daleks are my servers and plan9 is my control board mwahahaha - Matt d-rwxrwxr-x M 104260 rsc sys 0 Nov 19 2004 /n/sources/patch/applied/qmail incorporate local simplifications, better error codes, separate note group. d-rwxrwxr-x M 104260 rsc sys 0 Nov 20 2004 /n/sources/patch/applied/smtpd-hide-pass Hide the user/password combination in smtpd's logs. d-rwxrwxr-x M 104260 rsc sys 0 Nov 20 2004 /n/sources/patch/applied/kbmap-german Keyboard map for german keyboards d-rwxrwxr-x M 104260 rsc sys 0 Nov 20 2004 /n/sources/patch/applied/rc-rdwr-man update rc(1) to match the new sources (with <>) d-rwxrwxr-x M 104260 rsc sys 0 Nov 20 2004 /n/sources/patch/applied/rc-rdwr adds <>file and <>[fd]file to rc. it's a natural fit for /net and i've wanted it repeatedly. i've been running this shell for several days and nothing has broken. brucee has these changes and has been looking at them too. d-rwxrwxr-x M 104260 rsc sys 0 Nov 20 2004 /n/sources/patch/applied/roget Support for Roget's thesaurus from project Gutenberg. Datafiles are (temporarly) in /n/sources/steve/roget, roget and rogetindex are the data and index respectively. /sys/src/cmd/dict/mkroget will rebuild build these from roget-body.rtf (with the help of rtf2txt). I suggest the data and index should go in /n/sources/roget similar to /n/sources/pgw. -Steve d-rwxrwxr-x M 104260 jmk sys 0 Nov 21 2004 /n/sources/patch/applied/sed-typo 1 for 2 d-rwxrwxr-x M 104260 rsc sys 0 Nov 21 2004 /n/sources/patch/applied/ap-fenice Hope it will be less prone to site redesign d-rwxrwxr-x M 104260 jmk sys 0 Nov 23 2004 /n/sources/patch/applied/fossil-sleep fix typo d-rwxrwxr-x M 104260 rsc sys 0 Nov 23 2004 /n/sources/patch/applied/malloc-2 fixes malloc(2)'s synopsis of the get*alloctag routines to match the prototypes in /sys/include/libc.h. d-rwxrwxr-x M 104260 jmk sys 0 Nov 26 2004 /n/sources/patch/applied/usbman make formatting of filenames in usb(4) consistent. d-rwxrwxr-x M 104260 jmk sys 0 Nov 29 2004 /n/sources/patch/applied/man-updates while perusing manual d-rwxrwxr-x M 104260 jmk sys 0 Nov 29 2004 /n/sources/patch/applied/fossil-last-man change to the fossil(4) man page to mention the 'fossil/last' tool d-rwxrwxr-x M 104260 jmk sys 0 Nov 30 2004 /n/sources/patch/applied/factotum-precede As long as we're fixing spelling errors and typos, change "preceed" to "precede". It's inconsistent with "proceed", but that's English for you. d-rwxrwxr-x M 104260 rsc sys 0 Dec 1 2004 /n/sources/patch/applied/file-tar added code to file to recognise tar archives by checksum and distinguish posix from non-posix by presence or absence of "ustar" string in the "magic" field. d-rwxrwxr-x M 104260 rsc sys 0 Dec 1 2004 /n/sources/patch/applied/acme-bind Mutual exclusive patches: - the first fixes /acme/bin/README comment, - the last updates man page according to the way acme currently binds its binaries. d-rwxrwxr-x M 104260 rsc sys 0 Dec 1 2004 /n/sources/patch/applied/acme-loadfile acme -l /dev/null d-rwxrwxr-x M 104260 rsc sys 0 Dec 1 2004 /n/sources/patch/applied/tar-posix-exit fix a bug detecting end of an archive when reading a POSIX archive containing a long (>100 byte) directory name. this caused premature exit of tar x or tar t on some POSIX archives. it's my bug; sorry. d-rwxrwxr-x M 104260 rsc sys 0 Dec 4 2004 /n/sources/patch/applied/manpage-fix Fix typo and trofftropism d-rwxrwxr-x M 104260 jmk sys 0 Dec 7 2004 /n/sources/patch/applied/cron-catchup Limit the tasks (1 days worth) to envoke when the system clock moves forward more than this. d-rwxrwxr-x M 104260 jmk sys 0 Dec 7 2004 /n/sources/patch/applied/factotum-chap-ehg Several foolish bugs caught by Eric Grosse -Steve d-rwxrwxr-x M 104260 jmk sys 0 Dec 7 2004 /n/sources/patch/applied/vncv-scroll makes vncv send scroll button events to the other side (it only sent 3 bits worth previously) d-rwxrwxr-x M 104260 rsc sys 0 Dec 8 2004 /n/sources/patch/applied/sokoban-multistep A multi-step-move change to the original Plan 9 sokoban. The new behaviour is only visible in the Button 3 menu (new entry [no]animate), and when you click button 1 on a square not adjacent to glenda - otherwise the behaviour should be as before. This is essentially the same change I already reported to 9fans, only the code has been cleaned up slightly. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Dec 9 2004 /n/sources/patch/applied/newuser-cron Remove redundancy. d-rwxrwxr-x M 104260 rsc sys 0 Dec 9 2004 /n/sources/patch/applied/keyboard-sad an addition to the latin1.h patch, this adds the sad smiley to the list of available keyboard commands in /lib/keyboard andrey d-rwxrwxr-x M 104260 jmk sys 0 Dec 13 2004 /n/sources/patch/applied/assert-man Fix typo d-rwxrwxr-x M 104260 jmk sys 0 Dec 13 2004 /n/sources/patch/applied/my-face my face at home d-rwxrwxr-x M 104260 jmk sys 0 Dec 16 2004 /n/sources/patch/applied/rioattach if a rio attach failed (due to an overly large rectangle, for instance), the fid was still marked as busy, confusing subsequent 9p requests. d-rwxrwxr-x M 104260 jmk sys 0 Dec 18 2004 /n/sources/patch/applied/usb-endpoint-info Put information back into type, dir and isotype fields of Endpt which was removed in the November update. d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2004 /n/sources/patch/applied/mk9660-manpage brought the example up to date. d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2004 /n/sources/patch/applied/acme-fontnames2 This patch supersedes my acme-fontnames patch. It: • Establishes the invariant that fontnames[i] is either nil or allocated with malloc. The current code sometimes does free(static_string) as described in the acme-fontnames patch. That patch leaked memory. This patch should not. • With the old code for rfget, it seems possible that fontnames[fix] could get set to a pointer that had been freed. For example, calling rfget with save true and name nil would do something like "name = fontnames[fix]; ⋯; free(fontnames[fix]); fontnames[fix] = name;". I do not know if this ever happened in practice, but this patch avoids the possibility. • Establishes the convention for rfget that the caller owns rfget's name parameter; in particular, rfget now does estrdup(name) when appropriate and the use of rfget around /sys/src/cmd/acme/rows.c:518 no longer does estrdup. Note that the other uses of rfget are compatible with this convention and if you decide the other convention is better (ie, that rfget should own name and callers should do estrdup), then you need to change the use of rfget around /sys/src/cmd/acme/exec.c:965 to not free(aa). • Changes rfget to erealloc (rather than realloc) fontcache. d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2004 /n/sources/patch/applied/ftpfs-port Not sure if this is the correct fix. ftpfs to a tiny embedded ftp server which doesn't support PASV (rare these days) provokes the use of PORT. PORT replies with the IP address of 0.0.0.0 This is due to the contents of /net/tcp/n/local (after announce()) being ::!29274 (missing IP address). Maybe this is history? Perhaps announce used to give the local address before connection and now either doesn't or only does after the connection is successfull; the port is correct though. My fix is to use myipaddr() to fetch the IP address, which works. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2004 /n/sources/patch/applied/win-errstr Fix per-process errstr mess. d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2004 /n/sources/patch/applied/smell-hackish Exiting from acme while win still running left processes over. d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2004 /n/sources/patch/applied/acme-scrollgap Fill in blank upper left corner. d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2004 /n/sources/patch/applied/tar-abs-names preface only relative names (such as "#foo") with "./". without this fix, tar was unable to create archives using absolute path names. d-rwxrwxr-x M 104260 rsc sys 0 Dec 25 2004 /n/sources/patch/applied/bwk-face bwk at princeton d-rwxrwxr-x M 104260 jmk sys 0 Jan 4 2005 /n/sources/patch/applied/hubdesc i bought a storage key that for some peculiar reason contains an internal hub with the usb storage device attached to that. the internal hub won't accept a get descriptor request giving a length as large as 256. 128 is fine, and since no valid descriptor can be larger than 71 bytes (2*32+7), the change should not break access to any other existing hub. d-rwxrwxr-x M 104260 jmk sys 0 Jan 4 2005 /n/sources/patch/applied/file-elf-bo make the elf cpu test try both byte orders. perhaps bytes 18 and 19 are actually a short? d-rwxrwxr-x M 104260 rsc sys 0 Jan 5 2005 /n/sources/patch/applied/man-cpu-exportfs-fix Fix some sections of linked manual pages aan(8) not aan(1) and the like d-rwxrwxr-x M 104260 rsc sys 0 Jan 5 2005 /n/sources/patch/applied/calendar Feature creap I guess, support for "every friday" and "the first wednesday" in calender for regular jargon-bingo sessions. d-rwxrwxr-x M 104260 rsc sys 0 Jan 6 2005 /n/sources/patch/applied/fossil-walk Fossil crashes on walking .. into a removed directory d-rwxrwxr-x M 104260 rsc sys 0 Jan 6 2005 /n/sources/patch/applied/man-8-secstore-logfile mention the name of log file d-rwxrwxr-x M 104260 jmk sys 0 Jan 6 2005 /n/sources/patch/applied/secstoreman when diffing the latest small change i noticed that secstored's manual page needed more work to format properly and keep to man page conventions d-rwxrwxr-x M 104260 jmk sys 0 Jan 6 2005 /n/sources/patch/applied/xls2txt excel spreadsheet to text converter. d-rwxrwxr-x M 104260 jmk sys 0 Jan 7 2005 /n/sources/patch/applied/trap-uninit at least on the pc, the new trap.c fails to initialise clockintr: 8c -FTVw trap.c warning: trap.c:302 used and not set: clockintr d-rwxrwxr-x M 104260 jmk sys 0 Jan 7 2005 /n/sources/patch/applied/trap-clockintr make Ken happy d-rwxrwxr-x M 104260 jmk sys 0 Jan 11 2005 /n/sources/patch/applied/bmp-read My bad, stupidity in reading colour mapped images. d-rwxrwxr-x M 104260 jmk sys 0 Jan 15 2005 /n/sources/patch/applied/kl-reserved2 A reserved register field in floating point ops is not zerod as it should be. d-rwxrwxr-x M 104260 jmk sys 0 Jan 15 2005 /n/sources/patch/applied/devfs-errmsg trivial change to error message which might have saved me looking at the source. d-rwxrwxr-x M 104260 rsc sys 0 Jan 18 2005 /n/sources/patch/applied/fossilboot this patch courtesy of Latchezar Ionkov: 1. The second argument of the run command on line 235 should be f[0], not venti. 2. If f[1] contains '*' (like in tcp!*!17034), replace the '*' with 127.1 before setting 'venti' to f[1]. d-rwxrwxr-x M 104260 rsc sys 0 Jan 18 2005 /n/sources/patch/applied/readnvram-sdd0 look on /dev/sdD0 too. I've got a cpu server with no magnetic disks and the cfdisk won't fit in the first ide controller's slot (due to overcrowding inside the machine), so it ends up being /dev/sdD0. d-rwxrwxr-x M 104260 rsc sys 0 Jan 23 2005 /n/sources/patch/applied/libmach fix libmach bug, visible under acid. to reproduce the bug: acid /bin/ls acid: filepc("lsxxx:1") acid 1591: suicide: sys: trap: fault read d-rwxrwxr-x M 104260 rsc sys 0 Jan 23 2005 /n/sources/patch/applied/faces-vwhois vwhois and faces assume you have a upas/fs mail dir called mbox, mine are work and home, this makes them agnostic. Also fixed usage. d-rwxrwxr-x M 104260 rsc sys 0 Jan 23 2005 /n/sources/patch/applied/chars-not-bytes I must not assume pointer difference yeilds string length I must not assume pointer difference yeilds string length I must not assume pointer difference yeilds string length I must not assume pointer difference ... d-rwxrwxr-x M 104260 rsc sys 0 Jan 23 2005 /n/sources/patch/applied/mntgen_through_exportfs i believe i mentioned this issue to russ some time ago: when mntgen is used behind exportfs, the mount points get "lost", as exportfs does not hold an open file descriptor for each directory that has been walked to by its clients, so walking to a file below the mount point results in mntgen creating a new mountpoint with a new qid, so "forgetting" the old mount. reproduce the problem with: mkdir /tmp/mnttest mntgen /tmp/mnttestsrvfs srvfs mnttest /tmp/mnttest [in another window] mount /srv/mnttest /tmp bind /dev /tmp/blah ls -l /tmp/blah a deeper fix might change the plan 9 kernel to allow user processes to hold file descriptors that are not yet open (but may be walked). that fix doesn't seem so likely in the near future (and it's not a great problem for anything else), so this fix changes mntgen so it always generates the same qid (60 bits of md5sum) for a given mountpoint name, thus avoiding the problem. it's a small enough change that i hope it can be accepted for the time being, even if it is only an interim solution to a deeper problem... d-rwxrwxr-x M 104260 rsc sys 0 Jan 23 2005 /n/sources/patch/applied/libdraw-cf2subfont fix libdraw cf2subfont bug. depth var was equal to zero, and subfontname function was unable to locate subfont file. to reproduce the bug: scat moon moon .... plot stringwidth: bad character set for rune 0x0000 in /lib/font/bit/lucida/unicode.6.font ... d-rwxrwxr-x M 104260 rsc sys 0 Jan 28 2005 /n/sources/patch/applied/tftpd-sun Accept both and .SUN as sun boot filenames. d-rwxrwxr-x M 104260 rsc sys 0 Jan 28 2005 /n/sources/patch/applied/bgmap-rename please rename 'bgmap' to 'bg' so it will be in tune with all the rest :) andrey d-rwxrwxr-x M 104260 rsc sys 0 Jan 28 2005 /n/sources/patch/applied/kfs-dev-exists 9fs kfs on a diskless terminal causes disk/kfs to suicide when it fails to open the device. This patch adds a check. I cannot see how this will prevent kfs standalone machines from booting but somone who knows more should think about it. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Feb 3 2005 /n/sources/patch/applied/localboot sense of test for tcp!*!17034 was reversed d-rwxrwxr-x M 104260 rsc sys 0 Feb 3 2005 /n/sources/patch/applied/pop3resp This patch eliminates a line length limit in pop3resp. It changes pop3resp to use Brdstr rather than Brdline. The patched pop3resp still takes care of memory management for its callers: A line is freed when the next line is read. The limit causes upas/fs to blow up with "unexpected eof" and "unexpected end of conversation" when unreasonable messages exist on a POP server. d-rwxrwxr-x M 104260 rsc sys 0 Feb 3 2005 /n/sources/patch/applied/login-nopw - Allow login to be run on the cpu server, but with warning. - Allow machineowner to change user without a password. d-rwxrwxr-x M 104260 jmk sys 0 Feb 3 2005 /n/sources/patch/applied/kernel-devproc echo > /proc/ ^ `{cat /dev/ppid} ^ /noteid => system crash problem appears when any garbage will sent to noteid. in turn, id=atoi(a) will become 0, loop will find first zeroed proc entry and strcmp will kill system. However, one doubt still here: atoi call. If n equals to zero, it will dereference pointer a. d-rwxrwxr-x M 104260 jmk sys 0 Feb 3 2005 /n/sources/patch/applied/pragincomplete this is the patch i mentioned several weeks ago. it extends #pragma incomplete to allow -T to be switched off and on for whole libraries (eg, in the Freetype include files), mainly intended for use when the code is imported and so constructed that it is hard to add all required `incomplete' names. d-rwxrwxr-x M 104260 jmk sys 0 Feb 3 2005 /n/sources/patch/applied/gmt-rfc822 tz>0 to select + or "" in dates doesn't work if you're actually in GMT. a nearby Spam filter decided messages were spam because the date wasn't RFC822 format. i wonder who writes such stuff. anyhow, probably should change it to tz>=0 i couldn't find any other instances d-rwxrwxr-x M 104260 jmk sys 0 Feb 3 2005 /n/sources/patch/applied/unlnfs the lnfs(4) man page refers to unlnfs, but it doesn't appear to be distributed. i've included a version i wrote in case you haven't got it. (i don't know whether the version referred to was mine, as i vaguely recall posting it, or someone else's. it's trivial anyway). d-rwxrwxr-x M 104260 jmk sys 0 Feb 5 2005 /n/sources/patch/applied/kernel-devpccard I was try to install cardbus pccard with realtek 8139 chip on board. Card was detected fine, but does not work: #l0: rtl8139: 10Mbps port 0x400 irq 11: FFFFFFFFFFFF pci output: ... 9.0.0: 02.00.00 10ec/8139 11 0:00000401 256 1:04000000 512 ... xd /dev/pci/9.0.0raw output: 0000000 ec103981 04009002 10000002 00400000 ... as You can see - memory and io on bus are disabled: On boot, original devpccard.c correctly enable bus master, memory and io, but leave pci->pcr don't touched. in my case it was 0. pcisetbme call from ether8139.c leave pccard out of luck. P.S. cls and ltr registers are filled for consistency too. d-rwxrwxr-x M 104260 rsc sys 0 Feb 5 2005 /n/sources/patch/applied/libmach-elf2 Russ, Here's the updated patch. I made the elfdotout function more general so that it could handle cracking ELF headers for any architecture and added in support for POWER and MIPS (since ql and vl both generate ELF32 for -H5). This works with ql and vl emitted ELFs. The section search could be stricter (ie. right now the data section can have the X bit set), but the ql elf code spits out data with RWX. d-rwxrwxr-x M 104260 rsc sys 0 Feb 5 2005 /n/sources/patch/applied/new_kbmap Ukrainian keyboard map d-rwxrwxr-x M 104260 rsc sys 0 Feb 5 2005 /n/sources/patch/applied/rus_kbmap Russian keyboard map d-rwxrwxr-x M 104260 rsc sys 0 Feb 8 2005 /n/sources/patch/applied/upasfs-russian-charset add russian charset support d-rwxrwxr-x M 104260 rsc sys 0 Feb 8 2005 /n/sources/patch/applied/keyfs fix some minor bugs: bug0 (user entry encrypted with empty key and decrypted with different one - uninitialized authkey used) ---------------------------------------- % cd % rm -f keys % auth/keyfs -p keys % cd /mnt/keys % mkdir user Password: keyfs: warning: bad status in key file 1 keys read bug1 (user entries encrypted with different keys - no difference exists between empty and uninitialized authkey) ---------------------------------------- % cd % echo -n > keys % auth/keyfs -p keys % cd /mnt/keys % mkdir user Password: 1 keys read % mkdir user2 Password: keyfs: warning: bad status in key file keyfs: warning: bad status in key file 2 keys read bug2 (suicide on invalid status string) ---------------------------------------- % cd % echo -n > keys % auth/keyfs -p keys Password: 0 keys read % cd /mnt/keys % mkdir user 1 keys read % cd user % echo die, babe > status keyfs 2701: suicide: sys: trap: fault read addr=0x0 pc=0x00003986 bug3 (walk to ".." from user directory not implemented) ---------------------------------------- % cd % auth/keyfs -p keys Password: 1 keys read % cd /mnt/keys/user % cd .. Can't cd ..: '..' file not found bug4 (Remove does not clunk fids in some cases) ---------------------------------------- % pwd /mnt/keys/user % rm key key rm: key: permission denied rm: key: 'key' fid in use bug5 (memory leak) ---------------------------------------- allocated at installuser: 'u->name = strdup(name);', Wstat leaked at Clunk d-rwxrwxr-x M 104260 rsc sys 0 Feb 8 2005 /n/sources/patch/applied/zipfs zipfs a la tarfs. Allows me to vac my old zip archives int venti with their correct file and directory dates. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Feb 9 2005 /n/sources/patch/applied/upas-fs-nil-test the test at fs.c:1233 is too late: mb->d->qid.path will already have been dereferenced. i've had fs break due to this. d-rwxrwxr-x M 104260 rsc sys 0 Feb 9 2005 /n/sources/patch/applied/64-bit-mkext-mkfs make mkfs and mkext use 64-bit file sizes and seek offsets. d-rwxrwxr-x M 104260 rsc sys 0 Feb 9 2005 /n/sources/patch/applied/right-patch-name add rfork e; make create print the right patch name when requesting that the user run patch/email. d-rwxrwxr-x M 104260 rsc sys 0 Feb 9 2005 /n/sources/patch/applied/keep-mkfs-sync make mkfs cope with growing or shrinking files in the source tree by copying exactly the number of bytes it claimed in the archive header it wrote (as tar does), padding or truncating as necessary. d-rwxrwxr-x M 104260 rsc sys 0 Feb 9 2005 /n/sources/patch/applied/ld-4-ape first attempt at ld for ape. Linux code configure :-) Linux code :-( Linux #*%@#$@! -Steve d-rwxrwxr-x M 104260 rsc sys 0 Feb 11 2005 /n/sources/patch/applied/strip-consts Strip has constants hardwired for the minimum and maximum magic numbers. This patch moves the constants into the header. d-rwxrwxr-x M 104260 rsc sys 0 Feb 12 2005 /n/sources/patch/applied/better-errmsg added %r to cannot open file, handy for some of my more weird file servers where you can get some helpfull info from lower layers, eg p: cannot open 'junk' - cifs: server disconnected -Steve d-rwxrwxr-x M 104260 rsc sys 0 Feb 12 2005 /n/sources/patch/applied/macbody-cheese Fix (2nd time) round for: #define S "//foo" #define S "/*foo" #define S(s) "s" d-rwxrwxr-x M 104260 rsc sys 0 Feb 15 2005 /n/sources/patch/applied/ether-case Additions to warn that ndb ether addresses must be abcdef and not ABCDEF. d-rwxrwxr-x M 104260 rsc sys 0 Feb 20 2005 /n/sources/patch/applied/smtp-meta If you are not Russ, please let him do this one. This fixes the expand_meta for both auth (so factotum gets the expanded relay's name) and for the dns lookup for the mx record. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Feb 20 2005 /n/sources/patch/applied/pngupdate a major update. support added for all sample depths (1-16), greyscale and indexed images. interleaved images are supported as well. d-rwxrwxr-x M 104260 rsc sys 0 Feb 21 2005 /n/sources/patch/applied/jpgpngtorgbv colormap doesnt change to rgbv after the first image when doing png *.png, for example. torgbv now understands about crgb24 &c. d-rwxrwxr-x M 104260 jmk sys 0 Feb 25 2005 /n/sources/patch/applied/malloclong sizeof knows better than earthling d-rwxrwxr-x M 104260 rsc sys 0 Feb 27 2005 /n/sources/patch/applied/robs-bits-2 wrote manual pages changed lock to $argv0 in screenlock generated a re-sized bunny 30% of big jpg on sources added initial set screen to black to blend with glanda's background global var chatty which disables password prompt. Really we need somthing like a dialogue box to prevent the password entry messing up the picture. Rather than this I just made the decision to drop the prompt, its easily restored if you think this is a step too far. [myself, I prefered 'lock' as a name] -Steve d-rwxrwxr-x M 104260 rsc sys 0 Feb 27 2005 /n/sources/patch/applied/n97_kbmap kbmap file for NMB N-97 keyboard; commonly used by NCD X-terminals. only works under VMware. native kernel fails to initialize the kbd. d-rwxrwxr-x M 104260 rsc sys 0 Feb 27 2005 /n/sources/patch/applied/replica-mkdir-p replica gets confused when it tries to change files in a directory that doesn't exist. using mkdir -p instead of just mkdir solves that problem, i think. andrey d-rwxrwxr-x M 104260 rsc sys 0 Feb 27 2005 /n/sources/patch/applied/kbmap-norwegian kbmap for norwegian keyboards. d-rwxrwxr-x M 104260 rsc sys 0 Mar 1 2005 /n/sources/patch/applied/yesterday-usage d-rwxrwxr-x M 104260 rsc sys 0 Mar 1 2005 /n/sources/patch/applied/history-man d-rwxrwxr-x M 104260 rsc sys 0 Mar 1 2005 /n/sources/patch/applied/grumble-on-dismount2 Get rid of the 'mount read: i/o on hungup channel' errors when unmounting some common file servers tapefs now without wstat() patch. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Mar 4 2005 /n/sources/patch/applied/con-baud Handle empty argument to -b a little better and add it to usage() d-rwxrwxr-x M 104260 rsc sys 0 Mar 4 2005 /n/sources/patch/applied/acme-wiki-embedded-nl Don't become confused when an on-wiki link [contains a newline]. The fix assumes that each newline in the link represents a space in the actual page name. This assumption seems to be fine. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Mar 12 2005 /n/sources/patch/applied/man-manpage punctuation error: "function's" -> "functions" d-rwxrwxr-x M 104260 rsc sys 0 Mar 13 2005 /n/sources/patch/applied/keys.who-manpage Manpage for /adm/keys.who d-rwxrwxr-x M 104260 rsc sys 0 Mar 17 2005 /n/sources/patch/applied/mkfiles-man due to the many external programs that come to plan9 nowadays i've added a 'man' target for installing external man pages located in the same directory as the source. the command is common to the 'mkone' and 'mkmany' include files and works simply by: - the source code writer creates one or several man pages named '$TARG.man' - the source code writer defines MAN= in the mkfiles for their code, $MAN points to the section the man pages are destined for (i.e. /sys/man/1/ for section 1, commands) - the user types 'mk man' or 'mk somefile.man' to have the man page installed. for now, the 'man' target is not included in 'mk all'... d-rwxrwxr-x M 104260 rsc sys 0 Mar 17 2005 /n/sources/patch/applied/replica-fcp incorporate the new fcp command (and code) into replica to facilitate faster pulls from slower connections. this setup has been running on plan9.ucalgary.ca and a couple of other places for the past week without any significant issues. d-rwxrwxr-x M 104260 rsc sys 0 Mar 18 2005 /n/sources/patch/applied/acmescroll patch for some acme scrolling issues that've been bugging me for years. 1) when middle-button scrolling to the bottom of a long document, acme doesn't go right to the end - there's a bit more of the document left below. 2) acme constrains the mouse to remain within the boundaries of the scrollbar. i find this unnecessary and sometimes unhelpful. 3) clicking with b1 or b3 on the very top line's-worth of the scrollbar does nothing, meaning that it's essentially dead space, and also that it's awkward to scroll very small (one or two line) windows. one gratuitous change: a piece of code from optionally using fixed-point arithmetic to use a vlong; it looks simpler that way and the overhead is negligible by my reckoning. YMMV. d-rwxrwxr-x M 104260 jmk sys 0 Mar 18 2005 /n/sources/patch/applied/8ljpc JPC should be 7b/8b not 76/86 (b and 6 look similar i suppose). it doesn't matter much because parity is hardly ever checked this way... just happened to notice it when doing some other things. d-rwxrwxr-x M 104260 jmk sys 0 Mar 18 2005 /n/sources/patch/applied/profilev profile was calling vlong run-time support directly which is counter-productive when those operations are done by the compiler and might not even exist in the library. i've used A = A+B rather than A += B to ensure the generated code is the same (ie, calls to _addv and _subv) on those platforms that still use them. (those functions are handled specially for profiling purposes in vlop.s on some platforms.) it's probably time to put a bit more intelligent vlong support into the portable part of cc, but this change will allow alpha, amd64 and others to cope until i have a chance to look at that (eg, for powerpc). d-rwxrwxr-x M 104260 rsc sys 0 Mar 24 2005 /n/sources/patch/applied/doc-man-4-wikifs-pp Add missing .PP to wikifs(4); this is a small fix to test the procedure we will use for the manpages review. d-rwxrwxr-x M 104260 rsc sys 0 Mar 24 2005 /n/sources/patch/applied/acme-look This patch corrects "Edit [erw] filename" errors when "filename" is unrooted. Under this condition the effective file name becomes the entire tag line. (Button 3 clicking on an unrooted file name in the tag line also fails, but silently.) The propsed change to dirname() explicitly sets a terminating null for the unrooted file name case. d-rwxrwxr-x M 104260 rsc sys 0 Mar 24 2005 /n/sources/patch/applied/applylog-dometadb cd >db replica/updatedb -p /sys/lib/sysconfig/proto/usrproto -r /root db chmod -x tmp replica/updatedb -p /sys/lib/sysconfig/proto/usrproto -r /root db >>log replica/applylog -c /dev/null /root /root /n/newfs * kfs support added * ext2 support is commented out, because ext2srv does not fit onto the floppy. * fossil mountdist support works * on mountdist, the bins of it are binded into the /bin of the floppy * fmtventi does bind its binaries in /bin/venti * some /n paths are removed * the $wsys error is now fixed in textonly install * fixed pcflop to include loopback device d-rwxrwxr-x M 104260 rsc sys 0 Apr 1 2005 /n/sources/patch/applied/doc-man-8-newuser-usefossil - give hints for using fossil now that it's the default/recommended filesystem - be more clear about /mail/box/$user and /cron/$user/cron - give useful example - /sys/lib/newuser: start plumber and factotum for drawterm users (also changed in man page) d-rwxrwxr-x M 104260 rsc sys 0 Apr 1 2005 /n/sources/patch/applied/doc-man-1-winwatch-wasmissing - manual page for winwatch (was missing) - winwatch.c: complain if unrecognized options are used (it already complains at useless arguments) d-rwxrwxr-x M 104260 rsc sys 0 Apr 1 2005 /n/sources/patch/applied/acme-wiki-diff-fixes Fri Apr 1 04:06:16 EST 2005 -- uriel Includes the previous acme-wiki-diff-author-date; and fixes the order of the diff params so changes are not shown reversed and removes a useless diff execution. d-rwxrwxr-x M 104260 rsc sys 0 Apr 3 2005 /n/sources/patch/applied/ncr-debug Remove an irritating debug message d-rwxrwxr-x M 104260 rsc sys 0 Apr 3 2005 /n/sources/patch/applied/acme-wiki-diff-fix-2 Sat Apr 2 06:44:10 EST 2005 -- uriel Fix incorrect change attributions, sorry for sending so many small patches, should have tested it more carefully before. d-rwxrwxr-x M 104260 rsc sys 0 Apr 3 2005 /n/sources/patch/applied/slimsam remove extra code d-rwxrwxr-x M 104260 rsc sys 0 Apr 3 2005 /n/sources/patch/applied/installer-without-kfs Good, the patch is there. To be removed files: moveoldfs, iskfs, mountkfs d-rwxrwxr-x M 104260 rsc sys 0 Apr 4 2005 /n/sources/patch/applied/ftpfs-auth-tls This patch adds the option -t to ftpfs, which causes it to use AUTH TLS over the control and data connection. d-rwxrwxr-x M 104260 jmk sys 0 Apr 5 2005 /n/sources/patch/applied/t-mode-for-all -T mode - function call parameter checking passthru for pcc and ape/cc d-rwxrwxr-x M 104260 rsc sys 0 Apr 8 2005 /n/sources/patch/applied/file-png Makes file(1) to recognize PNG image files d-rwxrwxr-x M 104260 rsc sys 0 Apr 8 2005 /n/sources/patch/applied/acmemail-png Make Acme Mail to send attachments of type image/png to the plumber by right-clicking d-rwxrwxr-x M 104260 rsc sys 0 Apr 9 2005 /n/sources/patch/applied/jukefs-mapname This is a fix of jukefs, where there could have happen a reread() on a static buffer, which Plan9 does not allow. This patch was done in a distributed manner on IRC, with some beer, red wine and across four countries. d-rwxrwxr-x M 104260 rsc sys 0 Apr 9 2005 /n/sources/patch/applied/date-sender garbage d-rwxrwxr-x M 104260 rsc sys 0 Apr 9 2005 /n/sources/patch/applied/plan9.ini-nvram documented nvram= nvrlen= and nvroff= in plan9.ini d-rwxrwxr-x M 104260 rsc sys 0 Apr 12 2005 /n/sources/patch/applied/wikifs-hr-rellinks-html-fixes Sun Apr 10 21:36:46 GMT 2005 -- uriel Various fixes and improvements to wikifs: - Handle https: links properly - Handle #relative links, this is really needed for the FAQ page and others, it works only in the web interface but I'm working on support for it in Wiki, but it's not a big deal anyway. - Convert lines of the form /^-----*$/ to
in html and to an standard number of - in acme; this is very useful and will make the "Tip'o Day" page look much better right away. - Use real headers instead of hacked up font hacks, also generate anchors for them so they can be references with #rel links. This is the first step of the sanitization of the wiki HTML. d-rwxrwxr-x M 104260 rsc sys 0 Apr 12 2005 /n/sources/patch/applied/acmewin-paste if someone pastes in an acme window but cancels the paste win will die with 'reading data' error, this patch makes handling of the 'canceled paste' without dying. d-rwxrwxr-x M 104260 rsc sys 0 Apr 12 2005 /n/sources/patch/applied/install-mbr-question This is a patch for adding a question to the user in bootplan9, whether he wants to install the Plan9 MBR or not. The existing mechanism only asked when the Plan 9 partition was 2Gb into the disk. Now, when this happens, some hints about it are printed additionally to the main question. d-rwxrwxr-x M 104260 rsc sys 0 Apr 12 2005 /n/sources/patch/applied/winwatch-exclude A small patch to winwatch, to exclude certain windownames from being displayed. So winwatch -e '^(winwatch)' does not display winwatch itself. The code was done by Nils O. Selåsdal and the manpage by Christoph Lohmann <20h@r-36.net>. d-rwxrwxr-x M 104260 rsc sys 0 Apr 12 2005 /n/sources/patch/applied/drawterm-win Recompiled for arrow keys and -e. d-rwxrwxr-x M 104260 rsc sys 0 Apr 12 2005 /n/sources/patch/applied/winwatch-zerowin Winwatch never was any happy with showing 0 windows. (division by zero error) Sorry about that. d-rwxrwxr-x M 104260 rsc sys 0 Apr 12 2005 /n/sources/patch/applied/file-mp3-order correct byte order of raw-mp3 magic number (to be little-endian). d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/scuzz-big-exabyte crank MaxIOsize from 32k to 64512 (126*512) to permit reading exabytes tapes written with `tar b 126'. it might make more sense to get rid of the MaxIOsize limit altogether. d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/runq-keep-lock-alive if you have a slow upstream link, mailing multi-megabyte files can take minutes, and the C. DMEXCL lock can go stale, and running the queues can cause other smtps to start up trying to resend it. presotto acknowledged that this is a potential problem and that another process would be needed to keep the lock alive, but then left the labs. this fix recycles existing upas machinery to start a process to keep the lock alive until the file is sent. d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/isatty Fix various functions testing whether a fd looks like it is pointing to a console. Converts dirstat aproaches to checking the tail of fd2path. d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/9660srv-qttmp i usually put iso images on a "chmod +t"-ed directory. 9660srv refuses to mount such an image because it's "not a plain file". well, it is just a plain file with QTTMP bit on.d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/memdraw-type Fixes the two appearances of "#include " in the top. d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/libmemdraw-typo This patch fixes the changed parameters of memimagestring(), where the last one called drawop is in the manpage, but not in the source (/sys/src/libmemdraw/string.c). d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/tar-seek when printing a table of contents or skipping an archive member because it didn't match any of the command-line patterns, seek past large archive members when the archive is on a seekable medium. this is a big time-saver after archiving large files (such as other tar archives), when you use `tar t' to generate a table as contents as quick verification that everything got written into the archive. d-rwxrwxr-x M 104260 rsc sys 0 Apr 17 2005 /n/sources/patch/applied/doc-man-1-p9psync sync changes to p9p section 1 man pages back to plan9. mostly minor changes to reduce diff size. some files have not been touched. e.g. acid, calendar, colors, db, gview, jpg, kill, page, rio some files got a previously missing SOURCES section e.g. replica spin lp troff2html faces and one or two more d-rwxrwxr-x M 104260 jmk sys 0 Apr 23 2005 /n/sources/patch/applied/incl-uobj Reserve a constants for sparc64 machine files. d-rwxrwxr-x M 104260 rsc sys 0 Apr 25 2005 /n/sources/patch/applied/dist-smallfixes - cleaner clean in mkfile - dont bomb out of profile if no mouse is present d-rwxrwxr-x M 104260 jmk sys 0 Apr 26 2005 /n/sources/patch/applied/libmach-uobj-more Additional header files required by the previous patch. Additionally the referenced to u.out.h in uobj.c should be changed to use xc/u.out.h. d-rwxrwxr-x M 104260 jmk sys 0 Apr 26 2005 /n/sources/patch/applied/libmach-uobj libmach support for sparc64 .u and elf boot images. - mkfile,obj.c,setmach.c - glue - elf.h,executable.c - boot image support - u.c,udb.c,uobj.c - new bits, borrowed heavily from kobj.c et al d-rwxrwxr-x M 104260 jmk sys 0 Apr 26 2005 /n/sources/patch/applied/sparc64-acid Debugger defines for sparc64. d-rwxrwxr-x M 104260 rsc sys 0 Apr 28 2005 /n/sources/patch/applied/9load-mbrpart fold MBR/ext table reading loops together. Tested against s1,s2,s3,s4,s5 partitions created by disk/fdisk. d-rwxrwxr-x M 104260 rsc sys 0 Apr 28 2005 /n/sources/patch/applied/acmemail-errormsgs 1. cleaner error messages (no more fmtstr in acme). 2. install Mail in /acme/bin/$objtype so acme can find it d-rwxrwxr-x M 104260 rsc sys 0 Apr 28 2005 /n/sources/patch/applied/doc-man-1-nedmail-noupasfs remove upas/fs from the nedmail man page (the exact same text+extra is already in upasfs(4)). make it more clear that "nedmail -c dir" does nothing more than creating. sync usage in manpage with usage in binary. d-rwxrwxr-x M 104260 rsc sys 0 May 2 2005 /n/sources/patch/applied/tbl-argcfix It slightly annoyed me that tbl -h suicided. Try to prevent that. And while at it try to make -ms et al work. d-rwxrwxr-x M 104260 rsc sys 0 May 2 2005 /n/sources/patch/applied/pop3-header-loop Prevent infinite loop in upas/pop3 when reading a message header with a line more than 8kbytes long. d-rwxrwxr-x M 104260 rsc sys 0 May 2 2005 /n/sources/patch/applied/9660srv-apple-padding Hybrid iso9660/HFS+ CDs made by the MacOS Finder can have alignment padding between entries in the directory System Use area, which as far as I can see is contrary to the spec. This patch allows 9660srv to read such directories correctly instead of going into an infinite loop. d-rwxrwxr-x M 104260 rsc sys 0 May 6 2005 /n/sources/patch/applied/dont-clear-the-screen-so-fast this is my current main.c. the main changes are longer delay in shutdown() if #e/*debug is present; one-minute delay at end of shutdown() when a terminal panics; fiddled the computation of kpages, apparently to give kernels on machines >64MB more memory. i now also print the amount of memory at boot, as V6 Unix did, but that's commented out here. d-rwxrwxr-x M 104260 rsc sys 0 May 6 2005 /n/sources/patch/applied/exportfs-abugs This is a patch for exportfs and its manpage. It fixes the issue in exportfs, that if you specify -n (no none) or -N (other namespace) after -a (authenticate), that these options have no effect. This was due to the fact that -a was handled in and not after the argument processing. In the manpage the wrong stated '-n' is corrected to be the real '-N' and '-n' is now documented. d-rwxrwxr-x M 104260 jmk sys 0 May 9 2005 /n/sources/patch/applied/ez-ether-for-dell recognise vid 8086 did 1050 for the intel 82562EZ, which appears in some Dell systems. d-rwxrwxr-x M 104260 jmk sys 0 May 10 2005 /n/sources/patch/applied/i83546eb adds support for the i83546eb dual copper card (did 0x1010). basically follows the codepath for i83546gb. andrey d-rwxrwxr-x M 104260 rsc sys 0 May 15 2005 /n/sources/patch/applied/unicode.13.font_and_jis24 jis24 is added to these fonts. italic and bold shows the normal japanese font, though. d-rwxrwxr-x M 104260 rsc sys 0 May 15 2005 /n/sources/patch/applied/unicode_fonts add jis16 and jis24 to these fonts. d-rwxrwxr-x M 104260 rsc sys 0 May 15 2005 /n/sources/patch/applied/lucida.12-16 I found that jis24 is good enough in these fonts, too. d-rwxrwxr-x M 104260 rsc sys 0 May 15 2005 /n/sources/patch/applied/page-pdf-crop Correct problem with page(1) reported by Peter Cejchan, which turned out to be incorrect handling of (rare) PDF files with a CropBox inset within the MediaBox. The accompanying demo file croptest.ps (which I don't expect you to copy to sources) illustrates the effect: it prints text in four orientations, translated to each quadrant of the page - 16 pages in all. When translated with ps2pdf and viewed with page, the result should be cropped to the correct quadrant and displayed with text upright. d-rwxrwxr-x M 104260 jmk sys 0 May 16 2005 /n/sources/patch/applied/more-gbe-buffering allocate longer queues for gigabit interfaces d-rwxrwxr-x M 104260 rsc sys 0 May 17 2005 /n/sources/patch/applied/install-ventifmt You had to choose two times an "index" in the text displayed. Now it asks for "arenas" and "index". d-rwxrwxr-x M 104260 rsc sys 0 May 18 2005 /n/sources/patch/applied/doc-man-3-usb add SOURCE section to usb(3) d-rwxrwxr-x M 104260 rsc sys 0 May 18 2005 /n/sources/patch/applied/install-halt-nokfscmd Remove the section that runs kfscmd halt, as kfscmd no longer exits. d-rwxrwxr-x M 104260 rsc sys 0 May 19 2005 /n/sources/patch/applied/xls2txt-fixed fixed xls2txt to support rog's file: I didn't support intra string font changes. and lucio's: FP overflow due to a dropoff. Also included a kludge for better support collum widths - this is not well documented and the previous "more correct" code fails on some excel doucuments. d-rwxrwxr-x M 104260 rsc sys 0 May 19 2005 /n/sources/patch/applied/install-spelling This patches fixes some bad style sentences, an error in "halt", because of kfs, some bad spelled sentences and the GB form of "Initializing" in disk/format. Thanks to Vester Thacker for pointing this out. Still one problem is there: the description of mountfs, stop and finish are placed wrong in the menus because of their short names. I don't see a real solution for this. d-rwxrwxr-x M 104260 rsc sys 0 May 19 2005 /n/sources/patch/applied/doc-man-1-mlmgr-replyto The -r option of ml isn't yet mentioned. (for customizing the reply-to in outgoing mails) d-rwxrwxr-x M 104260 rsc sys 0 May 20 2005 /n/sources/patch/applied/correct-spelling correct spelling of error messages. d-rwxrwxr-x M 104260 rsc sys 0 May 20 2005 /n/sources/patch/applied/doc-man-4-keyfs-purgatory After too many failed authentication attempts: 'Purgatory' instead of 'hell'... ...temporarily inaccessible key-file instead of disabled account d-rwxrwxr-x M 104260 rsc sys 0 May 22 2005 /n/sources/patch/applied/art-timezone ART timezone d-rwxrwxr-x M 104260 rsc sys 0 May 22 2005 /n/sources/patch/applied/patch-create-without-consctl Check if /dev/consctl is writeable so patch/create works inside acme's win. d-rwxrwxr-x M 104260 rsc sys 0 May 22 2005 /n/sources/patch/applied/schtarb-2nd This time strecpy is used. In awd.c is no smprint used, because I don't see a reason why there should be something alloced for such a simple thing. d-rwxrwxr-x M 104260 jmk sys 0 May 26 2005 /n/sources/patch/applied/doc-man-2-thread-chanrecv CHANRECV -> CHANRCV d-rwxrwxr-x M 104260 jmk sys 0 May 26 2005 /n/sources/patch/applied/doc-man-8-newuser-fossilcons-ref fossilcons(4) -> fossilcons(8) d-rwxrwxr-x M 104260 rsc sys 0 May 26 2005 /n/sources/patch/applied/lib-ndb-auth-typo auth is now in section 8 of the man pages d-rwxrwxr-x M 104260 rsc sys 0 May 30 2005 /n/sources/patch/applied/juggle-color-revised [revision of patch juggle-color -- this patch replaces juggle-color, hopefully slightly easier to maintain diskinit/addisk] the balls now have color, using the color scheme of stats(8). this makes it easier to follow the balls - and it is more pretty too. Axel. d-rwxrwxr-x M 104260 jmk sys 0 Jun 1 2005 /n/sources/patch/applied/sudoku-resize bug in resize code was causing the window to be unhide-able. andrey d-rwxrwxr-x M 104260 jmk sys 0 Jun 7 2005 /n/sources/patch/applied/southbridge-vt8237 Add vid/did for VIA vt8237 southbridge. d-rwxrwxr-x M 104260 jmk sys 0 Jun 7 2005 /n/sources/patch/applied/juggle-drawing simplify drawing of juggled balls, consume less memory and fix an artifact when balls overlap (white angles) d-rwxrwxr-x M 104260 jmk sys 0 Jun 7 2005 /n/sources/patch/applied/man-games-sokoban-b1-juggle-pattern document sokoban multi-step moves and pushes; add impressive juggle pattern (need more of these!) Axel. d-rwxrwxr-x M 104260 jmk sys 0 Jun 16 2005 /n/sources/patch/applied/createuser-unused-usage Remove from usage msg the -m flag that is not implemented or documented. Probably a typo. d-rwxrwxr-x M 104260 jmk sys 0 Jun 16 2005 /n/sources/patch/applied/hget-setoffset-postbody Change usage string so that it matches the hget(1) manual page. Move the printing of u->postbody after the printing of cookies, since cookies must appear in the header. Add indentation to hhcrange(). Change setoffset() so that it sets out->written, and seeks to the appropriate file offset if the -o option is used. Without this, the assert(d >= 0) in output() fails and the output becomes garbage if the file was already partially downloaded. d-rwxrwxr-x M 104260 jmk sys 0 Jun 16 2005 /n/sources/patch/applied/4s-leftborder-whitening fix the problem where the long narrow yellow brick in vertical position erases the left border. It did this, because it was defined as a 2 column 4 row high brick, where the leftmost column was entirely empty (i.e. white). I have no clue why the brick was defined this way. Redefining the brick to be single column 4 row high seems to work and not have hidden repercussions. There may be a deeper fix that ``ignores'' the entirely white column (or masks it away), but I leave that for a more expert fixer. Axel. d-rwxrwxr-x M 104260 jmk sys 0 Jun 16 2005 /n/sources/patch/applied/games-xs-fatal-warp-blink replaced print/exits combinations by sysfatal, to overcome 'unusable' window resulting from fatal error. (experimental) warp the mouse to the middle of the window for each new brick that starts dropping - to have sufficient space left and right to move the falling brick. (experimental) replaced the white/black blinking when a row disappears by fading to white - looks nicer. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Jun 29 2005 /n/sources/patch/applied/ssh-suicide ssh suicides if the remote end disconnects in the middle of protocol negoitation. The hooks are all there to prevent this but where not used - perhaps the author was interrupted during the implementation? :-) -Steve d-rwxrwxr-x M 104260 jmk sys 0 Jul 5 2005 /n/sources/patch/applied/games-xs-responsive-suspend-cursor varous fixes/extensions to games/^(4s 5s): - be (more) responsive (to keyboard and reshape) in paused state (during blinking, between drop and appearance of next brick) - q/Q always quits (also in suspended state) (previously, to quit in suspended state you needed two q's: one to switch to non-suspended state, after which next one would quit) - also Del always quits - also p/P enables suspended state - moving mouse cursor out of resp. into playing field enables resp. disables suspended state - visual indication of suspended state (faded color, mouse cursor) - replace explicit flushing of channels by having a separate process that only passes on key events in non-suspended state, and keep mouse and 'previous mouse' info always up-to-date, to avoid the case where a late mouse click to lower one brick somehow gets 'delayed' and causes instantanious drop of the subsequent one, as soon as it appears after the pause - tweaked mouse warping a bit more. now also warp when leaving suspended mode. the hope is that by warping the mouse (horizontally) we can always keep it within the bounds of the board. previously, it might happen that the brick would be at the far right, and the mouse cursor at the far left, such that it would not be possible to move the mouse far enough left to move the brick entirely to the left. Axel. d-rwxrwxr-x M 104260 jmk sys 0 Jul 5 2005 /n/sources/patch/applied/dnsquery-whitespace trim leading and trailing whitespace from queries passed to dnsquery - perhaps I am the only one that has problems with this due to my inaccurate mouseing? -Steve d-rwxrwxr-x M 104260 jmk sys 0 Jul 5 2005 /n/sources/patch/applied/mimetype-pgp-encrypted an entry for application/pgp-encrypted no need for vf(1) to touch attachments of this type d-rwxrwxr-x M 104260 rsc sys 0 Jul 7 2005 /n/sources/patch/applied/httpd-webls-clickable-path - add hyperlinks to the path components in the ``Index of /foo/bar/baz'' line, (except for the part representing current directory) if going there is allowed (i.e hyperlinks for ``/'', ``foo/'' and ``bar/'' but not for ``baz'') This also supports the case when the root ("/") is readable. - avoid adding a ``Go to parent directory'' line if target == current dir (i.e. if current dir == '/') Right now giving empty string for the dir in /magic/webls?dir= gives a directory of ``.''. One _might_ consider adding at the start of :/^dols if(strcmp(dir, "") ==0) dir = "/"; (i.e. before the ``cleanname(dir);'' statement) to normalize on "/" as the root. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Jul 7 2005 /n/sources/patch/applied/smtpd-no-localhost reject calls with helo/ehlo strings of "localhost", "localhost.localdomain" or "localhost.example.com". these are never legit, and are always the mark of a spammer. one can imagine other possibilities, such as 127.0.0.1. d-rwxrwxr-x M 104260 rsc sys 0 Jul 7 2005 /n/sources/patch/applied/linesrcalign fixed the p0/sp correspondence for the optimized case (horizontal/vertical). line(dst, p0, p1, end0, end1, thick, src, sp) Line draws in dst a line of width 1+2*thick pixels joining points p0 and p1. The line is drawn using pix- els from the src image aligned so sp in the source cor- responds to p0 in the destination. d-rwxrwxr-x M 104260 rsc sys 0 Jul 7 2005 /n/sources/patch/applied/man-games-45s-sudoku document [45]s suspend/resume by moving the mouse. document [45]s key stroke commands. fix sudoku left/right button for menu mixup. document (existence of) [45]s score files [should we also document their format?] document [45]s mouse warping limitation bug/feature. [TODO? uniform mouse button names: right now we use both 'button 1' and 'left button' for the same thing.] d-rwxrwxr-x M 104260 rsc sys 0 Jul 8 2005 /n/sources/patch/applied/no-blank-lines blank lines aren't appropriate in manual pages, especially these, which appear just before .SH requests. d-rwxrwxr-x M 104260 rsc sys 0 Jul 9 2005 /n/sources/patch/applied/cpu-update A patch for /sys/src/cmd/cpu.c This patch is designed to be upper compatible to current cpu.c 1. execution local binary cpu -c localcommand exits. now perm bit of /dev/cpunote is 664 writing some message to /dev/cpunote will terminate exportfs (we should restrict the message such as "finished", but I am lazy) 2. grid capable now "user" is taken from /dev/user and the user's home and profile is assigned as follows: (using rc notation) if (! test -e /usr/$user) home=/usr/none if (! test -e $home/lib/profile) bind /usr/none $home thus grid user will execute /usr/none/lib/profile 3. added some log information 4. tiny bug fix for usage message -p to -P d-rwxrwxr-x M 104260 rsc sys 0 Jul 9 2005 /n/sources/patch/applied/nvidia-kernel-update add kernel support for newer nvidia 6XXX series cards. d-rwxrwxr-x M 104260 jmk sys 0 Jul 13 2005 /n/sources/patch/applied/segment-image-refcount Correct reference counting of cached text segment images. Image.ref should always be equal to the number of pointers from Segment and Page structures to the Image. Without this patch, the kernel will leak Images, as shown by this one-liner which will kill any Plan 9 system (change 200 to 2000 for a cpu server): % for (i in `{seq 1 200}) {cp /bin/echo /tmp; /tmp/echo -n & /tmpe/echo $i} d-rwxrwxr-x M 104260 jmk sys 0 Jul 18 2005 /n/sources/patch/applied/graphics-man s/int/void/ in lockdisplay and unlockdisplay d-rwxrwxr-x M 104260 rsc sys 0 Jul 23 2005 /n/sources/patch/applied/listen-usage correct usage messages d-rwxrwxr-x M 104260 rsc sys 0 Jul 23 2005 /n/sources/patch/applied/import-exec import used execl, which had the nil added yesterday, but i suspect it was intended to be execv all along in one case. d-rwxrwxr-x M 104260 rsc sys 0 Jul 25 2005 /n/sources/patch/applied/import-exec-really >right you are how wrong i was! i'm still not sure how i managed to think it worked locally before submitting it. i suspect now that i said mk import.c not mk import.install but i haven't got CCTV tapes to check. d-rwxrwxr-x M 104260 rsc sys 0 Aug 1 2005 /n/sources/patch/applied/auth-warning change two "r"s to " %r"s; add some missing ": %r"s. visible in /sys/log/auth when keyfs starts up: cpu Aug 1 01:13:26 keyfs starting warnings: 42edd9a6 42ec881f warning: can't open upas/expire:r d-rwxrwxr-x M 104260 rsc sys 0 Aug 4 2005 /n/sources/patch/applied/ap-update small update to cope with changes in newsday's website. d-rwxrwxr-x M 104260 jmk sys 0 Aug 7 2005 /n/sources/patch/applied/i82541ip This patch adds support for the Intel PRO/1000 GT Desktop, that uses the i82541PI chipset, which is compatible with i82541GI. d-rwxrwxr-x M 104260 rsc sys 0 Aug 10 2005 /n/sources/patch/applied/dns-soa-expire As DeNIC wants to have other values, then what is default and not changeable in /lib/ndb/local, was this patch created. Now you can change the SOA values of "retry" and "expire", where the last one is expected to be higher than 604800, when host- ing a .de domain - in contrast to the 86400 (Day) default. d-rwxrwxr-x M 104260 jmk sys 0 Aug 16 2005 /n/sources/patch/applied/kbmap-dvorak kbmap files for Dvorak key layout. d-rwxrwxr-x M 104260 rsc sys 0 Aug 22 2005 /n/sources/patch/applied/tlshand-msgrecv-hcertreq-ignore-error don't know if this warrants a patch. it seems I need to relax reporting this length error to be able to make it through the handshake. might be just a bug in the implementation I'm trying to handshake with (still, I have to work around it to make it through the handshake...), or there is something else happening -- if so, I missed it - ignoring this check makes it work: the length fields of the individual DistinguishedName (cas) read in the while loop are correct in this case. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Aug 22 2005 /n/sources/patch/applied/tiny-fixes better suicide evasion and some more comments - nothing very important. d-rwxrwxr-x M 104260 rsc sys 0 Aug 22 2005 /n/sources/patch/applied/tlshand-session-key a 802.1x-ttls supplicant (client) needs to compute a session-key using material gathered in the tls handshake: the prf, master secret, client-random and server-random. Either all these have to be exposed to the user, or the session-key computation has to be done as part of the tls handshake. I'm suggesting to do the latter here, by extending the TLSconn. As far as I can see, this change should not break any existing code, although recompilation is needed. I could imagine other situations where also a session-key has to be computed, but in a different way. I tried to cater for that by the sessionType field which allows/forces the user to specify some kind of computation scheme. If there is a better way to do this please let me know (I tried to get feedback via 9fans, without success). Axel. d-rwxrwxr-x M 104260 rsc sys 0 Aug 22 2005 /n/sources/patch/applied/petty-bug fix a petty bug in ipquery, the wrong pointer passed to ndbfree(). I saw it and couldn't leave it alone. d-rwxrwxr-x M 104260 rsc sys 0 Aug 22 2005 /n/sources/patch/applied/howard-email trickey formaly known as howard - this patch restores his face d-rwxrwxr-x M 104260 rsc sys 0 Aug 22 2005 /n/sources/patch/applied/webfs-typos fixed a typo and made the description of opening up a new connection clearer d-rwxrwxr-x M 104260 rsc sys 0 Aug 22 2005 /n/sources/patch/applied/mit Maybe you consider this an old joke? however I hadn't heard it before - I laughed. d-rwxrwxr-x M 104260 rsc sys 0 Aug 29 2005 /n/sources/patch/applied/file-gzip After my stupidity with vncv (russ knows) I picked somthing off my to-do. File now understands gzip and bz2 files, also shifted the ismung() file histogram's sampling down fthe file as unreprisentitive headers tend to live at the start. d-rwxrwxr-x M 104260 rsc sys 0 Aug 29 2005 /n/sources/patch/applied/acme-enable-menu-via-ctl interpret menu and nomenu in the ctl file to set resp. unset filemenu which controls whether or not the automagic Undo, Redo, Put are in the tag. this is needed to allow automagic Undo/Redo/Put for acme Wiki -- because all files edited in the wiki look like directories to acme the automagic Undo/Redo/Put are switched off there. a separate patch (acme-wiki-filemenu-in-tag) depends on this one. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Aug 29 2005 /n/sources/patch/applied/tlshand-msgrecv-hcertreq-fix fix for the HCertificateRequest in tlshand.c:/^msgRecv - use the right sizes for length fields - add additional check before the get16 at tlshand.c:1150 - add p += nn; n -= nn; which were missing after makebytes at tlshand.c:1145 the latter fixes the problem I had with the test at tlshand.c:1153 where the (n != nn) part succeeded while we would want it (would expect it) to fail. I've not yet tested this on the real system but - it feels and looks convincingly ok, - it explains the values of n and nn that I saw, and, better than that, - it checks out with the data I gathered earlier while debugging this (unfortunately I have only a single sample -- the one from our campus net) as before, please doublecheck; I'm feeling good about this. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Aug 29 2005 /n/sources/patch/applied/acme-wiki-use-winismodified use the acme ctl file ismodified field instead of maintaining our own dirtied variable. When the user tries to Get a modified file, we do remember if we have warned already, and reset that information in the same places where we used to set the dirtied variable. this patch includes the acme-wiki-filemenu-in-tag patch (the change in wiki.c around 441c445,447) which depends on acme-enable-menu-via-ctl , and the acme-wiki-fix-dirtystate patch (ignore writes to body or tag; winclean after setting file name and tag commands). Axel. d-rwxrwxr-x M 104260 rsc sys 0 Aug 29 2005 /n/sources/patch/applied/acme-wiki-filemenu-in-tag This patch depends on acme-enable-menu-via-ctl . It includes patch acme-wiki-fix-dirtystate . It adds enabling of the automagic Undo/Redo/Put in the tag using the acme 'menu' ctl added by acme-enable-menu-via-ctl (in addition to what acme-wiki-fix-dirtystate already did). Axel. d-rwxrwxr-x M 104260 rsc sys 0 Aug 29 2005 /n/sources/patch/applied/acme-wiki-fix-dirtystate avoid that Wiki thinks the file is diry after a Get (and thus complains when we Get multiple times, without changing the file) also: avoid that we can go too far back with Undo. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Sep 1 2005 /n/sources/patch/applied/spin425 Upgrade the version of Spin included with Plan 9 to 4.2.5. NOTE: I do not use Spin, so I'm not sure it's 100% correct, but it seems to work with casual testing. I also did not update the Spin documents in /sys/doc because I am unsure of their origin. d-rwxrwxr-x M 104260 rsc sys 0 Sep 1 2005 /n/sources/patch/applied/memo-fixes some bug fixes, code cleanup. fgb d-rwxrwxr-x M 104260 rsc sys 0 Sep 4 2005 /n/sources/patch/applied/termrc-kbmap A small change to the /rc/bin/termrc, to include /dev/kbmap into the default namespace of a terminal. d-rwxrwxr-x M 104260 rsc sys 0 Sep 4 2005 /n/sources/patch/applied/wikifs-dotdot Fix qid generation for root so you can dotdot out of a page directory (e.g. /n/wiki/sandbox). I think this is the same idea as Russ's recent patch to devsd.c. d-rwxrwxr-x M 104260 rsc sys 0 Sep 5 2005 /n/sources/patch/applied/sokoban-bfs-responsive-leak - be gentle to glenda, by allowing her to walk the shortest distance in a multi-step walk (found by using breadth-first search instead of depth-first search) - while walking with animation enabled, remain responsive, by splitting the animation of a walk in single-step moves that are 'scheduled' using etimer(2). This allows e.g. redirecting a walk during the animation, while glenda is still walking, or disabling animation during a walk. - reduce the amount of leaks reported by leak(1) by building the levels menu contents 'by hand' instead of relying on menuhit to call genlevel (would there be a way to avoid leaking while letting menuhit call genlevel?) Note: I still get some leaks, after popping up the B2 menu: acid: src(0x00009627); // 4 /sys/src/libc/fmt/vsmprint.c:38 I don't quite see where they come from. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Sep 7 2005 /n/sources/patch/applied/utfnlen_missing add utfnlen declaration chan.c calls it d-rwxrwxr-x M 104260 rsc sys 0 Sep 10 2005 /n/sources/patch/applied/mv-remove I tracked down an anoyance in my cifs client -it will not let me delete open files - which I should fix. The only cluprit I found was mv(1) and there seems to be no reason for it do do this; so I "fixed" it. I will not be surprised if this patch gets sorry'ed. PS: tried to preserve the codeing conventions used in mv(1) but they appear to be mixed. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Sep 10 2005 /n/sources/patch/applied/prep-memsize remove unused function d-rwxrwxr-x M 104260 rsc sys 0 Sep 10 2005 /n/sources/patch/applied/libdisk-zero opendisk seemed to assume that memory was zeroed, so that part and h/c/s were initially zero, but user-level malloc no longer does that. changed to use mallocz. scsi.c did include the memset but i've changed it to use mallocz as well. d-rwxrwxr-x M 104260 rsc sys 0 Sep 10 2005 /n/sources/patch/applied/dist-termrc A change to the install floppy (floppy and CD)'s termrc, so people are asked, when textinstall is set to "ask", if they want to do a text install or the graphical one. Some people complained on #plan9, that the installer uses too much memory in graphical mode. d-rwxrwxr-x M 104260 rsc sys 0 Sep 10 2005 /n/sources/patch/applied/dd_error dd returns a clear completion code even when it encounters an error. I had reason to hack a simple alterntive, but there are probably better ways to go about it. And there may be other places deserving attention. d-rwxrwxr-x M 104260 rsc sys 0 Sep 15 2005 /n/sources/patch/applied/cddbutf8 set protocol level to 6 before doing anything, it makes the server talk utf-8 back. d-rwxrwxr-x M 104260 rsc sys 0 Sep 18 2005 /n/sources/patch/applied/man-3-usb_mention_usbn_port mention #U/usbn/port. d-rwxrwxr-x M 104260 rsc sys 0 Sep 18 2005 /n/sources/patch/applied/acmemail-errornewlines remove some newlines from output messages. error() appends the newline already. d-rwxrwxr-x M 104260 rsc sys 0 Sep 18 2005 /n/sources/patch/applied/libsec-plug-some-leaks an attempt at plugging some holes reported by leak(1). for me this does not break anything, and my 802.1x ttls-pap thingy now runs without leaks, as far as leak(1) tells me. there may be more leaks-- I did not exercise the code of x509.c:/^verify_signature beyond the "expected 1" message. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Sep 19 2005 /n/sources/patch/applied/setmalloctag-libsec-libmp add the setmalloctags I used to analyse the libsec-plug-some-leaks and find who should be responsible for cleanup of the leaked memory. also removed one now superflous ...=nil; in x509.c Axel. d-rwxrwxr-x M 104260 jmk sys 0 Sep 21 2005 /n/sources/patch/applied/as-quad allow assemblers' yylex to work for 64-bit decimal values when yylval is vlong, but act as before for 32-bit targets; and skip any trailing Us and LLs as might be used in mem.h to define addresses usable by C. d-rwxrwxr-x M 104260 rsc sys 0 Sep 22 2005 /n/sources/patch/applied/dns.h-update This patch adds all not included result messages, rr types and opcodes from http://www.iana.org/assignments/dns-parameters to the DNS server in Plan 9. d-rwxrwxr-x M 104260 rsc sys 0 Sep 22 2005 /n/sources/patch/applied/cb-col-split This patch has two changes for cb. It avoids special processing for a colon that is followed by another colon. If you are forced to look at C++ code, then you sometimes want to hit small fragements of it with cb. This patch prevents identifiers a::b from being broken up. It prefixes split lines with tab rather than four spaces. This looks better in acme. d-rwxrwxr-x M 104260 rsc sys 0 Sep 22 2005 /n/sources/patch/applied/ramfs-man-sync This patch cleans up ramfs(4) and /sys/src/cmd/ramfs.c, to be in sync. d-rwxrwxr-x M 104260 rsc sys 0 Sep 23 2005 /n/sources/patch/applied/nforce2-dma Added an entry for my nForce2 IDE controller as a passthru. d-rwxrwxr-x M 104260 rsc sys 0 Sep 24 2005 /n/sources/patch/applied/fix-typo Typographical error: Specfying -> Specifying d-rwxrwxr-x M 104260 jmk sys 0 Sep 26 2005 /n/sources/patch/applied/wavelan-new-card recognize my wlan card: "INTERSIL;HFA384x/IEEE;Version 01.02;", d-rwxrwxr-x M 104260 rsc sys 0 Oct 1 2005 /n/sources/patch/applied/lm78-dir Add a Dirtab entry for ".", which some of the devgen routines blindly assume is always present as entry 0. Without this nothing at all, not even ".", appears when I list the directory. After adding it I can read the temperature off my ASUS P2B motherboard. d-rwxrwxr-x M 104260 rsc sys 0 Oct 1 2005 /n/sources/patch/applied/wikifs-tohtml-improvements Minnor changes to wikifs/tohtml.c so it generates more 'standard'(hah!) html that can be styled thru css. Also remove spaces from anchor names and other various fixes. d-rwxrwxr-x M 104260 rsc sys 0 Oct 2 2005 /n/sources/patch/applied/snoopy-eapol-eap-ttls-rc4keydesc Add support for eapol, eap, ttls and rc4 key descriptors. These things are used by 802.1X. Though I may still tweak this a bit more, I think it should already be useful to others and more or less ready for prime time. Some possible critisism on this version: Maybe some of the sub-protocol names and field names I use are a bit (too?) verbose. The naming of fields is not entirely uniform. The capitalization of the subprotocols as printed in the `encapsulating' header might be a bit too much (though dhcp/bootp capitalize too). Right now I also `dump' the contents of ttls packets, which may be superflous. I don't know yet what I prefer. I'm not convinced I got the support for filtering etc. exactly right (i.e. all fields of Proto) but what I have seems to work for me. Nevertheless, I'm submitting this anyway because I think it is already useful as-is, even when improvement is possible. Axel. d-rwxrwxr-x M 104260 rsc sys 0 Oct 6 2005 /n/sources/patch/applied/newnsquote Permits quotes in the spec string in namespace files. Adds a debug flag to see what failed while doing a newns. Adds a couple of checks to abort instead of going Broken. d-rwxrwxr-x M 104260 rsc sys 0 Oct 9 2005 /n/sources/patch/applied/tbl-text have tbl, not troff, diagnose a missing T} d-rwxrwxr-x M 104260 rsc sys 0 Oct 9 2005 /n/sources/patch/applied/webfs-auth added basic auth to webfs. d-rwxrwxr-x M 104260 rsc sys 0 Oct 13 2005 /n/sources/patch/applied/httpd-range-compliance This patch solves some problems in wget and Firefox, when doing a partial request (Range:) on a file on a Plan 9 httpd. The Plan 9 httpd does not send out a Content-Length: when doing so, but these applications expect to have one and show it directly to the user. One clause in the HTTP 1.1 RFC says in section 4.4 Message Length: "For compatibility with HTTP/1.0 applications, HTTP/1.1 requests contain- ing a message-body MUST include a valid Content-Length header field unless the server is known to be HTTP/1.1 compliant." Compliant or not complianet - here is the patch. d-rwxrwxr-x M 104260 rsc sys 0 Oct 15 2005 /n/sources/patch/applied/chanfix namec(..) advances the name-pointer, so free fails if it has been advanced enough. (dont) try ls '#I10'. free(name) replaced with free(aname) d-rwxrwxr-x M 104260 rsc sys 0 Oct 19 2005 /n/sources/patch/applied/tarfs-ustar typo in fs/tarfs's ustar support, test against /n/sources/9grid/xcpu.tar d-rwxrwxr-x M 104260 rsc sys 0 Oct 19 2005 /n/sources/patch/applied/bc-suicide prevent suicide when bc(1) is envoked with bad args d-rwxrwxr-x M 104260 rsc sys 0 Oct 28 2005 /n/sources/patch/applied/window-scroll add -scroll and -noscroll options to window(1). d-rwxrwxr-x M 104260 rsc sys 0 Oct 28 2005 /n/sources/patch/applied/fgui-diagnostic fgui now aborts with a diagnostic if it cannot create its hidden window. stops stupid people like me from starting it from $home/profile. d-rwxrwxr-x M 104260 rsc sys 0 Oct 28 2005 /n/sources/patch/applied/lensmenu adds a menu under button 3 that has the keyboard commands in it. in addition to 'q' and ctrl-d, del now also exits the program. d-rwxrwxr-x M 104260 rsc sys 0 Oct 29 2005 /n/sources/patch/applied/cb-colon-drops-next-char this appears to fix the bug of cb dropping the character after a colon. d-rwxrwxr-x M 104260 rsc sys 0 Oct 29 2005 /n/sources/patch/applied/lib9p In file.c there seemed to be a leak. Also, the order of sons in a file is preserved wrt creation order. If we create ./a ./b then rm ./b and create ./c, the list of sons has ./a ./c. This is important for omero, where the order of files determines the order of widgets. In srv.c an incref (already marked in sources(?)) was missing. d-rwxrwxr-x M 104260 rsc sys 0 Oct 30 2005 /n/sources/patch/applied/rawon A non blocking rawon that preserves existing chars and does not insert vapor-newlines. d-rwxrwxr-x M 104260 rsc sys 0 Oct 31 2005 /n/sources/patch/applied/mousekeys Provides keys Kmouse|mouse_button that can be placed in the kbmap to provide mouse buttons through keyboard keys. BEWARE: The kbmap supplied already maps left ctrl,start,alt to mouse keys (as well as f1,f2,f3). Use the previous tables instead to provide the previous behaviour as a default. d-rwxrwxr-x M 104260 rsc sys 0 Nov 1 2005 /n/sources/patch/applied/memo-hide axel spotted that memo doesn't hide properly with rio. this patch fixes the problem. andrey d-rwxrwxr-x M 104260 rsc sys 0 Nov 9 2005 /n/sources/patch/applied/rhineiii-ids Add the vendor and device id for a dlink card, BSDs claims it is a VIA Rhine III card. d-rwxrwxr-x M 104260 rsc sys 0 Nov 11 2005 /n/sources/patch/applied/rio-resize-warp Warp mouse to outer window rectangle when starting a resize drag. d-rwxrwxr-x M 104260 rsc sys 0 Nov 11 2005 /n/sources/patch/applied/drawmalloctagsandkeys adds a few setmalloctag to debug programs using images and fonts. defines a few keys in keyboard.h, like Esc, Del, etc.. Nothing serious :-) d-rwxrwxr-x M 104260 rsc sys 0 Nov 12 2005 /n/sources/patch/applied/rawfix Applies the rawon/rawoff fix to vncs. I used the code from 9/port/devcons.c, which works perfectly and can be understood :-) d-rwxrwxr-x M 104260 rsc sys 0 Nov 12 2005 /n/sources/patch/applied/pingbroken Ping was broken. It always exited with an wrong status. It sometimes waited for a timeout due to a race condition. Both problems are solved by this patch. d-rwxrwxr-x M 104260 rsc sys 0 Nov 13 2005 /n/sources/patch/applied/tar-v-man small change to tar(1) to describe the new behaviour of `tar v' (no function letter output). d-rwxrwxr-x M 104260 rsc sys 0 Nov 13 2005 /n/sources/patch/applied/tar-r-fix Whoops. blkoff was getting set in the wrong place, overwriting the old archive contents under `tar rf'. d-rwxrwxr-x M 104260 rsc sys 0 Nov 19 2005 /n/sources/patch/applied/smtpd-log-recomp-errs ensure that regcomp() errors get logged by having regerror() syslog them, and by having smtpd capture send's stderr and logging it even if send's exit status is good. i also changed "smtpd" to "smtpd.db" for the log file for -d output; feel free to keep or discard that change. d-rwxrwxr-x M 104260 rsc sys 0 Nov 19 2005 /n/sources/patch/applied/matrox On my matrox millennium II, aux/vga failed with mga2164w: can't attach mga2164wmmio segment This patch adds the segment. d-rwxrwxr-x M 104260 rsc sys 0 Nov 19 2005 /n/sources/patch/applied/realmode The "realmode" entry was amongst "misc" and seems intended to be amongst "link" instead. d-rwxrwxr-x M 104260 rsc sys 0 Nov 19 2005 /n/sources/patch/applied/setmalloctags A couple of setmalloctag() for getenv and smprint d-rwxrwxr-x M 104260 rsc sys 0 Nov 19 2005 /n/sources/patch/applied/namefixandleak Avoids a leak of uid, and changes the default name for the root file of the tree to "dev", because "/" is an illegal name that may lead to errors for file system multiplexors. This patch is not against the patched version of lib9p I'm using, it's against the one in sources. d-rwxrwxr-x M 104260 rsc sys 0 Nov 20 2005 /n/sources/patch/applied/ramfs-senseless-if This patch removes a senseless if() clause in ramfs. d-rwxrwxr-x M 104260 jmk sys 0 Nov 26 2005 /n/sources/patch/applied/db-imm64 here are a few changes i hadn't sent along before. some literals weren't displayed properly in 64-bit code. distinguish RIP mode in the output (mainly for linker debugging) d-rwxrwxr-x M 104260 rsc sys 0 Nov 28 2005 /n/sources/patch/applied/9fat_devfs Mount /dev/fs/9fat if available d-rwxrwxr-x M 104260 rsc sys 0 Nov 28 2005 /n/sources/patch/applied/plumbopenwithcreate I did not like the one sent in the previous patch. I think this one is more clean and correct. I tried with acme and omero with plumber and youps. d-rwxrwxr-x M 104260 rsc sys 0 Nov 28 2005 /n/sources/patch/applied/import-no-auth This adds a -A option to import, which forces it to use no auth- entication at all. This is required for connecting to Inferno on other host systems than Plan 9, where no authentication can be performed. d-rwxrwxr-x M 104260 rsc sys 0 Nov 28 2005 /n/sources/patch/applied/lib9pbugfix Fixes a bug in lib9p caused because respond may free the flush list and then call closereq, which would use the flush list to respond. Happens when you provide no flush, yet you handle requests concurrently. Note: this patch is against the version in sources, not against the patch I sent before for lib9p to do other things. d-rwxrwxr-x M 104260 jmk sys 0 Dec 2 2005 /n/sources/patch/applied/vtupdate some 'improvements' like (partial) support for xterm controls, colors are now different, -w switch makes resize events propagate properly when using ssh -w. it still isn't perfect, but it can be used with programs like screen or pine now. -f switch selects a font, -x switch exports term as xterm instead of vt100. d-rwxrwxr-x M 104260 rsc sys 0 Dec 3 2005 /n/sources/patch/applied/man3 Just minor typo adjustments that seem obvious. d-rwxrwxr-x M 104260 rsc sys 0 Dec 3 2005 /n/sources/patch/applied/man_smtp2 Sorry, forgot to save the final copy :-( d-rwxrwxr-x M 104260 rsc sys 0 Dec 3 2005 /n/sources/patch/applied/man_smtp Got rid of a couple of typos and removed an old (?) description of the "a" option to smtp(8). d-rwxrwxr-x M 104260 rsc sys 0 Dec 3 2005 /n/sources/patch/applied/smtp-man update smtp manual to better reflect reality. d-rwxrwxr-x M 104260 rsc sys 0 Dec 3 2005 /n/sources/patch/applied/syslog-udp syslog udp port definition d-rwxrwxr-x M 104260 rsc sys 0 Dec 3 2005 /n/sources/patch/applied/factotum-cs-fix If authdial fails, try with what we know. suggested by rsc. d-rwxrwxr-x M 104260 jmk sys 0 Dec 6 2005 /n/sources/patch/applied/vtman explain the new -f, -x and -w switches d-rwxrwxr-x M 104260 rsc sys 0 Dec 7 2005 /n/sources/patch/applied/fortune Deleted code is debugged code - Jeff Sickel d-rwxrwxr-x M 104260 rsc sys 0 Dec 7 2005 /n/sources/patch/applied/scuzz-file-size-vlong set the default file size to the largest vlong. `& ~iosize' in the old code just looks like a bug. d-rwxrwxr-x M 104260 rsc sys 0 Dec 10 2005 /n/sources/patch/applied/stdio.file.leak FILE's closed in ERR state are not marked as CLOSED and thus cannot be reused. d-rwxrwxr-x M 104260 rsc sys 0 Dec 11 2005 /n/sources/patch/applied/man1-sync1 Solved the critics to man1-sync0. d-rwxrwxr-x M 104260 rsc sys 0 Dec 12 2005 /n/sources/patch/applied/9p-auth-and-listen Changes lib9p to add helpers to perform authentication for clients and to add netsrv/threadnetsrv wrappers to handle network clients. respondstrerr() does a respond of %r, and the client is single threaded by default (i.e., srv does lock to serialize calls to service procedures, as they would be if [thread]netsrv() was never called. d-rwxrwxr-x M 104260 rsc sys 0 Dec 12 2005 /n/sources/patch/applied/i810-cursor Fixes problem with the mouse cursor color on the i810. d-rwxrwxr-x M 104260 rsc sys 0 Dec 14 2005 /n/sources/patch/applied/spellcheck Just correcting an irritating spelling error. d-rwxrwxr-x M 104260 rsc sys 0 Dec 14 2005 /n/sources/patch/applied/ip-fixes I've been running with these changes for a long time. The change to icmp6.c just corrects packet counts. The changes to il.c were jmk's suggestion to avoid exhausting the supply of rpc mount buffers. d-rwxrwxr-x M 104260 rsc sys 0 Dec 14 2005 /n/sources/patch/applied/wikifs-not-allow-questionmark-in-title Pages with '?' in their name break links in the web interface. The proper solution is maybe to generate links that are properly url-encoded, that would remove the limitations on <> too, but url-encoding is trickier than it seems. But for now it's easier to just keep people from creating such pages. d-rwxrwxr-x M 104260 rsc sys 0 Dec 14 2005 /n/sources/patch/applied/tlssrv_syslog Removed the trailing linefeed that gives rise to double spacing in the syslog. d-rwxrwxr-x M 104260 rsc sys 0 Dec 14 2005 /n/sources/patch/applied/man1-sync2 A bit more sync. eqn(1) is the hardest part. d-rwxrwxr-x M 104260 rsc sys 0 Dec 18 2005 /n/sources/patch/applied/c-no-number-sign avoid opening files as '#foo' during creation of archives (not just during extraction). d-rwxrwxr-x M 104260 rsc sys 0 Dec 20 2005 /n/sources/patch/applied/rc-no-break-on-bad-here-doc This prevents an rc suicide when processing an ill-formed here document. d-rwxrwxr-x M 104260 rsc sys 0 Dec 20 2005 /n/sources/patch/applied/hget-factotum-attrs change the attributes passed to factotum by hget to be the same as those passed by webfs, and correct their spelling. (my fault in the first place). d-rwxrwxr-x M 104260 rsc sys 0 Dec 20 2005 /n/sources/patch/applied/i810-vga-blanking Fixes blanking on i81x. d-rwxrwxr-x M 104260 rsc sys 0 Dec 20 2005 /n/sources/patch/applied/man1-sync3 More sync to man(1). Filter was really down to date. d-rwxrwxr-x M 104260 rsc sys 0 Dec 21 2005 /n/sources/patch/applied/tar-redundant-call quotefmtinstall() call seems to be redundant. [came across this while trying to find out programs that quote printed file names like ls and du. i missed -Q option (ls(1)) in du (in pipelines) till i discovered this unquote mechanism: du -a | sed 's/^/echo /' | rc ] Arvindh d-rwxrwxr-x M 104260 rsc sys 0 Dec 24 2005 /n/sources/patch/applied/ping-byteorder-fix Fix sequence number byte order mismatch. d-rwxrwxr-x M 104260 rsc sys 0 Dec 24 2005 /n/sources/patch/applied/marshal-p-fix invoke pgp correctly. d-rwxrwxr-x M 104260 jmk sys 0 Dec 29 2005 /n/sources/patch/applied/smc8041tx Add SMC 8041TX ne2k clone id string. Thanks to 20h and Oksel for figuring it out. d-rwxrwxr-x M 104260 rsc sys 0 Dec 31 2005 /n/sources/patch/applied/adict-fix2 changed threadprint() by fprint() and added a STACK enum, to avoid stack overflows. d-rwxrwxr-x M 104260 rsc sys 0 Jan 1 2006 /n/sources/patch/applied/acme-2-3-chord After selecting text, with button 1 down, executed 2-3 chord repeatedly. When button 3 was pressed the second time, the previous Cut was not Undone. Hope this patch fixes it. Arvindh d-rwxrwxr-x M 104260 rsc sys 0 Jan 4 2006 /n/sources/patch/applied/page-angle If the page being viewed is rotated 180° by selecting "upside down" (B2 menu), "orig size" retains the angle. But if the page is rotated 180° by selecting "rotate 90" twice, "orig size" doesn't retain the angle. Arvindh d-rwxrwxr-x M 104260 rsc sys 0 Jan 4 2006 /n/sources/patch/applied/man1-sync4 New year, new work. I left gs(1) alone, as it is an import. d-rwxrwxr-x M 104260 rsc sys 0 Jan 8 2006 /n/sources/patch/applied/pic-unicode unicode data was filtered out by pic. d-rwxrwxr-x M 104260 rsc sys 0 Jan 8 2006 /n/sources/patch/applied/tbl-unicode fix segment violation on some input data. d-rwxrwxr-x M 104260 rsc sys 0 Jan 8 2006 /n/sources/patch/applied/ape-dirstat ape/make reporting when in deep directories "Directory .: Make: Cannot open. Stop." Turned out to be a buglet in the implementation of dirstat in ape, obvious when compared to /sys/src/libc/9sys/dirstat.c Unfortunately this will mean most of APEs commands will need to be recompiled and pushed out. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Jan 8 2006 /n/sources/patch/applied/fcp-err-msg error messages from fcp contain "cp: " instead of "fcp: ". also, "man fcp" doesn't return a man page. i believe the man indices need to be rebuilt: diff /sys/man/1/INDEX <{/sys/lib/man/secindex /sys/man/1} thanks, arvindh d-rwxrwxr-x M 104260 rsc sys 0 Jan 8 2006 /n/sources/patch/applied/aquarela-man aquarela manual, improved usage message, logfile written to /sys/log if it already exists. -Steve d-rwxrwxr-x M 104260 rsc sys 0 Jan 8 2006 /n/sources/patch/applied/prep-8-fix-new Again the kernel added and "386" stays there. d-rwxrwxr-x M 104260 rsc sys 0 Jan 11 2006 /n/sources/patch/applied/nedmail-png-display add a table entry to display .png files. d-rwxrwxr-x M 104260 jmk sys 0 Jan 13 2006 /n/sources/patch/applied/mahjongg-control -f causes mahjongg to indicate non-blocked tiles on mouse-over. d-rwxrwxr-x M 104260 rsc sys 0 Jan 20 2006 /n/sources/patch/applied/usbd-vid-did Adds vid and did output to usb/usbd. First, this is required because of the USB spec, that says, that the USB software should first look for vid and did and then for the device class. Second, for example an UART driver has no USB device class and needs checking of the vid and did. d-rwxrwxr-x M 104260 rsc sys 0 Jan 20 2006 /n/sources/patch/applied/nvidia-mem Fix nvidia driver to map the dma ring. Changed cursor location (into 64MB mapped area). d-rwxrwxr-x M 104260 rsc sys 0 Jan 23 2006 /n/sources/patch/applied/usbaudio-mouse-work Get usb/usbaudio and usb/usbmouse to work again. In usbmouse is the dependency on stdio.h removed, so it got 30Kb smaller on my dev machine. d-rwxrwxr-x M 104260 rsc sys 0 Jan 24 2006 /n/sources/patch/applied/killbig-nits-2 Supersed killbig-nits Fix print calls in swap.c^page and proc.c^killbig so they make sense and are in a more logical order. Change killbig to not have hardcoded the file server name and instead avoid killing process owned by eve that have non-writeable /ctl. Still not sure exactly where in the rc scripts this line would fit best: for(i in `{ps|grep '(kfs|fossil|venti)'|awk '{print $2}'}) { chmod -w /proc/$i/ctl; } d-rwxrwxr-x M 104260 rsc sys 0 Jan 24 2006 /n/sources/patch/applied/msexceltables-fix I think, now, finally, I understand, and have fixed msexceltables.c in the face of unicode and rich text (coloured fonts) and asian phonetic hints. d-rwxrwxr-x M 104260 rsc sys 0 Jan 24 2006 /n/sources/patch/applied/vncv-wheel-fix The plan9 /dev/mouse protocol doesn't send *release* button message when using wheel buttons. With that fix the vnc server can now handle successive wheel button messages. d-rwxrwxr-x M 104260 rsc sys 0 Jan 24 2006 /n/sources/patch/applied/worthy-dull Revision control is like any other form of accountancy -- worthy but dull d-rwxrwxr-x M 104260 rsc sys 0 Jan 24 2006 /n/sources/patch/applied/read2-dirread-also add SEE ALSO for dirread(2) d-rwxrwxr-x M 104260 rsc sys 0 Jan 25 2006 /n/sources/patch/applied/ethervgbe The latest revision of the VIA Velocity Gigabit Ethernet driver. It has been tested on a 100Mbps Full duplex network and seems to work quite well. d-rwxrwxr-x M 104260 rsc sys 0 Jan 25 2006 /n/sources/patch/applied/uart-update-manpage Next try for uart(3). I've only seen manpages in the console and as HTML, where formatting doesn't matter that much, in my life. Maybe that's the problem. d-rwxrwxr-x M 104260 rsc sys 0 Jan 25 2006 /n/sources/patch/applied/vtdial-better-err-msg force vtDial to name the offending venti server. if there isn't one, make that clearer too. d-rwxrwxr-x M 104260 rsc sys 0 Jan 25 2006 /n/sources/patch/applied/man-6-smtpd-ratfs Smtpd doesn't look into /mail/lib/blocked itself (anymore). Ratfs does that. Smtpd looks into /mail/ratify d-rwxrwxr-x M 104260 rsc sys 0 Jan 28 2006 /n/sources/patch/applied/htmlroff-warn make warnings start with "htmlroff: ". d-rwxrwxr-x M 104260 rsc sys 0 Jan 28 2006 /n/sources/patch/applied/htmlroff-clusmy fix typo d-rwxrwxr-x M 104260 rsc sys 0 Jan 28 2006 /n/sources/patch/applied/ms2html-man finish ms2html(1) and improve it. d-rwxrwxr-x M 104260 rsc sys 0 Jan 30 2006 /n/sources/patch/applied/snoopy-gre make snoopy dissect gre. there doesn't seem to be a registry that keeps track of protocol numbers used in gre. the current ones (used in in the mux structure) come from what was in the old gre.c and an older gre rfc (with all too-ancient-looking protocols removed (criteria i used is that they had to be present in ip.c). the eopol, dot1q, llc were used in other software. d-rwxrwxr-x M 104260 rsc sys 0 Jan 30 2006 /n/sources/patch/applied/snoopy-namenits give compile the correct name argument. only used for the possible error message. d-rwxrwxr-x M 104260 rsc sys 0 Jan 30 2006 /n/sources/patch/applied/lex-debug allprint prototype correction for 8-bit cleanness weird debug message fix (my case was negative input() return values of pic) d-rwxrwxr-x M 104260 rsc sys 0 Jan 30 2006 /n/sources/patch/applied/tbl-args fallback to stdin when some flags with no files present d-rwxrwxr-x M 104260 rsc sys 0 Jan 31 2006 /n/sources/patch/applied/tar-x-vs-non-empty-links I just ran into this reading an old 4bsd tar archive: if the size on the tape of a link is non-zero, tar x and tar t will get confused about position on the tape and typically blow up with a checksum error. (unix tar evidently had the same bug; I tried a late research unix tar and it blew up the same way.) The fix is one line; I added some comments and a little paranoia against similar future problems. d-rwxrwxr-x M 104260 rsc sys 0 Feb 3 2006 /n/sources/patch/applied/closescsi closescsi(2) is documented in the manpage but not in the header or source. Perhaps it was missed when the code was written, or removed when the thread locking was added? -Steve d-rwxrwxr-x M 104260 rsc sys 0 Feb 3 2006 /n/sources/patch/applied/usbd-hub-fix This fixes the issue, that usbd was unable to get the desc- riptor of the HUB and some cleanups in the OR creation of some requests. d-rwxrwxr-x M 104260 rsc sys 0 Feb 10 2006 /n/sources/patch/applied/9fs-patch-anon A patch to fix replica/pull ... network to work with anonymous sources and add options to 9fs, that make it possible for new users to access 9fs. The solution to have a second case was choosen, so there won't be irritations in the future, about; if we are now using sources anonymiously or not. d-rwxrwxr-x M 104260 rsc sys 0 Feb 10 2006 /n/sources/patch/applied/faster-awk per my mail, avoid slow strtod() and ferror(). maybe the strtod screen should go into all 3 or 4 versions of strtod (or maybe there should be fewer versions) instead. d-rwxrwxr-x M 104260 rsc sys 0 Feb 12 2006 /n/sources/patch/applied/touch-time-arg-validation Make touch -t more foolproof so it quits when passed invalid values rather than set time to the epoch. Got bitten by this in a script when Make touch -t more foolproof so it quits when passed invalid values rather than set time to the epoch. And a tiny style fix to make the whole file have a consistent style. d-rwxrwxr-x M 104260 rsc sys 0 Feb 12 2006 /n/sources/patch/applied/2l-formatting fix up formatting of tagged paragraphs. d-rwxrwxr-x M 104260 rsc sys 0 Feb 13 2006 /n/sources/patch/applied/getflags-script-changes aux/getflags changes to the various scripts. d-rwxrwxr-x M 104260 rsc sys 0 Feb 13 2006 /n/sources/patch/applied/getflags-formatting add missing `.SH NAME' and delete extra blank line from SOURCE. d-rwxrwxr-x M 104260 rsc sys 0 Feb 13 2006 /n/sources/patch/applied/venti-printarenas mkfile to build printarena by default, dumpvacroots to pick up $venti, makes retreiving lost vac scores less hassle when disks die. d-rwxrwxr-x M 104260 rsc sys 0 Feb 13 2006 /n/sources/patch/applied/ape-dumb-autohell-fixes Small 'fixes' to APE to make auto*hell a bit happier. Had to create a fork of ls to better emulate Unix ls: Auto*hell tries to use ls -L because some sort of symlink braindamage which obviously doesn't apply to Plan 9, allow that flag but ignore it. Auto hell does some weird tests comparing the arguments it gives to ls with the output of ls, cleanname() breaks those tests by rewriting parts of the path, so removed the calls to cleanname(). Auto*hell also often calls printf with a single argument and no formating, so I created a really dummy printf that just prints it's first argument. This patch is not of much value, but when trying to wade thru auto*hell land it removes lots of noise. Further patches might follow as I go thru the experience of trying to get a single auto*hell script to run, if I don't go mad first. d-rwxrwxr-x M 104260 rsc sys 0 Feb 16 2006 /n/sources/patch/applied/replica-pull-usage make usage consistant with manpage d-rwxrwxr-x M 104260 rsc sys 0 Feb 16 2006 /n/sources/patch/applied/mpman documentation for mpsignif was incorrect. actually, although the documentation is not wrong, mplowbits0(0) and mplowbits0(1) both return 0, which could be construed as an inconsistency. d-rwxrwxr-x M 104260 rsc sys 0 Feb 16 2006 /n/sources/patch/applied/iostats a few minor problems: - modes weren't printed properly (fcallfmt assumes dirmodefmt and dirfmt). - the number of workbufs wasn't sufficient for lots of parallel reads (e.g. when testing fcp) the fix should probably allow an indefinite number, but 40's better than 16! - it didn't follow $path when executing a file. i'm aware that $path's not so important in plan 9, but i (and perhaps others too) have path=(/bin .) and it can lead to surprises when iostats tries things the other way around. it's a pity that rdenv() is so painful, but i'm not aware of an alternative. d-rwxrwxr-x M 104260 rsc sys 0 Feb 18 2006 /n/sources/patch/applied/mkfile-addsudoku add sudoku to mkfile so it will be compiled and installed d-rwxrwxr-x M 104260 rsc sys 0 Feb 18 2006 /n/sources/patch/applied/ether83815-changes Here are the relevant changes from /sys/src/fs/pc/ether83815.c to /sys/src/9/pc/ether83815.c. Some of the changes include more device IDs supported and better handling of ethernet addresses. Federico d-rwxrwxr-x M 104260 rsc sys 0 Feb 18 2006 /n/sources/patch/applied/auth-stderr-tags I've added a tag with the program's name to each fprint(2, ...) in /sys/src/cmd/auth that wasn't obviously debugging output or part of a protocol ('!' seems to be magic at the start of some messages). In some cases, I switched to using sysfatal(). It all compiles and I'm about to install it and reboot the world. d-rwxrwxr-x M 104260 rsc sys 0 Feb 20 2006 /n/sources/patch/applied/man4_nfs Two very minot tweaks. d-rwxrwxr-x M 104260 rsc sys 0 Feb 20 2006 /n/sources/patch/applied/svorak Svorak layout by Magnus Engström d-rwxrwxr-x M 104260 rsc sys 0 Feb 21 2006 /n/sources/patch/applied/nothing couldn't resist d-rwxrwxr-x M 104260 rsc sys 0 Feb 22 2006 /n/sources/patch/applied/nv-vga-mmio While reading that file, I realize that getpcixdid needs nv->mmio to be inited. At least, it seems the patch I provide doesn't change the behavior (tested with my GeForce 5200). Priority: Low d-rwxrwxr-x M 104260 rsc sys 0 Feb 22 2006 /n/sources/patch/applied/igbe-no-trailing-space remove whitespace at ends of lines to simplify diffs. d-rwxrwxr-x M 104260 rsc sys 0 Feb 23 2006 /n/sources/patch/applied/hmac-init allow the caller to provide the DigestState structure, memset to 0, and have it work, as with the other digest functions d-rwxrwxr-x M 104260 jmk sys 0 Feb 25 02:33 /n/sources/patch/applied/ether82557 Adding support to Pro 100 VE as found in Asus Pundit PE3d-rwxrwxr-x M 104260 rsc sys 0 Feb 26 05:42 /n/sources/patch/applied/cursorswitch-man cursorswitch(2) seems to be superseded by setcursor(2) d-rwxrwxr-x M 104260 rsc sys 0 Feb 26 22:33 /n/sources/patch/applied/tapefs-g8byte-parens add missing parens around `x' in g8byte body. d-rwxrwxr-x M 104260 rsc sys 0 Feb 26 23:59 /n/sources/patch/applied/tapefs-41bsdfs I created this a while ago; it's a trivial change from 32vfs.c and you might want to (somehow) fold it into 32vfs. I'm pretty sure I needed it to read a 4.1 file system and 32vfs wouldn't do the job. d-rwxrwxr-x M 104260 rsc sys 0 Feb 27 00:01 /n/sources/patch/applied/tapefs-util-fix These are changes I've been running with for a long time. I think the intent is produce a more meaningful error message when there's a file type difference. d-rwxrwxr-x M 104260 rsc sys 0 Feb 27 10:22 /n/sources/patch/applied/tarfs-bytes g8byte -> b8byte d-rwxrwxr-x M 104260 rsc sys 0 Mar 1 01:41 /n/sources/patch/applied/ql-hunk change the allocation strategy to match that of other ?l: if more and more space is needed, allocate in larger chunks d-rwxrwxr-x M 104260 rsc sys 0 Mar 1 01:41 /n/sources/patch/applied/ql-range the instructions added after AEND require ALAST to be used instead of AEND for the oprange table (and the bad op check in list.c) d-rwxrwxr-x M 104260 rsc sys 0 Mar 1 01:42 /n/sources/patch/applied/32vfs-delete-unused-buffer delete buffer found to be unused by the compiler. d-rwxrwxr-x M 104260 rsc sys 0 Mar 1 02:17 /n/sources/patch/applied/trace-man-ctl-msg typo d-rwxrwxr-x M 104260 rsc sys 0 Mar 3 03:25 /n/sources/patch/applied/pci-buslogic add buslogic/mylex entry, plus subentry for 958 scsi host adaptor. d-rwxrwxr-x M 104260 rsc sys 0 Mar 3 03:30 /n/sources/patch/applied/mga4xx-cursor bad addr in cursor load ... d-rwxrwxr-x M 104260 rsc sys 0 Mar 3 10:49 /n/sources/patch/applied/gif-tolerance gif now more tolerant to mal-formed files, and no longer sucides on error. d-rwxrwxr-x M 104260 rsc sys 0 Mar 7 01:26 /n/sources/patch/applied/webfs-http-c-with-405-corrected-typo Fixed small typo - sorry for the duplication. -- Marina Brown . d-rwxrwxr-x M 104260 rsc sys 0 Mar 7 01:30 /n/sources/patch/applied/drawterm-win-mouse-wheel-and-unicode-support +mouse wheel support +unicode support rs_rlab@mail.ru d-rwxrwxr-x M 104260 rsc sys 0 Mar 9 23:15 /n/sources/patch/applied/expect-man typo d-rwxrwxr-x M 104260 rsc sys 0 Mar 9 23:17 /n/sources/patch/applied/webfs-usage-reports-wrong-name d-rwxrwxr-x M 104260 rsc sys 0 Mar 9 23:18 /n/sources/patch/applied/marshal-inline-c9x another c9x inline keyword woes ... d-rwxrwxr-x M 104260 rsc sys 0 Mar 9 23:19 /n/sources/patch/applied/eqn-inline-c9x 'inline' is now a C keyword d-rwxrwxr-x M 104260 rsc sys 0 Mar 11 01:44 /n/sources/patch/applied/awkbugs.again [resubmitted having pulled from sources and fixed careless mistake caused by manually entering a change that i'd made on another system, without testing it.] for the record, these are the changes: printf("%c", 0); could produce unexpected results (it read uninitialised memory) the utf function was in the wrong place, so never worked. d-rwxrwxr-x M 104260 rsc sys 0 Mar 11 03:33 /n/sources/patch/applied/riotopped.delete two unrelated little fixes: bottomwin: if bottom was used a few times, rio lost track of the window ordering, because w->topped was zero for several windows. setting it to a lower (negative) number each time fixes that. deleting windows: a change so that if you point to a window to delete it, releasing the button off the window aborts the action. yes, it is possible to do this by chording with another button, but i feel this way is more intuitive and less easy to muck up (important when you might lose important info). BTW, i think the cursor when deleting a window should be distinguished, as it was in 8½ - i'll submit a patch to reinstitute the old bones cursor, if it's not considered a bad idea. d-rwxrwxr-x M 104260 rsc sys 0 Mar 11 04:31 /n/sources/patch/applied/xigh-face nothing special... d-rwxrwxr-x M 104260 rsc sys 0 Mar 11 23:15 /n/sources/patch/applied/vgadb-2001fp add dell2001fp. i'm not sure that this worked, but i believe it's derived from the X11 modeline. d-rwxrwxr-x M 104260 rsc sys 0 Mar 12 23:07 /n/sources/patch/applied/rc-lexer out of range d-rwxrwxr-x M 104260 rsc sys 0 Mar 12 23:14 /n/sources/patch/applied/man-acme-addr-correction Brings acme(4)'s description of reading from the addr file in line with the code. d-rwxrwxr-x M 104260 rsc sys 0 Mar 12 23:15 /n/sources/patch/applied/tar-2-wstats split time setting from gid setting via wstat. d-rwxrwxr-x M 104260 rsc sys 0 Mar 14 22:16 /n/sources/patch/applied/rdbfs-exit exit gracefully on echo exit>/proc/pid/ctl remove old code d-rwxrwxr-x M 104260 jmk sys 0 Mar 18 03:00 /n/sources/patch/applied/iiyama-pro400 iiyama Vision master 400 monitor d-rwxrwxr-x M 104260 jmk sys 0 Mar 18 23:58 /n/sources/patch/applied/vgadb-cut-paste fix a dreg leftover from cutting and pasting. d-rwxrwxr-x M 104260 jmk sys 0 Mar 18 23:59 /n/sources/patch/applied/nfsserver-8-opts Clarify that -n is a no-op; describe -s and -t. d-rwxrwxr-x M 104260 rsc sys 0 Mar 20 22:46 /n/sources/patch/applied/http-bug httpd suicides occasionaly if simple log files 8c found it - p used and not set. d-rwxrwxr-x M 104260 rsc sys 0 Mar 24 00:06 /n/sources/patch/applied/games-4s-scorefiles It's about file permissions. Highscore files are only writable for members of sys, but '4s' and '5s' want to open it read-write. d-rwxrwxr-x M 104260 rsc sys 0 Mar 24 00:14 /n/sources/patch/applied/win-del-delete Make Del in win always close the window. d-rwxrwxr-x M 104260 rsc sys 0 Mar 24 03:11 /n/sources/patch/applied/html-char_entities a more complete set of character entities. fgb d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 02:22 /n/sources/patch/applied/gif-robustness I opened the door to grotty (virused ?) gifs too far, this closes it a little. d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 02:23 /n/sources/patch/applied/tcs-html-chars more html character entities for tcs(1). fgb d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 21:37 /n/sources/patch/applied/vc-l cc.h: add `little' flag; lex.c: recognise -l, sort the list of options; swt.c: don't do big-endian adjustment under -l. d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 21:37 /n/sources/patch/applied/0c simulate 0c with vc -l (can't use -L, it's used but undocumented). d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 22:35 /n/sources/patch/applied/vl-little add -L for generating little-endian code. this is a merger of anyrhine's 0l (he calls it pl) and vl. i'm not convinced that it generates the correct byte order for all header types, but it's what he's been using. i also have no convenient way to test it. d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 22:36 /n/sources/patch/applied/va-little add -L to generate little-endian code. d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 22:55 /n/sources/patch/applied/0l simulate old 0l command with vl -L d-rwxrwxr-x M 104260 rsc sys 0 Mar 25 22:55 /n/sources/patch/applied/0a simulate old 0a command with va -L d-rwxrwxr-x M 104260 rsc sys 0 Mar 26 23:42 /n/sources/patch/applied/libhtml-pretag '\n' aren't ignored inside 
 anymore,
tested with abaco and htmlfmt(1).

Federico
d-rwxrwxr-x M 104260 rsc sys 0 Mar 29 01:06 /n/sources/patch/applied/tweak-hex-noreturn
warning: tweak.c:538 no return at end of function: tohex
d-rwxrwxr-x M 104260 rsc sys 0 Mar 29 01:11 /n/sources/patch/applied/mga-hwscroll-fix2
hwscroll fix
cursor fix
code cleanup
d-rwxrwxr-x M 104260 rsc sys 0 Mar 30 07:37 /n/sources/patch/applied/missing-return-stmts
warning: 5db.c:493 no return at end of function: armcondpass
warning: 5db.c:555 no return at end of function: armshiftval
warning: 5db.c:592 no return at end of function: armmaddr
warning: x509.c:2466 no return at end of function: tagdump
warning: smbbuffer.c:402 no return at end of function: smbbuffergetv
warning: acidleak.c:104 no return at end of function: finddata
warning: disksim.c:433 label declared and not used "next"
warning: conv.c:218 no return at end of function: Jconv
warning: mmc.c:563 no return at end of function: mmcsetbs
warning: iobuf.c:143 no return at end of function: xread
warning: db.c:115 no return at end of function: lockopen
warning: /sys/src/cmd/upas/smtp/smtpd.y:298[smtpd.tab.c:155] no return at end of function: cat
warning: message.c:120 no return at end of function: getaddr
d-rwxrwxr-x M 104260 rsc sys 0 Mar 30 07:47 /n/sources/patch/applied/uname-configure
make uname() return values that gnu configure can understand
d-rwxrwxr-x M 104260 rsc sys 0 Mar 30 08:00 /n/sources/patch/applied/tcs-tune
This patch extends tcs(1) to understand TUNE.
http://www.tunerfc.tn.gov.in

TUNE is a 16 bit encoding scheme for Tamil.
Unlike Unicode, TUNE maps each character to a
single glyph.  This property of TUNE makes it
straight forward to view/edit Tamil in Plan 9.

Please use unicode.8.font from
http://www.tip9ug.jp/who/art/lib/font/madhavi.tgz
to view the Tamil glyphs in tune.c.

Sample text to test tune.c:
http://www.tip9ug.jp/who/art/lib/kp-tune.utf
http://www.tip9ug.jp/who/art/lib/kp-uni.utf

Pointers to TUNE and Unicode Tamil code charts are
available in the Links section in
http://www.tip9ug.jp/who/art/tamil

Please let me know if any other information is
needed.

Thanks.

Arvindh
d-rwxrwxr-x M 104260 rsc sys 0 Mar 30 22:01 /n/sources/patch/applied/tcs-tune-2
- removed enum state from tune.c
- fixed syntax error in tune.c:227

thanks.
arvindh
d-rwxrwxr-x M 104260 rsc sys 0 Mar 30 22:01 /n/sources/patch/applied/tune-paren
remove extra left paren from EPR invocation to eliminate syntax error.
d-rwxrwxr-x M 104260 rsc sys 0 Mar 30 22:01 /n/sources/patch/applied/uname-sysname
make uname actually set sysname to Plan9 if no /env/osname exists.
d-rwxrwxr-x M 104260 rsc sys 0 Mar 31 23:10 /n/sources/patch/applied/plumb-http-path-allow-dots
Dots are valid path/dir elements in HTTP
d-rwxrwxr-x M 104260 rsc sys 0 Mar 31 23:12 /n/sources/patch/applied/ndb-8-format
fix formatting around examples
d-rwxrwxr-x M 104260 rsc sys 0 Mar 31 23:13 /n/sources/patch/applied/mach.h-spim
add ObjSpim back for .0 files (little-endian mips).
d-rwxrwxr-x M 104260 rsc sys 0 Apr 1 07:38 /n/sources/patch/applied/indiana-dst
indiana starts honoring DST on sunday 2 apr 2006.

- erik
d-rwxrwxr-x M 104260 rsc sys 0 Apr 2 06:29 /n/sources/patch/applied/libmach-0
since we're recompiling the world...
i should have included these libmach changes with the
[0v][cal] changes.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 2 06:31 /n/sources/patch/applied/2007timezones
fix DST start and end dates for 2007-; the new changeover dates are
	starting	2d sun in march
	ending	1st sun in nov

- erik
d-rwxrwxr-x M 104260 rsc sys 0 Apr 2 06:32 /n/sources/patch/applied/ape-ls-l-fix
Was using ape sed with native regexps, changed to native sed.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 2 06:45 /n/sources/patch/applied/mga550-native
Add support for MGA550 to the mga4xx native driver. Do some small cleanups and fix a silly bug while at it.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 2 07:19 /n/sources/patch/applied/upas-tcs-2047
there are two parts to this patch

1.  	remove all the character-set tables from fs.c and mbox.c; 
	allow tcs to do the translations

2.  	the rfc-2047 decoding was improved. adjacent 2047-encoded
	"atoms" are pasted together as per the rfc.

i also tightened up the processing of 2047 strings.  the code had been scanning
the insides of 2047 strings for the start of a new 2047-string.

- erik
d-rwxrwxr-x M 104260 rsc sys 0 Apr 2 12:03 /n/sources/patch/applied/upas-2047
this allows 2047-encoded atoms sepearted by only whitespace to be pasted together,
as the rfc requires.

the usage string was corrected and upas/fs will now complain about extra arguments.

i'm including the test examples from rfc-2047 as rfc2047-test
d-rwxrwxr-x M 104260 rsc sys 0 Apr 2 21:05 /n/sources/patch/applied/iwhois-dejpcnin
This patch fixes *.de lookups and adds them for *.jp, *.cn and
*.in.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 4 20:33 /n/sources/patch/applied/unix-readme
update drawterm's description in /sys/src/cmd/unix/README.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 4 20:33 /n/sources/patch/applied/disksim-man-mbr
disksim example missing disk/mbr which disk/fdisk requires.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 4 20:41 /n/sources/patch/applied/drawterm-tls
add TLS (devtls and pushtls plus encodefmt); also create Make.osx
and fix up devaudio-none for testing on OS X 10.4: ls '#a'
shows '#a/tls'; I haven't exercised it any further.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 8 00:47 /n/sources/patch/applied/setjmp-man
fix pathname
d-rwxrwxr-x M 104260 rsc sys 0 Apr 8 00:47 /n/sources/patch/applied/consolefs-pp
fix consolefs(4)'s formatting by adding a .PP
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 00:20 /n/sources/patch/applied/readtimezone-fd-leak
always close fd `i' in readtimezone, even if there's an error.
also eliminate a bunch of gotos.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 00:42 /n/sources/patch/applied/whitelist-format
changed whitelist format to be:
ipaddress      com.domain.subdomain
This allows simpler maintence, sort +1 of the file groups all the smtpXX.gmail.com 
entries together so I can take a guess at a CIDR mask for them. Gmail is a bad example
as their spf data gives me all I need but others are less helpfull.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 01:01 /n/sources/patch/applied/dnstcp-norecursion
still getting DNS poisioning I think - some spam appearing from sites which
appear to have domains when the spam arrives but which has disappeared
when I look later. Iven if I am missunderstanding the problem this is a
more complete fix.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 01:09 /n/sources/patch/applied/kmem-man-index
Add kmem to 1/INDEX and under NAME in leak(1).

And *PLEASE* could we have the lookman index re-built daily? it takes just a few seconds to rebuild, and it is a pain in the ass when looking for things and you can't find them because turns out the indexes are out of date.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 01:10 /n/sources/patch/applied/webfs-redir-contenttype
Provide correct contenttype after a redirect. Needed by abaco.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 01:11 /n/sources/patch/applied/rc-1-m
document -m; correct formatting of -c.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 01:32 /n/sources/patch/applied/ps-start-run-time
As Russ suggested in the 'patch/list sorry/proc-mtime' thread on 9fans here is a modified ps.
A -s flag prints process local start time.
Not sure about the -r option, maybe two extra options to ps is over kill?

Cheers,
John

d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 01:32 /n/sources/patch/applied/drawterm-tls-old-compilers
vc6 & vc7 can't compile [] initialisation syntax.
tweaked devtls.c similarly to what was done for devssl.c.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 02:27 /n/sources/patch/applied/cc-varadic-macros
this patch enables support for c99 varadic macros for ?c. the best example of use
would be

	#define debug(...)	fprint(2, __VA_ARGS__)
or
	#define debug(...)	
or
	#define xyzdebug(...)	if(xyz->debug) fprint(2, ___VA_ARGS__)

d-rwxrwxr-x M 104260 rsc sys 0 Apr 11 02:53 /n/sources/patch/applied/cpp-varadic-macros
this patch should be sufficient for full-blown c99 varadic macro support.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 13 14:11 /n/sources/patch/applied/usb-lib-errstr
only close open fds, thus avoiding overwriting the errstr,
thus giving better messages.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 14 23:00 /n/sources/patch/applied/patch-man
Corrected patch/create usage in the EXAMPLES
section.  Email argument was missing.

Thanks.
Arvindh
d-rwxrwxr-x M 104260 rsc sys 0 Apr 14 23:01 /n/sources/patch/applied/httpd-readfile-nul
NUL-terminate the string returned by readfile().
d-rwxrwxr-x M 104260 rsc sys 0 Apr 14 23:01 /n/sources/patch/applied/diffy-man
Added path to diffy's source.

Thanks.
Arvindh
d-rwxrwxr-x M 104260 rsc sys 0 Apr 16 21:59 /n/sources/patch/applied/nfsserver-more-free
crank up the statfs numbers to the max for those systems (e.g., OS X)
that pay attention to them.  cp on OS X will refuse to copy into
an nfs partition that appears to not have enough space.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 16 22:04 /n/sources/patch/applied/win-eventsize
	for(i in `{seq 0 5467})
		echo -n ～ >>/tmp/win.test # ～ == U+FF5E

Acme win hangs if I cat the file created above.

This patch restricts the event buffer size to 8192 bytes
in fs.c:/^fswrite.

Thanks.
Arvindh
d-rwxrwxr-x M 104260 rsc sys 0 Apr 16 22:04 /n/sources/patch/applied/win-pending-s
Acme win hangs when I cat a file containing bad
UTF sequences.  Sample file (runeerrors.txt)
included.

Fixed pendingS calculation.  Terminal partial
runes, if any, are actually Runeerrors and
shouldn't be ignored using utfnlen().

Thanks.
Arvindh
d-rwxrwxr-x M 104260 jmk sys 0 Apr 20 00:15 /n/sources/patch/applied/devpccard-semi
old plan 9 c compilers reject line 34, which has two
semi-colons, with

: bootes; mk
8c -FVw -I. devpccard.c
devpccard.c:35 syntax error, last name: ulong
devpccard.c:39 name not declared: ioaddr
devpccard.c:41 name not declared: ioaddr
d-rwxrwxr-x M 104260 jmk sys 0 Apr 24 21:27 /n/sources/patch/applied/unicode-man
fixed unicode usage.
d-rwxrwxr-x M 104260 jmk sys 0 Apr 24 21:30 /n/sources/patch/applied/tcs-mime-windows-1252
fix typo in windows-1252 mime charset name. 
d-rwxrwxr-x M 104260 jmk sys 0 Apr 24 21:30 /n/sources/patch/applied/9pcon-8-...
finish the final sentence of DESCRIPTION in 9pcon(8).
d-rwxrwxr-x M 104260 jmk sys 0 Apr 24 21:31 /n/sources/patch/applied/exportfs-p
add -P to exportfs's usage message; correct exportfs -P example in man page.
d-rwxrwxr-x M 104260 jmk sys 0 Apr 24 21:38 /n/sources/patch/applied/jpg_doc
Spelling fix
d-rwxrwxr-x M 104260 jmk sys 0 Apr 24 21:39 /n/sources/patch/applied/timesync-typo
Tiny typo on timesync's manpage.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 26 23:18 /n/sources/patch/applied/ape_hdrs
An incomplete attempt at rationalising the non-Posix/ANSI
header files, required by a fresh Zlib and Libpng
implementation
d-rwxrwxr-x M 104260 rsc sys 0 Apr 26 23:19 /n/sources/patch/applied/nfsserver-opts-fix
per russ's suggestion, fix the option parsing of the nfsserver suite.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 27 20:29 /n/sources/patch/applied/pcc-spim
These changes should have been included with earlier spim
support.  Add spim and 68000 objtypes and flesh out allos.
d-rwxrwxr-x M 104260 rsc sys 0 Apr 29 22:32 /n/sources/patch/applied/strings-valid-rune
strings didn't recognize Runeerror something other than a utf character, in
contradiction of the man page (and utility.)
d-rwxrwxr-x M 104260 rsc sys 0 Apr 29 22:33 /n/sources/patch/applied/lp-options
flagfmt didn't mention the -p or -M flags. i was getting
a usage message from {echo hello|lp}. i can't see why it hasn't
already been noticed though, which makes me suspicious that
this fix might be wrong. here it is anyway, FWIW.
d-rwxrwxr-x M 104260 rsc sys 0 May 3 00:02 /n/sources/patch/applied/utf-runeerror
out-of-date reference to 0080 used as Runeerror.  use fffd instead.
d-rwxrwxr-x M 104260 rsc sys 0 May 12 22:07 /n/sources/patch/applied/faces-times
try to make time displayed approximately correct by synchronising
date-and-time update with minute roll-over.  currently it's a minute
late about half the time, on average.
d-rwxrwxr-x M 104260 rsc sys 0 May 12 22:08 /n/sources/patch/applied/cron-accuracy
various changes: remove some magic numbers, break some
long lines, remove unwanted USED(argc,argv), fix the
`don't go mad' test (was off by a factor of 60), permit
jobs to be initiated each minute, not each two minutes,
and initiate them starting in the first second of the minute.

tested and it seems to work.
d-rwxrwxr-x M 104260 rsc sys 0 May 12 22:08 /n/sources/patch/applied/smtp-8-names
update the greylist file names.  also touch up the formatting.
d-rwxrwxr-x M 104260 rsc sys 0 May 12 22:09 /n/sources/patch/applied/smtpd-dot-in-helo
require a dot in helo/ehlo strings not coming from a trusted
(ournet) machine.  This lets broken microsoft mail clients
submit to the local smtp server, but stops a lot of spam.
d-rwxrwxr-x M 104260 rsc sys 0 May 16 00:17 /n/sources/patch/applied/file-update
here are some updates to file(1).  there are some taste decisions to be made, i'll
resubmit if i got it wrong.

1. beef up latin1 support.  0xa0-0x100 were previously not recognized as Latin.
    also, calling chartorune before anything else caused character like "ä" (in latin1)
    to be parsed as a null.  (try echo ä | tcs -t latin1 > junk ; file junk)
    i also changed "ascii latin" to "Latin" because i kept thinking "ascii latin" was 
    latin1, not utf-encoded codepts that are in latin1.

    i added the -h flag for debugging and extra information.  file reports that /lib/glass
    is binary because it exceeds the "language limit".  i added the flag to figure out what
    was really going on.  remove it if it's not useful.
2. try iself before long0 which calls crackheader.  iself is more informative, and
    it also results in more uniform results for elf files. (i was getting "elf executable"
    for some executables and "i386 ELF executable" for everything else.)
3. allow fonts to contain an offset and to be larger than 6000 bytes.
    code2000, which has near-full coverage, is just a little bit bigger than 6k.
    (there are several fonts in /lib/font/bit/*/*.font that need fixing.)
4. added support for old-style faces.
5. added support for XPM and ppm images.
d-rwxrwxr-x M 104260 rsc sys 0 May 20 13:07 /n/sources/patch/applied/libframe-alpha
this patch allows libframe to take advantage of accellerated stringbg drawing.
d-rwxrwxr-x M 104260 rsc sys 0 May 22 02:40 /n/sources/patch/applied/dt.devhostopen
Added /dev/hostopen to replace winplumb stuff. also changes to compile using cygwin+mingw
d-rwxrwxr-x M 104260 jmk sys 0 May 22 23:31 /n/sources/patch/applied/dt-getcallpcamd64
missing ulong -> uintptr change
d-rwxrwxr-x M 104260 rsc sys 0 May 24 23:37 /n/sources/patch/applied/file-elf
prevent uninitialized memory access by declaring tab array static and
classify elf file by type (object, executable, shared library, core dump)
d-rwxrwxr-x M 104260 rsc sys 0 May 24 23:37 /n/sources/patch/applied/file-utf-16
support for utf-(16|32)(be|le)
d-rwxrwxr-x M 104260 rsc sys 0 May 24 23:54 /n/sources/patch/applied/utfecpy-boundary
utfecpy() ignores any valid UTF sequence
at the end of its input.

for example:
	utfecpy(to, to+5, "1234567890");
copies only 3 bytes while
	strecpy(to, to+5, "1234567890");
copies 4.  also,
	utfecpy(to, to+7, "123க456");
copies only the first 3 bytes.

this patch replaces utfecpy() with a modified
version of /acme/bin/source/win/main.c:/^utfncpy

thanks.
arvindh
d-rwxrwxr-x M 104260 rsc sys 0 May 24 23:56 /n/sources/patch/applied/smtpd-trust-inside-sender
Only run validatesender if the client is untrusted.
This means that validatesender will get run once when
the message arrives from outside but not as relayed
internally and run through spam checking.
Previously we were running validatesender twice for
each incoming message, and the second one was redundant.
d-rwxrwxr-x M 104260 rsc sys 0 May 31 23:00 /n/sources/patch/applied/smtpd-helo-letter
for untrusted clients, require at least one ASCII letter in
the helo/ehlo argument, unless it's an address-literal.
this rejects

	helo 192.168.1.1

among others.  It still accepts the legal form

	helo [192.168.1.1]

but that's not what spammers send.

I've been running with this change overnight with no problems.
d-rwxrwxr-x M 104260 rsc sys 0 May 31 23:01 /n/sources/patch/applied/replica-cd-fossil
Update /dist/replica/cd to allow a `pull' from a Plan 9 iso
on CD to a fossil file server. The relevant sections from
/dist/replica/network have been inserted, and the file has
been made executable.
d-rwxrwxr-x M 104260 jmk sys 0 Jun 11 23:37 /n/sources/patch/applied/xls2txt-reliable
Manual page update - bugs fixed long ago.
d-rwxrwxr-x M 104260 jmk sys 0 Jun 11 23:38 /n/sources/patch/applied/82801dbl-southbridge
Add yet another Intel south bridge.
This one exists in the Toshiba Sattellite 2435-S255.
d-rwxrwxr-x M 104260 jmk sys 0 Jun 11 23:39 /n/sources/patch/applied/man-9p
9p's man page failed to include listensrv/threadlistensrv
d-rwxrwxr-x M 104260 jmk sys 0 Jun 11 23:43 /n/sources/patch/applied/varargchkverbs
handle "u" and "b".
d-rwxrwxr-x M 104260 jmk sys 0 Jun 11 23:53 /n/sources/patch/applied/strtochan3
resubmitted as requested.
d-rwxrwxr-x M 104260 jmk sys 0 Jun 13 21:26 /n/sources/patch/applied/consoles
make sure that we don't attempt to read from /dev/null on chat consoles
d-rwxrwxr-x M 104260 rsc sys 0 Jun 20 22:54 /n/sources/patch/applied/conswdir-restore-dirs
Save and restore label and wdir on startup and exit.
I tend to hop on and off remote machines quite a bit.
d-rwxrwxr-x M 104260 jmk sys 0 Jul 7 06:51 /n/sources/patch/applied/cron-fix
this fixes some latent bugs, notably coping with
time jumping backward and with very fast processors
(amd64s).  andrey and i have been running this for
about a day and it's been behaving.
d-rwxrwxr-x M 104260 geoff sys 0 Jul 13 03:52 /n/sources/patch/applied/usbuhci-iso-deadlock
There's a deadlock in isoio() on a multiprocessor, arising  from a
conflict in order of acquiring locks.  (Not just theory: I've seen
it happen on my dual-processor Epia system.)  Consider this scenario:

isoio() on cpu1:
1 isoio calls sleep(&e->wr, isoready, e) /sys/src/9/pc/usbuhci.c:1173
2	sleep calls lock(&e->wr) /sys/src/9/port/proc.c:740
3	sleep calls isoready(e) /sys/src/9/port/proc.c:755
4		isoready calls ilock(&ctlr->activends) /sys/src/9/pc/usbuhci.c:1093
5		isoready calls iunlock(&ctlr->activends) /sys/src/9/pc/usbuhci.c:1095
6	sleep calls unlock(&e->wr) /sys/src/9/port/proc.c:762

interrupt() on cpu0:
7 interrupt calls ilock(&ctlr->activends) /sys/src/9/pc/usbuhci.c:1043
8 interrupt calls cleaniso(e, ) /sys/src/9/pc/usbuhci.c:1052
9	cleaniso calls wakeup(&e->wr) /sys/src/9/pc/usbuhci.c:1015
10		wakeup calls lock(&e->wr) /sys/src/9/port/proc.c:864
11		wakeup calls unlock(&e->wr) /sys/src/9/port/proc.c:876
12 interrupt calls iunlock(&ctlr->activends) /sys/src/9/pc/usbuhci.c:1055

An interleaving of events containing 2 - 7 - 10 - 4 leads to deadlock.

I developed a fix which involved locking the individual endpoints instead
of locking ctlr->activends, but when I checked it with Sape he pointed
out that a simple revision to isoreadyx() would mean that it can be
called without locking at all.  So here's a simple patch based on Sape's
suggestion.  (Tested by playing audio for 24 hours on the above dual-
processor system, without incident.)

-- Richard Miller
d-rwxrwxr-x M 104260 rsc sys 0 Jul 16 02:00 /n/sources/patch/applied/i82557-0x103d
This adds support for the Intel 82562 PRO/100 VE, in a newer
revision, that is built into X31 Thinkpads.

d-rwxrwxr-x M 104260 rsc sys 0 Jul 16 02:00 /n/sources/patch/applied/parsereq-fix
fixed abspath() to not filter space chars in url. they can be part of a file name.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:48 /n/sources/patch/applied/nforce4-sata
I've added the device IDs for the nForce4 SATA devices. It's 
been tested on an Asus A8N5X with 2 drives on each of the 
2 devices. DMA seems to be working fine.

The nVidia device IDs in the boot code have been brought into 
sync with the kernel.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:49 /n/sources/patch/applied/boot-devsd-probe
The code to walk through the list of discovered devices 
updates the index of the first device (sdev) when it should 
be updating the index of the currect device (tail).
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:52 /n/sources/patch/applied/atofman
update the bugs section to mention atoll.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:53 /n/sources/patch/applied/man-4-preproc
[128 files]
small typo corrected. "| |" always yields no output.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:55 /n/sources/patch/applied/aquarela-delete
This patch makes aquarela respond to delete requests from XP clients.
Previously, aquarela would perform the delete but would not respond to the client, resulting in a timeout.
The content of the response message is irrelevant, as it is ignored by the client.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:58 /n/sources/patch/applied/test-invalid-args-and-man-page
Make test(1) complain when we pass unexpected extra args
instead of silently ignoring them or returning TRUE.
This is a bit more consistent with PoSix systems and
avoids confusion (fgb thought that test -dw /tmp would
check if /tmp was a writable dir).

Also complain when the argument to -t is not a valid
filedes.

Fix small typo in the man page and document that there
is on way to test if a dir is writeable.

uriel
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:58 /n/sources/patch/applied/libhtml-tableak
fix a few leaks.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 14:59 /n/sources/patch/applied/subfont-leak
fix leak.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:02 /n/sources/patch/applied/webfs-contenttype
fix post form's content type; add support for image/png and image/ppm.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:02 /n/sources/patch/applied/webfs-shortread
webfs was truncating pages.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:03 /n/sources/patch/applied/manrio-typo-enq
typo in rio(1): ENQ is Ctrl-E, not Ctrl-F
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:04 /n/sources/patch/applied/doc-man-8-fossilcons-users-r
Add missing -r.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:10 /n/sources/patch/applied/replica-applychanges-fix-usage
Correct usage message for replica/applychanges.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:16 /n/sources/patch/applied/plumb-basic-quote-hget-args
Make sure we quote the args for hget to keep rc from interpreting special chars in the URL.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:23 /n/sources/patch/applied/needsrcquote
fix inadvertent quoting of runes by needsrcquote.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:24 /n/sources/patch/applied/webfs-httpread-too-smart
this fix httpread trying to be too smart and ocationating problems.
"webfsget https://mail.google.com/mail/help/images/logo1.gif" is an example.

federico
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:27 /n/sources/patch/applied/hoc-rune
allow hoc to accept utf-8 variables.  this patch is a bit noisy because i changed "int c" to "int r", and 'x'
to L'x' in hoc.y but the basic change is from Bgetc to Bgetrune.
d-rwxrwxr-x M 104260 rsc sys 0 Aug 12 15:30 /n/sources/patch/applied/mga4xx-newmodes-resub
2006-08-07: Patches for the Matrox 4xx driver to support 16 an 24 bits modes (in addition to the existant 8 and 32). Also the G200 cards now work with hardware acceleration.

2006-08-12: Resubmit of the previous patch with corrected indentation, tabs are now used instead of hardcoded space (sorry about that one).

Leonardo Valencia