| 01:24.42 | *** join/#gnu-kbsd xsun (n=xsun@20150078090.user.veloxzone.com.br) |
| 01:42.24 | CIA-12 | 03aurel32 * r1173 10/trunk/glibc-2.3-head/debian.sh: Support for glibc 2.3.6 |
| 02:16.39 | *** join/#gnu-kbsd asac_ (n=asac@debian/developer/asac) |
| 05:16.33 | braindmg | hey guys check today dpkg --version, and check the same tomorrow after dinstall =P |
| 05:17.54 | braindmg | (hmm or well today after dinstall ;) |
| 05:28.34 | tarzeau | hmm |
| 05:28.41 | tarzeau | what should i notice about that? |
| 05:47.21 | nyu | braindmg: how about p7zip ? :) |
| 06:02.48 | braindmg | tarzeau: you didn't notice anything? ;) |
| 06:02.54 | braindmg | nyu: I've to review it first |
| 06:03.04 | braindmg | nyu: there's more important stuff on the queue now =) |
| 06:03.19 | braindmg | nyu: but I'll try to put it in before the release |
| 06:03.25 | braindmg | so it can be used for etch+1 |
| 06:03.43 | braindmg | nyu: ok I've all the patches split now |
| 06:04.18 | braindmg | I'll start reviewing one by one, rediffing, building and submitting ... |
| 06:13.19 | tarzeau | braindmg: i didn't check |
| 06:18.40 | nyu | braindmg: yeah, that's the essential |
| 06:18.56 | nyu | if it's not in etch, we'll have to wait for etch+2 ;) |
| 06:22.23 | braindmg | (the patches I'm talking about now are for util-linux ...) |
| 06:22.52 | braindmg | (the split ones) |
| 06:24.31 | nyu | weird.. Thomas Dickey just replied to me in a Gentoo bug |
| 06:24.33 | nyu | https://bugs.gentoo.org/show_bug.cgi?id=121637 |
| 06:24.45 | nyu | I think he wants to know GNU/kFreeBSD's ttynames |
| 06:24.52 | nyu | braindmg: ah, good! |
| 06:25.11 | nyu | braindmg: is the ioctl one included? |
| 06:25.27 | nyu | + /* |
| 06:25.27 | nyu | + * TIOCSCTTY: steal tty from other process group. |
| 06:25.27 | nyu | + */ |
| 06:25.27 | nyu | + ioctl(0, TIOCSCTTY, 1); |
| 06:25.59 | nyu | note that it's been moved |
| 06:26.04 | nyu | want me to split that? |
| 06:26.26 | nyu | uhm actualy i can't, it depends on the termios hunk |
| 06:26.42 | braindmg | I think I never synced with your patch |
| 06:27.03 | braindmg | my current patches: www.hadrons.org/~guillem/tmp/ |
| 06:30.57 | braindmg | nyu: they are not yet rediffed, not fully reviewed ... |
| 06:34.25 | nyu | braindmg: I'll sent you the part you're missing as split patch |
| 06:34.30 | nyu | s/sent/send/g |
| 06:35.52 | nyu | braindmg: wrt http://www.hadrons.org/~guillem/tmp/util-linux-2.12r_sys_user.h.patch, why don't you use #elif ? |
| 06:36.11 | nyu | or just sys/user.h |
| 06:37.41 | nyu | braindmg: your patches, taken as a whole, as much more intrusive than what we have in svn. have you considered sending them in separate "waves" ? |
| 06:37.42 | braindmg | because the test is to check if that file exists, and I just copy pasted what was on mkswap.c |
| 06:37.50 | nyu | so that at least the critical ones are accepted |
| 06:38.09 | braindmg | nyu: those patches are the original ones that were introduced in the util-linux debian package |
| 06:38.41 | braindmg | and then reverted later as massive hunk rejection and lamont didn't wanted to fix it |
| 06:38.55 | braindmg | nyu: all are critical |
| 06:39.04 | braindmg | otherwise the desired apps do not build |
| 06:39.12 | braindmg | nyu: and that's why I've splitted them |
| 06:39.18 | nyu | how receptive is the maintainer? |
| 06:39.27 | braindmg | the old patch was a monolithic beast of 120 KiB |
| 06:39.42 | nyu | in http://www.hadrons.org/~guillem/tmp/util-linux-2.12r_termios.patch, the second line in second hunk is unnecessary (since you have a kludge to define TAB3 above) |
| 06:39.45 | braindmg | nyu: he should be receptive, he is adrian bunk, old ex maint |
| 06:39.56 | nyu | I know.. |
| 06:40.12 | nyu | never treated with him |
| 06:40.18 | nyu | could be like the sysvinit guys or like Md |
| 06:40.50 | nyu | as for TAB3/TABDLY, can't these be added for GNU/Hurd ? |
| 06:40.53 | braindmg | (old ex devel sorry) |
| 06:41.03 | braindmg | nyu: no, roland already said those would not be added |
| 06:41.18 | braindmg | didn't I give you the webmail pointer? |
| 06:41.22 | nyu | Roland maintains the debian package? :P |
| 06:41.33 | braindmg | nyu: roland maintains the glibc dude ;) |
| 06:41.59 | nyu | precisely, not the debian glibc |
| 06:42.09 | braindmg | nyu: and all this is glibc and util-linux upstream, debian does not matter at all |
| 06:42.39 | nyu | have you considered doing "#define termio termios" ? |
| 06:42.47 | nyu | this makes the patch much smaller (look at the svn version) |
| 06:42.57 | braindmg | nyu: yeah I saw, but I don't like it ;) |
| 06:43.13 | braindmg | nyu: and the thing is that the code is using termios in other places |
| 06:43.22 | braindmg | so I don't see a reason to not migrate cleanly |
| 06:43.39 | nyu | it uses termios only, without fallback? |
| 06:43.43 | nyu | that's good |
| 06:43.53 | braindmg | if he wants "backwards compat" I can add a termios compat layer |
| 06:44.10 | braindmg | but on the source it should be termios IMO |
| 06:44.15 | braindmg | termio is old crap ;) |
| 06:44.32 | braindmg | nyu: the same for other unix compat stuff, the code has a legacy for portability |
| 06:44.32 | nyu | ooh, compat layer. good idea :) |
| 06:44.46 | nyu | but libc5 should die ;) |
| 06:45.02 | braindmg | nyu: but for now I'll try to push it, if he refuses then I'll do the work ;P |
| 06:46.29 | nyu | ok got it |
| 06:46.34 | nyu | http://pastebin.com/550753 |
| 06:46.42 | nyu | could you incorporate this? |
| 06:47.06 | braindmg | nyu: I think I did the define, because I didn't want to introduce a macro loop ... |
| 06:47.14 | braindmg | and changed the macro used afterwards |
| 06:47.47 | braindmg | nyu: is this really necessary? |
| 06:47.50 | nyu | yes |
| 06:47.59 | braindmg | and why does linux does not need it? |
| 06:48.14 | nyu | it seems Linux is more strict |
| 06:48.24 | nyu | erm, less |
| 06:48.47 | nyu | on gnu/kfreebsd it is the only place where such ioctl is accepted by the kernel |
| 06:48.56 | braindmg | hmm ok |
| 06:48.59 | nyu | login can't do setsid(), sysvinit is a different process |
| 06:50.07 | braindmg | have you checked the kernel src or is this by inference? |
| 06:50.09 | nyu | uhm |
| 06:50.17 | braindmg | and why no setsid()? |
| 06:50.39 | nyu | I checked the kernel src to find out that TIOCSCTTY is rejected when you're not the session leader |
| 06:50.45 | nyu | setsid() returns EPERM |
| 06:50.49 | nyu | (even when you're root) |
| 06:50.55 | nyu | wait |
| 06:50.58 | nyu | glibc has something |
| 06:51.09 | braindmg | then, the question is why setsid is rejected |
| 06:51.19 | braindmg | do you fork twice and then setsid? |
| 06:51.23 | braindmg | that's usually needed |
| 06:51.41 | nyu | login_tty |
| 06:51.51 | nyu | in login? |
| 06:52.26 | braindmg | hmm well at least on daemons if you want to becamo a session leader you fork twice |
| 06:52.55 | nyu | I don't think this applies here. the code to fork twice would be already present |
| 06:53.06 | nyu | login_tty() in glibc seem to do the whole thing right |
| 06:53.07 | braindmg | but probably if login forks then init will crap out, |
| 06:53.35 | nyu | it's not a problem with login. login is just a program |
| 06:53.46 | nyu | if you replace /bin/login with a testcase, it'll work too |
| 06:53.55 | nyu | (tested on gnu/linux) |
| 06:54.40 | braindmg | nyu: no I'm talking about login behaving like a daemon =) |
| 06:54.55 | braindmg | well the thing is why cannot you setsid on kbsd? |
| 06:55.02 | braindmg | maybe we have a bug with that code |
| 06:55.04 | nyu | ah right, found it |
| 06:55.07 | nyu | I forgot about this |
| 06:55.33 | nyu | $ rgrep TIOCSCTTY . |
| 06:55.33 | nyu | ./login-utils/agetty.c: * TIOCSCTTY: steal tty from other process group. |
| 06:55.33 | nyu | ./login-utils/agetty.c: ioctl(0, TIOCSCTTY, 1); |
| 06:55.33 | nyu | ./login-utils/simpleinit.c: ioctl (0, TIOCSCTTY, 0); /* I want my control-C */ |
| 06:55.36 | nyu | ./login-utils/login.c: * TIOCSCTTY: steal tty from other process group. |
| 06:55.39 | nyu | ./login-utils/login.c: if (ioctl(0, TIOCSCTTY, 1)) |
| 06:55.55 | nyu | on pure util-linux, login is responsible for TIOCSCTTY, instead of getty |
| 06:56.03 | nyu | but debian login is from shadow, which doesn't do this |
| 06:56.30 | braindmg | hmm then shouldn't we fix shadow instead? ;) |
| 06:56.41 | braindmg | where does freebsd do that? |
| 06:57.37 | nyu | no, shadow is portable |
| 06:57.58 | nyu | AFAIK it plays with other gettys |
| 06:58.09 | nyu | doesn't hurt to add it in shadow as well, though |
| 06:58.20 | nyu | i'm looking at freebsd |
| 07:00.08 | nyu | neither getty nor login do TIOCSCTTY in freebsd |
| 07:00.10 | nyu | ugh |
| 07:00.26 | nyu | maybe it's because of O_NOCTTY |
| 07:00.30 | *** join/#gnu-kbsd DocPheniX (n=Im@ip68-13-63-20.om.om.cox.net) |
| 07:00.43 | *** join/#gnu-kbsd ushio (n=Im@ip68-13-63-20.om.om.cox.net) |
| 07:00.48 | nyu | on gnu/kfreebsd it disables controlling term when opening |
| 07:00.59 | *** part/#gnu-kbsd ushio (n=Im@ip68-13-63-20.om.om.cox.net) |
| 07:01.00 | nyu | on other glibc systems it's a no-op |
| 07:01.12 | nyu | but i tried this already |
| 07:01.23 | nyu | I disabled all O_NOCTTY instances in sysvinit |
| 07:01.38 | nyu | getty doesn't have any of them |
| 07:01.46 | nyu | login shouldn't care since it doesn't open the tty |
| 07:01.50 | nyu | uhm... |
| 07:02.04 | nyu | it's almost surely a kernel restriction of some sort |
| 07:02.20 | nyu | braindmg: how about just using login_tty() from libc? |
| 07:02.48 | nyu | it performs a TIOCSCTTY (for a reason, it seems) |
| 07:03.15 | nyu | it also does some other common things for preparing a login tty |
| 07:04.06 | braindmg | hmm I'd like to know why this is failing though |
| 07:04.26 | braindmg | and not blindly adding something ;) |
| 07:06.56 | nyu | the question is more like why it was working on linux |
| 07:07.22 | nyu | if you don't call TIOCSCTTY, it's normal you don't get a controlling terminal |
| 07:07.41 | nyu | see what other util-linux places do: |
| 07:07.44 | nyu | ./login-utils/simpleinit.c: ioctl (0, TIOCSCTTY, 0); /* I want my control-C */ |
| 07:07.48 | nyu | ./login-utils/login.c: * TIOCSCTTY: steal tty from other process group. |
| 07:07.51 | nyu | ./login-utils/login.c: if (ioctl(0, TIOCSCTTY, 1)) |
| 07:08.16 | nyu | sysvinit does TIOCSCTTY too (like simpleinit.c) |
| 07:09.04 | braindmg | nyu: but you said that this is not done in freebsd ... |
| 07:09.13 | nyu | right |
| 07:09.22 | nyu | i also wonder why it works there |
| 07:09.29 | nyu | perhaps they have it in libc |
| 07:09.32 | braindmg | nyu: do that's the question |
| 07:09.38 | braindmg | s/do/so/ |
| 07:09.47 | nyu | let's check |
| 07:10.05 | braindmg | nyu: I just want to know what's the proper fix, and a justification, and not, because they do that ;) |
| 07:10.51 | nyu | the justification is: because you need a controlling terminal, TIOCSCTTY is the way to get it, and nothing garantees that you have it unless you called TIOCSCTTY. |
| 07:11.13 | nyu | if you have it without calling TIOCSCTTY, it's because of system-specific details that aren't relevant |
| 07:11.23 | *** join/#gnu-kbsd jharrison (n=jharriso@24.244.173.31) |
| 07:11.42 | nyu | as for the place where you should call it, it seems there's no standard either |
| 07:11.54 | nyu | their libc doesn't do it either |
| 07:12.55 | nyu | perhaps their init doesn't do it |
| 07:13.13 | nyu | uhm no, forget that |
| 07:14.02 | nyu | braindmg: have you looked at what login_tty does? |
| 07:14.14 | nyu | it can replace other hacks in getty, not just TIOCSCTTY |
| 07:14.18 | nyu | looks cleaner to me anyway |
| 07:20.51 | braindmg | nyu: yeah I agree that you need a controlling terminal, the question remains why does it work on Linux, and on FreeBSD |
| 07:25.59 | nyu | why do we need to answer that? we can use login_tty() and trust the glibc maintainers have a clue and added TIOCSCTTY for a reason :) |
| 07:26.30 | nyu | the same applies to login-utils/login.c |
| 07:27.00 | braindmg | does uclibc provide login_tty? |
| 07:27.48 | braindmg | yes it does |
| 07:39.25 | nyu | is that fine, then? |
| 07:42.33 | braindmg | yes i can use login_tty, I'll add it to the patch |
| 07:44.09 | nyu | nice thx |
| 07:44.53 | braindmg | I'll start with the small and trivial patches |
| 07:46.43 | nyu | aah, there we go |
| 07:46.44 | nyu | libexec/getty/main.c: if (login_tty(i) < 0) { |
| 07:46.44 | nyu | libexec/getty/main.c: if (login_tty(i) < 0) { |
| 07:46.44 | nyu | libexec/getty/main.c: syslog(LOG_ERR, "login_tty %s: %m", tty); |
| 07:46.56 | nyu | freebsd uses login_tty as well |
| 07:47.45 | braindmg | ok then this is resolved ;) |
| 07:47.53 | braindmg | I'm satisfied now |
| 07:47.55 | braindmg | =P |
| 07:47.56 | nyu | :) |
| 08:22.27 | braindmg | ok first patch sent |
| 08:48.40 | nyu | yikes, /etc/ld.so.conf disappeared |
| 08:48.49 | nyu | tarzeau: I think I found your links2 problem |
| 08:50.29 | nyu | tarzeau: try copying ld.so.conf from a gnu/linux system and see if it works |
| 10:14.04 | CIA-12 | 03rmh * r1174 10/trunk/kfreebsd-6/fetch: Add fetch script, to aid in fetching upstream source. |
| 10:34.39 | nyu | braindmg: could you update http://kfreebsd-gnu.debian.net/debian/obsolete/ls-lR.gz ? |
| 10:34.48 | nyu | i'm missing newer stuff |
| 10:34.54 | nyu | (or setup a cronjob ;) |
| 10:59.23 | *** join/#gnu-kbsd ema (n=ema@debian/developer/ema) |
| 12:13.47 | CIA-12 | 03rmh * r1175 10/trunk/web/patches/xterm.diff: Add xterm.diff. |
| 13:07.24 | *** join/#gnu-kbsd HawkStreet (n=HawkStre@tor/session/x-937b47b5790c27ac) |
| 13:07.34 | HawkStreet | anybody here? |
| 13:12.33 | HawkStreet | how new is this project? |
| 14:26.52 | wobin | HawkStreet: I think it's more than a year |
| 14:27.05 | wobin | last time I installed a Debian GNU/kFreeBSD box was: |
| 14:27.09 | wobin | -rw-r--r-- 1 root root 17 Jun 6 2005 /etc/debian_version |
| 14:27.59 | *** join/#gnu-kbsd Afr0_Style (n=under@217.124.25.135) |
| 15:29.04 | *** join/#gnu-kbsd l3m (n=l3m@84-73-62-186.dclient.hispeed.ch) |
| 15:30.53 | *** join/#gnu-kbsd ptitlouis (n=ptitloui@sysif.net) |
| 16:19.16 | nyu | since 2003-04 |
| 16:48.33 | CIA-12 | 03rmh * r1176 10/trunk/freebsd6-buildutils/debian/patches/000_fix_make_double_free.diff: patches/000_fix_make_double_free.diff: New. Patch from upstream (1.155:1.158). Thanks to Petr for spotting this. |
| 16:48.53 | *** join/#gnu-kbsd otavio (n=otavio@debian/developer/otavio) |
| 16:50.55 | CIA-12 | 03rmh * r1177 10/trunk/freebsd6-buildutils/debian/changelog: Update changelog (oops). |
| 17:11.13 | *** join/#gnu-kbsd dinesh (n=dinesh@84-74-72-172.dclient.hispeed.ch) |
| 17:22.18 | nyu | braindmg: ignore what i said about ls-lR.gz. it seems the missing packages have never been in gnuab |
| 17:24.14 | *** join/#gnu-kbsd aurel32 (n=aurel32@farad.aurel32.net) |
| 17:56.31 | *** part/#gnu-kbsd lt_kije (n=emx@mdsnwikwbas08-pool4-a199.mdsnwikw.tds.net) |
| 18:10.17 | *** join/#gnu-kbsd Afr0_Style (n=under@217.124.24.149) |
| 18:31.10 | *** join/#gnu-kbsd jharrison (n=jharriso@24.244.173.31) |
| 19:33.54 | tarzeau | nyu: really? /me tries... |
| 20:46.55 | tarzeau | sucks when konqueror crashes |
| 21:15.16 | nyu | braindmg: you asked me to nag you when the weekend is over about libbsd and util-linux |
| 21:18.24 | braindmg | nyu: no I said tomorrow (which was this morning or yesterday ;) |
| 21:18.28 | braindmg | to start with it |
| 21:19.11 | nyu | uhm |
| 21:19.17 | nyu | anyway, consider it done =) |
| 21:20.24 | nyu | now it hangs.. sigh |
| 21:21.03 | braindmg | I've still to do the libbsd stuff ... |
| 21:29.17 | *** join/#gnu-kbsd ema (n=ema@debian/developer/ema) |