Daily blurbs Aug. 2005


daily blurbs

30 Aug 2005 (Tue)

09:19:02 # Life I will be working for the Debian JP office as Sub-leader for this term. Hopefully I can get the rusty situation a bit improved to get the momentum to have a DebConf gathering in Japan sometime in the future. On an unrelated note, I've requested binfmtc and cowdancer sponsored uploads; which should be interesting. cowdancer probably needs some improvements in efficiency, but it's somewhat functional as it is.

28 Aug 2005 (Sun)

13:07:47 # Life Maintanancing pbuilder. Incidentally, it's the fourth anniversary of pbuilder or something. The original decision with pbuilder was to preserve as much environmental variables as possible; and see how packages work. That might in fact be a wrong approach, but I don't think it's really easy to fix that with shell. You can prove me otherwise. The main problem being that you can use env -i but restoring the necessary (and finding out what is really necessary) needs some work. I'm inclined to initialize only those environmental variables that are really known to break pbuilder badly. Ones such as SHELL, and PATH.

On the other side of the story, I am wondering if anyone would be interested in co-maintenance of pbuilder? It's definitely in what I would call a maintenance phase, but bugs are quite regularly filed.

20 Aug 2005 (Sat)

11:38:14 # Life linda/lintian should probably error out if Build-Depends field is syntactically incorrect. There isn't a place where syntax error in Build-Depends field is really noticed. Most parsers of Build-Depends are by nature more permissive about minor problems. I would like it to detect problems like the following:

Build-Depends: package > 4.0

Build-Depends: something [hurd, i386]

Build-Depends: other, \

The more correct versions are:

Build-Depends: package (>> 4.0)

Build-Depends: something [hurd i386]

Build-Depends: other, 

Marc blogged that it's actually already implemented in lintian. If that's the case, I really don't feel the necessity to handle 310941 on pbuilder side.

12:01:24 # Life Tried soundtracker jack support again. It's hanging somewhere there; weird. I think jack initialization is not doing the correct thing first time. Selecting jack after the thing started up works fine.

(gdb) select-frame
(gdb) thread apply all bt

Thread 2 (Thread -1210672208 (LWP 28787)):
#0  0xb7e3d5f8 in clone () from /lib/tls/i686/cmov/libc.so.6
#1  0x007d0f00 in ?? ()
#2  0x00000000 in ?? ()
#3  0xb7f29dc0 in __make_stacks_executable ()
   from /lib/tls/i686/cmov/libpthread.so.0
#4  0xb7f29999 in pthread_create@@GLIBC_2.1 ()
   from /lib/tls/i686/cmov/libpthread.so.0
#5  0xbff71968 in ?? ()
#6  0xb7dfedbc in setresuid () from /lib/tls/i686/cmov/libc.so.6

Thread 1 (Thread -1210668832 (LWP 28784)):
#0  0xffffe410 in __kernel_vsyscall ()
#1  0xb7f28a90 in __nptl_setxid () from /lib/tls/i686/cmov/libpthread.so.0
#2  0xb7dfedbc in setresuid () from /lib/tls/i686/cmov/libc.so.6
#3  0x0806e393 in main (argc=1, argv=0xbff71a24) at main.c:123
#0  0xb7e3d5f8 in clone () from /lib/tls/i686/cmov/libc.so.6

15 Aug 2005 (Mon)

08:56:01 # Life I've noticed that quite a lot of mail servers nowadays deny mail from my mail server. This is due to the fact that my mail server machine does not have a matching reverse-DNS entry. I don't think it's practical to request IPs to resolve to what it's trying to tell you in HELO, for instance, for virtual servers; so I would consider it's broken, but they won't see my mail. Sigh...

13 Aug 2005 (Sat)

11:38:58 # Life It's nice to see 'blocker' feature for debbugs. The most interesting talk (for me) at debconf was the one about debbugs by ajt. That however doesn't mean I've actually managed to get anything done for debbugs; which is a bit of a shame really. My wishlist is to have RSS feed for each bug, and dumping XML. I seem to remember starting doing that long time ago, but I don't know where the source code went.

21:41:32 # Life I've managed to get cowdancer released. It's a project that I've been motivated to look around; but not found a solution; and decided to actually implement it after I've been to Debconf5. I didn't really have much time for hacking after debconf5, and time seems to fly, but anyway. Packages are available here. cowdancer implements a copy-on-write filesystem semantics from a userland-only way. There also is an introductory page. I'm intending to use this along with pbuilder to speed things up significantly.

7 Aug 2005 (Sun)

13:15:09 # Life cdebootstrap support for pbuilder. I've worked on pbuilder this morning, and tried to see how cdebootstrap was doing. debootstrap is not able to bootstrap etch/sid right now. cdebootstrap won't bootstrap sid, but apparently it is able to bootstrap etch. It seems to use up so much diskspace for some reason, and does not seem to be able to run within 250MB or so I have allocated for it to work, when it's trying to test the upgrade procedure. Hmm...

6 Aug 2005 (Sat)

15:58:45 # Life dpatch maintenance. I worked on a few dpatch bug reports, and updated them within the arch repo. Only trivial changes and updates. Just a maintenance.

5 Aug 2005 (Fri)

23:49:07 # Life 6 Aug 2005 is the 60th year since the atom-bombing of Hiroshima. Around 250000 civilians died due to the brutal inhumane attack. I tribute a minute of silence for those who were destroyed. It did prove what E=mc2 meant; but at the same time it tells me that war is not about winning or losing, but death and mass destruction. And it's quite difficult to rationally discuss or explain such a villainous act.

I received an email against this entry, so I will comment further. I am mourning for my family; and the still on-going state of the result of Atomic weapon that has been cast upon the whole city. I consider it a mission that I convey the information to the next generation that use of the atomic weapon should never happen.

ermm... I do hope no Debian Developer decides to quit the Debian project because of this entry.

4 Aug 2005 (Thu)

08:18:11 # Life There was a wishlist bug filed for pbuilder to support unionfs. Most of the parts are already in pbuilder to support such configuration since pbuilder-user-mode-linux supported COW filesystem operations. Unionfs is not quite the best way to play with it since it seems to require a filesystem that is not in the mainline kernel, not even in the current git tree.

It's interesting that unionfs seems to claim it preserves UNIX semantics, even the i-node expectations. I will be interested to check what could be the case.

1 Aug 2005 (Mon)

01:06:30 # Life This weekend it will be the Monthly Tokyo-area Debian Meeting. First meeting after the Debconf. I've actually managed to not finish the keysigning process, and it doesn't look like I get around to doing it anytime soon. I've been playing with ptrace and LD_PRELOAD hacks for the past few days. ptrace is not quite an easy beast to use; it's a useful tool to find exploitable flaw, but not a easy-to-use tool for making an application on top of it. strace, gdb and user-mode-linux are real monsters.

Junichi Uekawa

$Id: 200508.html.en,v 1.16 2005/08/30 00:20:48 dancer Exp $