Re: Side effect of webwml alioth transition
[nekral, added you to explicit recipients because it might be a po4a
* Osamu Aoki <email@example.com> [2009-07-05 14:40:19 CEST]:
> It looks like I mixed up few things and issue is not related to webwml
No worries, actually we are facing three things that more or less broke
at the same time, so it's easy to think of a connection. :)
For a start, the build-logs are accessible again thanks to weasel who
adjusted the new default vhost that got installed to also allow access
to them again. :)
> On Sun, Jul 05, 2009 at 12:09:46PM +0200, Gerfried Fuchs wrote:
> > * Osamu Aoki <firstname.lastname@example.org> [2009-07-05 04:50:55 CEST]:
> > > Following may be related problem to recent webwml transition.
> > Not completely sure about that because the doc part is unrelated to the
> > webwml cvs from what I understand, it rather might also be related to
> > some recent commits:
> I see.
> > #v+
> > :~$ ps www 23329
> > PID TTY STAT TIME COMMAND
> > 23329 ? R 1333:59 /usr/bin/perl /usr/bin/po4a-translate --format docbook --keep 0 -m beyond-pkging.dbk -p po4a/ja/beyond-pkging.po -l ja/beyond-pkging.dbk
> > #v-
> /usr/bin/po4a-translate is slow CPU intensive process.
That might be true (and hopefully a place for improvements ;)), but it
shouldn't last _that_ long.
> This is developer reference Japanese translation by Hideki Yamane
> <email@example.com>, I think.
Thanks for digging that up. And I think it really might be related to
/proc/2441/fd$ ls -l| cut -d\ -f10-
0 -> pipe:
1 -> /org/www.debian.org/ddp-svn/make.log
2 -> /org/www.debian.org/ddp-svn/make.log
3 -> /org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po
/proc/2441/fd$ msgfmt --statistics -o /dev/null /org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po
/org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po:492: end-of-line within string
/org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po:492:3: syntax error
/org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po:495: end-of-line within string
/org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po:507: end-of-line within string
/org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po:507:3: syntax error
/org/www.debian.org/ddp-svn/trunk/developers-reference/po4a/ja/beyond-pkging.po:508: end-of-line within string
msgfmt: found 6 fatal errors
The process 2441 is now already running since more than an hour. An
strace on the process doesn't really yield much:
0.000000 brk(0xe10e000) = 0xe10e000
0.000104 brk(0xe10d000) = 0xe10d000
9.409890 brk(0xe12e000) = 0xe12e000
0.000093 brk(0xe12d000) = 0xe12d000
9.437423 brk(0xe14e000) = 0xe14e000
0.000085 brk(0xe14d000) = 0xe14d000
(that's time between the different syscalls in the first column)
I just killed it again.
> > There is something extremely fishy with this - yes, the TIME _is_ 1333
> > minutes aka 22 hours.
> I suspect that this is taking long since it was not translated much or
> po file is out of sync with pot files.
But _that_ long? I rather suspect there to be a bug somewhere in po4a
related to not being able to work well with broken po files.
> (I am not quite sure how DDP build script is run on master.)
You can see it in webwml cron repository: