On 2010-07-12 at 11:14 +0200, Matthias Foerste wrote:
> Today i set up a 'test environment'. It looks like gmtime() still wants
> to lock something:
Okay, I see it, thanks for checking that. Weird that gmtime() even goes
near timezone locks. How unusual. This appears to be:
http://www.google.com/codesearch/p?hl=en#EKZaOgYQHwo/unstable/sources/stash_libc_sources.redhat.com__20050627.tar.gz%7C4C6REHtbCA4/libc/time/gmtime.c&q=gmtime%20package:libc
rather than the glibc-2.0.4 or dietlibc variants that I could find.
So some Linux systems have a gmtime_r() which locks timezone data, so
that whole approach is out.
Looks like changing much of the LOG_PROCESS logic is called for.
Christof Meerwald suggests leaving the time conversion logic entirely to
exiwhat and just store a more raw form. That seems like a good
approach.
-Phil
