On Tue, Aug 26, 2014 at 09:47:54PM +0200, Lennart Poettering wrote: > On Sat, 23.08.14 22:39, Zbigniew Jędrzejewski-Szmek ([email protected]) wrote: > > > We assumed that objects in a unique chain are good enough, > > and only checked object type. But mmap code crashes when some object > > has zero size. This most likely is caused by a corrupted journal > > file, but we should fail gracefully. > > Shouldn't we check the full field name too with memcmp()? I mean, if we > look at the length we really can also compare the string for good, no? True, not necessary to fix the crash, but good for correctness.
> But looks good otherwise, except that I dont like "log_error()" being > invoked from a library. This should be downgraded to > "log_debug()". Libraries should never write anything to stdout/stderr by > default... Yeah, I noticed your commit for sd-journal.c. I'll fix up the patch and push. Zbyszek _______________________________________________ systemd-devel mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/systemd-devel
