First off - I appreciate FSVS so much; it's been a lifesaver helping me 
figure out how I messed up my postfix install, or why apache is suddenly 
unhappy. I routinely install it on every *nix box I stand up.

I wanted to drop you a line to confirm the strange results reported in 
http://www.mail-archive.com/[email protected]/msg00829.html  ("fsvs 
on Emdebian"). I'm running Debian Squeeze on a Marvell SheevaPlug 
('armel' architecture). The version in squeeze is 1.1.17-1, and 
segfaults in direnum just as Lukas indicated in that thread (faults on 
fsvs stat or fsvs commit).

I built the package locally from the debian source, with --enable-debug 
added to the ./configure line, and it still segfaulted. I then added 
your patch from 
http://www.mail-archive.com/[email protected]/msg00826.html 
(manually applied, since the debian line numbers are slightly off your 
baseline for the patch) - and somehow it suddenly worked?

All I can think is that this is uncovering a bug in the toolchain - I 
simply can't see why adding some debug printfs would change the 
semantics of the program (especially since it works even without -d!). 
Either I'm missing something in that patch, or there's something truly 
bizarre going on (a broken optimization?) and adding the conditional 
print is enough to bring the compiler back to sanity?

Anyway, it looks like my home-built package is working fine now, but I 
thought I'd pass along a confirmation of this pretty bizarre symptom. If 
you're interested I'd be happy to try experiments locally, though the 
Sheevaplug is pretty resource-bound for any serious debugging effort (I 
was happy just to get the compilation working without having to set up a 
cross-compile chain).

Thanks,
Seth

------------------------------------------------------
http://fsvs.tigris.org/ds/viewMessage.do?dsForumId=3928&dsMessageId=2685984

To unsubscribe from this discussion, e-mail: 
[[email protected]].

Reply via email to