PDA

View Full Version : 6.2.x seg faults -- help requested to narrow down the problem



kevine
2005-11-22, 04:18
I recently ripped a few more of my CDs (all flac format, individual
songs) and now slimserver seg faults (debian linux, perl 5.8.4) part
way through rescan. I went from roughly 190 CDs to 230 CDs.

I was running the svn head of the 6_2 branch, revision 5282 to be
precise. So I switched to 6.2.1 download and same problem, seg fault
part way through scan.

I have deleted .slimserversql.db (for both version) and when tested,
either version still seg faults.

I tried build-perl-modules.pl on svn branch head which ran to
completion, still seg faults.

I tried running each version on the command line, with --d_scan
--d_server and nothing I see is indicative of an error. The log simply
scrolls past, and it still seg faults.

I tried -diag and got

2005-11-22 21:31:59.2829 Use of uninitialized value in substitution (s///) at /usr/local/slimserver/Slim/DataStores/DBI/DBIStore.pm line 1474.
2005-11-22 21:31:59.4157 Use of uninitialized value in substitution (s///) at /usr/local/slimserver/Slim/DataStores/DBI/DBIStore.pm line 1474.
2005-11-22 21:31:59.4201 Use of uninitialized value in substitution (s///) at /usr/local/slimserver/Slim/DataStores/DBI/DBIStore.pm line 1474.

repeated for each file, which seems to be an unrelated issue with
replay gain tags.

I tried gdb on the core, and did not get anything enlightening (no
valid backtrace).

I tried memtest on the box (suspecting hardware) and it runs fine,
like everything else except slimserver.

So any hints on how to narrow this down to something close enough to
file a meaningful bug report?

kevine
2005-11-22, 04:40
Additionally, after enough attempts, the scan "completes" indicating 224 albums of 3176 songs, and promptly core dumps. Oddly, a little playing with ls */* | wc -l , and find -name \*.flac | wc -l ,etc, gives 225 albums of 3176 songs.

The album number seems out by one.

cheers

- Kevin

kevine
2005-11-22, 17:45
Investigating further, I attempted to build a later version of perl with debugging symbols to see if I could narrow down the problem. I did, but not how I would have like to.

I found gcc would also seg fault at arbitrary points in the perl build process (say 2-5 times), but "make" succesfully continues when restarted.

I'm chalking this one up as a hardware fault that only seems to pop up under longer periods of heavy I/O load (or some obscure config/kernel issue on old AMD K6 hardware).

cheers
- Kevin

Jacob Potter
2005-11-22, 18:20
On 11/22/05, kevine <kevine.1yxcpb (AT) no-mx (DOT) forums.slimdevices.com> wrote:
> I'm chalking this one up as a hardware fault that only seems to pop up
> under longer periods of heavy I/O load (or some obscure config/kernel
> issue on old AMD K6 hardware).

I think compilation seg faults are usually a symptom of bad RAM...
have you tried memtest86+?

- Jacob

kevine
2005-11-23, 21:16
On 11/22/05, kevine <kevine.1yxcpb (AT) no-mx (DOT) forums.slimdevices.com> wrote:
> I'm chalking this one up as a hardware fault that only seems to pop up
> under longer periods of heavy I/O load (or some obscure config/kernel
> issue on old AMD K6 hardware).

I think compilation seg faults are usually a symptom of bad RAM...
have you tried memtest86+?

- Jacob

Yes, it ran trouble free for 5 hours.

These hardware-like issues are now getting off-topic, so thanks for the tip.

ciao

- Kevin