Topic: Seg-faults in libapr-1.dll and ntdll.dll when using socache

[cbj4074]

In short, when I attempt to use mod_authn_socache and mod_socache_dbm, I experience segmentation faults in libapr-1.dll and ntdll.dll.

The faults are sporadic and do not occur with every page request. But within a few seconds of "normal browsing", a fault occurs. The fault kills the child process, but not the parent; Apache spawns a new child and retries the request.

I have described this issue at length on the Apache httpd Users mailing list: http://mail-archives.apache.org/mod_mbox/httpd-users/201302.mbox/browser . Unfortunately, nobody seems to have insight to share.

In order to file a bug report with ASF, I need to provide a backtrace, which requires .PDB debugging symbol files for the .exe, .dll, and .so files.

Given that the binaries I'm using were obtained at Apache Lounge, I'm seeking assistance in this regard.

If debugging symbols are available, their location is non-obvious; if they are not available, is there a specific reason for that (beyond free time and man-power)?

Or, if anyone here recognizes this problem or has suggestions, I'm all ears!

I obtained my binaries from Apache Lounge : httpd-2.4.3-win32-ssl_0.9.8-VC9.zip .

Thank you for providing Windows binaries in the first place; I am very grateful for that.

[glsmith]

Have you tried using mod_socache_shmcb instead?

[cbj4074]

No, but that's brilliant. I switched to shmcb and the faults do not occur. Thank you, thank you, thank you!

As much as I'd love to determine why, exactly, dbm socache doesn't appear to work in my environment, I don't have the time to dedicate when there is a readily-available "workaround".

And, as much I'd love to believe that it's a problem with Apache on Windows, or a problem with this particular build, it seems more likely to be related to another issue about which I've posted (the thread has yet to receive a response): http://www.apachelounge.com/viewtopic.php?t=5033 .

There seems to be a problem with stat-like API calls (files are read-locked or otherwise unavailable, sporadically), and I suspect this to be the cause of the dbm socache seg-faults I'm experiencing.

Thanks again for your help with this one.

Now, if you can just figure out why that locking issue occurs, you will be my personal hero.