On 10/30/06, Jani Averbach <jaa@jaa.iki.fi> wrote:
> I compiled svn trunk / apr 0.9.x with serf and tried the svntest
> system with it. It didn't go so well. Either the system was
> segfaulting (1), or I am seeing following errors:
>
> svn: Error running context: Error string not specified yet
>
> I have compiled apr with "--enable-pool-debug=yes".
Okay, I've been focused on getting Win32 to work. I'll try the
regression tests on my Mac as time permits.
I don't expect that any changes I made in the last few days would
break stuff though; a few might fix things seen on non-Win32 platforms
- specifically r22516.
> Secondly, I ran serf with valgrind and I got following results:
>
> ==12141== Memcheck, a memory error detector.
> ==12141== Copyright (C) 2002-2006, and GNU GPL'd, by Julian Seward et al.
> ==12141== Using LibVEX rev 1658, a library for dynamic binary translation.
> ==12141== Copyright (C) 2004-2006, and GNU GPL'd, by OpenWorks LLP.
> ==12141== Using valgrind-3.2.1, a dynamic binary instrumentation framework.
> ==12141== Copyright (C) 2000-2006, and GNU GPL'd, by Julian Seward et al.
> ==12141== For more details, rerun with: -v
> ==12141==
> ==12141== Warning: invalid file descriptor -1 in syscall close()
> ==12141== Conditional jump or move depends on uninitialised value(s)
> ==12141== at 0x5638E3F: handler_default (util.c:763)
> ==12141== by 0x5746648: serf_context_run (context.c:744)
> ==12141== by 0x563622B: finish_report (update.c:2105)
> ==12141== by 0x4C5B588: svn_wc_crawl_revisions2 (adm_crawler.c:599)
> ==12141== by 0x4B47601: svn_client__update_internal (update.c:151)
> ==12141== by 0x4B2ED6A: svn_client__checkout_internal (checkout.c:157)
> ==12141== by 0x4B2EFF4: svn_client_checkout3 (checkout.c:222)
> ==12141== by 0x405E2B: svn_cl__checkout (checkout-cmd.c:159)
> ==12141== by 0x40AA24: main (main.c:1544)
> ==12141== Warning: invalid file descriptor -1 in syscall close()
> ==12141== Warning: invalid file descriptor -1 in syscall close()
> ==12141== Warning: invalid file descriptor -1 in syscall close()
> ==12141== Warning: invalid file descriptor -1 in syscall close()
> ==12141==
> ==12141== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 1)
> ==12141== malloc/free: in use at exit: 24,688 bytes in 5 blocks.
> ==12141== malloc/free: 365,371 allocs, 365,366 frees, 60,583,066 bytes allocated.
> ==12141== For counts of detected errors, rerun with: -v
> ==12141== searching for pointers to 5 not-freed blocks.
> ==12141== checked 649,224 bytes.
> ==12141==
> ==12141==
> ==12141== 24,576 bytes in 3 blocks are definitely lost in loss record 3 of 3
> ==12141== at 0x4A1FB4B: malloc (in /usr/lib64/valgrind/amd64-linux/vgpreload_memcheck.so)
> ==12141== by 0x6479144: apr_allocator_alloc (apr_pools.c:293)
> ==12141== by 0x57485C4: serf_bucket_mem_alloc (allocator.c:183)
> ==12141== by 0x5745C5E: serf_connection_request_create (context.c:1056)
> ==12141== by 0x56321F7: svn_ra_serf__deliver_props (property.c:567)
> ==12141== by 0x5632406: svn_ra_serf__retrieve_props (property.c:611)
> ==12141== by 0x5638430: svn_ra_serf__discover_root (util.c:913)
> ==12141== by 0x5634753: svn_ra_serf__get_latest_revnum (serf.c:211)
> ==12141== by 0x4B461AF: svn_client__get_revision_number (revisions.c:65)
> ==12141== by 0x4B4340F: svn_client__repos_locations (ra.c:760)
> ==12141== by 0x4B43A2E: svn_client__ra_session_from_path (ra.c:922)
> ==12141== by 0x4B2EABF: svn_client__checkout_internal (checkout.c:84)
> ==12141==
> ==12141== LEAK SUMMARY:
> ==12141== definitely lost: 24,576 bytes in 3 blocks.
> ==12141== possibly lost: 0 bytes in 0 blocks.
> ==12141== still reachable: 112 bytes in 2 blocks.
> ==12141== suppressed: 0 bytes in 0 blocks.
> ==12141== Reachable blocks (those to which a pointer was found) are not shown.
> ==12141== To see them, rerun with: --show-reachable=yes
I'll try to take a look at this, but the allocator stuff is likely to
be a red herring due to the way the memory allocator works.
> I will try to compile the serf with apr-1.x and see if that makes any
> difference.
>
> In anycase, when I get the serf to working with our regression tests,
> I can/will start running nightly tests with it. It would be nice if
> someone could confirm if or not he or she can run serf with regtests.
Very cool.
> 1) I tested the system just before your windows fixes and now it seems
> that I can't anymore reproduce the segfault. However I will include
> the backtrace which I generated at the time.
So that means with trunk, you're not seeing the segfault?
Thanks! -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Oct 30 21:23:17 2006