"C. Michael Pilato" <cmpilato@collab.net> writes:
> Hey, dev-folk. If you get a few spare cycles, I'd really appreciate
> some code review on the various changes I made tonight (between
> revisions r13198 and r13208). The work was pretty straightforward:
>
> - Non-txn_body_* functions that took trail_t's also take pools.
>
> - Make non-txn_body_* callers of those functions pass the pools
> they likely just grew to the functions.
>
> - Make txn_body_* callers of those functions pass trail->pool to
> the functions.
>
> - And then go fix some of the more obvious subpool-needing code
> loops.
Looks like at least one of your subpool changes is wrong:
$ subversion/tests/libsvn_fs_base/fs-test 23
Segmentation fault
$ valgrind -q --num-callers=9 subversion/tests/libsvn_fs_base/.libs/lt-fs-test 23
==21878== Invalid write of size 4
==21878== at 0x4025F6DE: uncache_node_revision (dag.c:117)
==21878== by 0x4026B49D: do_retry (trail.c:232)
==21878== by 0x4026B5CE: svn_fs_base__retry_txn (trail.c:286)
==21878== by 0x4026F63C: svn_fs_base__commit_txn (tree.c:2659)
==21878== by 0x40251AE4: svn_fs_commit_txn (fs-loader.c:409)
==21878== by 0x8049D2C: test_commit_txn (fs-test.c:107)
==21878== by 0x804F0DD: merging_commit (fs-test.c:1866)
==21878== by 0x4022B41C: do_test_num (svn_tests_main.c:154)
==21878== by 0x4022B8C4: main (svn_tests_main.c:283)
==21878== Address 0x417E6FF4 is 12 bytes inside a block of size 20 free'd
==21878== at 0x4002B2C5: free (vg_replace_malloc.c:186)
==21878== by 0x403E478E: pool_clear_debug (apr_pools.c:1375)
==21878== by 0x403E49F9: apr_pool_clear_debug (apr_pools.c:1410)
==21878== by 0x4026EEAA: merge (tree.c:2347)
==21878== by 0x4026F228: txn_body_merge (tree.c:2476)
==21878== by 0x4026B489: do_retry (trail.c:227)
==21878== by 0x4026B5CE: svn_fs_base__retry_txn (trail.c:286)
==21878== by 0x4026F63C: svn_fs_base__commit_txn (tree.c:2659)
==21878== by 0x40251AE4: svn_fs_commit_txn (fs-loader.c:409)
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Mar 2 19:37:05 2005