--On Monday, March 7, 2005 7:27 PM -0600 cmpilato@tigris.org wrote:
> Author: cmpilato
> Date: Mon Mar 7 19:27:45 2005
> New Revision: 13300
>
> Modified:
> trunk/subversion/libsvn_fs_base/tree.c
> Log:
> Revert r13242, which itself was a reversion of r13208. In other
> words, re-introduce the changes from r13208.
>
> See, first time 'round, these changes exposed a nasty bug in the pool
> lifetimes of cached node_revision_t's. With pool debugging enabled,
> fs-test would die like so:
>
> $ ./fs-test
> ...
> PASS: lt-fs-test 23: merging commit *** glibc detected *** free():
> invalid next size (fast): 0x0a1f1948 *** Aborted
> $
>
> I've killed that cache altogether now (see r13299), and have verified
> that now the tests run to completion with failure.
>
> * subversion/libsvn_fs_base/tree.c
> (merge, svn_fs_base__commit_txn): Use an iteration subpool.
>
>
> Modified: trunk/subversion/libsvn_fs_base/tree.c
> Url:
> http://svn.collab.net/viewcvs/svn/trunk/subversion/libsvn_fs_base/tree.c?
> view=diff&rev=13300&p1=trunk/subversion/libsvn_fs_base/tree.c&r1=13299&p2
> =trunk/subversion/libsvn_fs_base/tree.c&r2=13300
> =========================================================================
> ===== --- trunk/subversion/libsvn_fs_base/tree.c (original)
> +++ trunk/subversion/libsvn_fs_base/tree.c Mon Mar 7 19:27:45 2005
...snip...
> @@ -2394,6 +2396,7 @@
> * it.
> */
>
> + svn_pool_destroy (subpool);
> return SVN_NO_ERROR;
> }
>
Sun's compiler complains that this section is unreachable. Based on my
examination, I think that's right: it's not possible to ever exit that
while loop any more. So, any reason to keep these two lines? I'll
replace it with a /* NOTREACHABLE */ comment... -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Mar 8 23:45:43 2005