You only know this for the success path. The fs api should hide these details..
There is no way we can document these guarantees for public APIs and really assume that all existing and future users follow all these new requirements for a new Fs layer cache.
We had similar caches in the past, but hid the ugly details by installing cleanup handlers...
We should really look at this well before 1.9, as adding such requirements is a breaking change. And the commit API already existed since well before 1.0 without these pool requirements. Strict dual pools are only common since 1.7, but especially generated bindings (like swig) have strange pool handling much longer.
From: "stefan2_at_apache.org" <stefan2_at_apache.org>
Sent: 4-1-2014 15:05
To: "commits_at_subversion.apache.org" <commits_at_subversion.apache.org>
Subject: svn commit: r1555350 -/subversion/trunk/subversion/libsvn_client/mtcc.c
Date: Sat Jan 4 14:04:36 2014
New Revision: 1555350
Fix segfault in svnmucc.
(svn_client_mtcc_commit): We explicitly destroy the MTCC pool and
have no knowledge about its relation to
SCRATCH_POOL. Thus, we can't use the
for anything non-trivial.
--- subversion/trunk/subversion/libsvn_client/mtcc.c (original)
+++ subversion/trunk/subversion/libsvn_client/mtcc.c Sat Jan 4 14:04:36 2014
@@ -1337,12 +1337,15 @@ svn_client_mtcc_commit(apr_hash_t *revpr
"is not a directory"),
+ /* Beware that the editor object must not live longer than the MTCC.
+ Otherwise, txn objects etc. in EDITOR may live longer than their
+ respective FS objects. So, we can't use SCRATCH_POOL here. */
SVN_ERR(svn_ra_get_commit_editor3(mtcc->ra_session, &editor, &edit_baton,
NULL /* lock_tokens */,
FALSE /* keep_locks */,
err = editor->open_root(edit_baton, mtcc->base_revision, scratch_pool, &root_baton);
Received on 2014-01-04 21:49:32 CET