http://colabti.org/irclogger/irclogger_log/svn-dev?date=2010-09-07#l40
describes a scenario where a file handle would be leaked due to a pool
not being freed; specifically,
parent(pool) calls child(pool);
child(pool) allocates handle from POOL;
child(pool) triggers an SVN_ERR();
handle remains allocated.
Are there any other cases where this would be a problem?
Should we adopt a strategy of
#define SVN_ERR_CLEAR(expr, scratch_pool)
do {
svn_error_clear(scratch_pool);
SVN_ERR(expr);
} while(0)
at least in some functions taking a scratch_pool argument?
(We could have this only for maintainer-mode builds, and
#define SVN_ERR_CLEAR(expr, pool) SVN_ERR(expr)
for normal builds.)
