subversion-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Branko ─îibej <br...@wandisco.com>
Subject Re: svn commit: r1502305 - in /subversion/trunk/subversion: bindings/cxxhl/src/exception.cpp bindings/swig/python/libsvn_swig_py/swigutil_py.c libsvn_subr/error.c libsvn_subr/nls.c mod_dav_svn/mod_dav_svn.c
Date Thu, 11 Jul 2013 19:52:23 GMT
On 11.07.2013 20:08, danielsh@apache.org wrote:
> Author: danielsh
> Date: Thu Jul 11 18:08:23 2013
> New Revision: 1502305
>
> URL: http://svn.apache.org/r1502305
> Log:
> Use svn_pool_create() instead of apr_pool_create().
>
> Presently, this means that if an apr_pool_create() fails, abort_fn() will be
> called.  None of those plafces check for NULL results from the allocator,
> so the net effect is changing a NULL dereference to calling our pool.c
> function abort_on_pool_failure() (which is marginally better).
>
> * subversion/bindings/cxxhl/src/exception.cpp
>   (Error::compile_messages):

This change is wrong, please revert it. I agree the code needs to check
for the null return, however, replacing the current mode with an abort
is not "marginally better", it's completely wrong. I seem to recall
we've been through this before, we avoid using the svn_pool APIs there
on purpose.

Not to mention that your change breaks the cxxhl build.

-- Brane



-- 
Branko ─îibej | Director of Subversion
WANdisco // Non-Stop Data
e. brane@wandisco.com

Mime
View raw message