On 14.12.2012 13:14, Bert Huijben wrote: > >> -----Original Message----- >> From: br...@apache.org [mailto:br...@apache.org] >> Sent: vrijdag 14 december 2012 05:58 >> To: comm...@subversion.apache.org >> Subject: svn commit: r1421668 - in /subversion/trunk/subversion: >> include/svn_error.h libsvn_subr/error.c tests/libsvn_subr/error-test.c >> >> Author: brane >> Date: Fri Dec 14 04:57:42 2012 >> New Revision: 1421668 >> >> URL: http://svn.apache.org/viewvc?rev=1421668&view=rev >> Log: >> Remove SVN_ERR__TRACED from the public headers. No-one should be >> able to create >> traces without calling svn_error_trace, so there's no reason to expose this >> private symbol. Hiding it incidentally allows us to slightly optimize the >> is-tracing-link predicate. > 08:48 < subversion-bot> build #5574 of svn-slik-w2k3-x64-ra is complete: > Failure [failed Test fsfs+svn] Build details > are at > http://ci.apache.org/builders/svn-slik-w2k3-x64-ra/builds/5574 blamelist: > brane, breser > 08:52 <@brane> oh please don't tell me that msvc generates code where > addresses of static, constant arrays aren't > unique! > 08:52 <@breser> :( > > How are you going to transfer a constant pointer through the different RA > layers? > > We support passing error chains over svn:// and http://, and this is why we > have/had to use the strcmp().
Yes indeed, I'm beginning to realize that I shouldn't have tested only ra-local. I'll put the strcmp back. -- Brane -- Branko Čibej Director of Subversion | WANdisco | www.wandisco.com
