On Tue, Feb 7, 2012 at 12:49 AM, Philip Martin <philip.mar...@wandisco.com> wrote: > Johan Corveleyn <jcor...@gmail.com> writes: > >> I actually don't see anything either in the Apache error log (except >> the child process starting up etc). Or should I look somewhere else? > > The error log is the right place. > >> Just to add some more info: I can't reproduce with mod_dav_svn from >> trunk@1237720 or trunk@1239596 (the two revisions which were >> backported in r1239697). >> >> Makes me wonder whether something maybe went wrong during the backport >> itself (merge conflict resolution)? Or the backported revisions >> depended on something else which was not backported? > > The backport looks correct to me. One effect of the change is that the > server no longer sends the bogus '###error###' MD5 checksum property. I > suppose you could remove the svn_fs_check_path call at line 699 in > subversion/mod_dav_svn/liveprops.c and set kind to svn_node_file to > force the bogus property to be sent. > >> (the above findings may need to be confirmed by someone else though. >> It's getting late here, and I'm trying all sorts of different >> configurations, so I hope I haven't made any mistakes). > > If you want to try debugging the failing command then you can setup the > test repository: > > svnadmin create repo > svnadmin load repo < > subversion/tests/cmdline/svnrdump_tests_data/mergeinfo_included_full.dump > > Setup Apache to serve that repository and then the command to debug is: > > svnrdump -q -r1:10 http://localhost.../repo
Sorry, need to get some sleep first :-). Maybe tomorrow-night, but perhaps you guys will have found it by then. Anyway, I tried to help a bit with the "bisection", but I'm not sure if I'll have the time to dig deeper. I hope someone else can reproduce and debug it ... -- Johan
