Branko Čibej <br...@wandisco.com> writes: > On 15.12.2014 18:13, Julian Foad wrote: >> I see a test failure on current trunk over HTTP: >> >> $ basic_tests.py ... 38 >> ... >> CMD: svnmucc ... put .../C</C< ... >> exited with 1 >> .../subversion/svnmucc/svnmucc.c:225, >> .../subversion/libsvn_ra_serf/xml.c:915: >> (apr_err=SVN_ERR_RA_DAV_MALFORMED_DATA) >> svnmucc: E175009: The XML response contains invalid XML >> .../subversion/libsvn_ra_serf/xml.c:915: (apr_err=SVN_ERR_XML_MALFORMED) >> svnmucc: E130003: Malformed XML: not well-formed (invalid token) >> FAIL: basic_tests.py 38: use folders with names like 'c:hi' >> >> I understand from Philip and Ben on IRC that this is a known bug with some >> versions of httpd. On our end we >> >> <philipm> Brane added a check the knows about the broken versions in some >> distributions (OS X and RedHat possibly) but not all the broken version. >> [...] >> <breser> Yeah and he said something to me about looking at the list and I >> forgot to do it. >> <philipm> Ben added a configure check that prevents building with some >> broken versions, Brane added an XFAIL for other broken versions. >> Ben's regex in apache.m4 is 2.4.[56], Brane's blacklist is 2.4.9. > > My blacklist is based on what was written in the issue noted in the test > case decorator. Could be the issue was not complete.
The issue identifies 2.4.9 as broken, however it also points to the cause as being the erroneous patch in https://issues.apache.org/bugzilla/show_bug.cgi?id=54611 which was applied to 2.2.25 and 2.4.5. The bug, or a variant of the bug, is present in all subsequent versions until fixed in 2.2.28 and 2.4.10. Julian's point remains: how should we handle the broken versions? Should we blacklist at configure time and refuse to build, or should we allow the build and FAIL the regression tests, or should we allow the build and XFAIL? At present we do all three. -- Philip Martin | Subversion Committer WANdisco // *Non-Stop Data*
