Eric Wong wrote. > > So should we just change all "unknown foo" tests to verify that 'git > > svn info' errors out too? > > Yes, I see no reason to differ from plain svn here. This starts getting more complicated at every turn. The included mini-series (probably textually depends on the other 6 patches though) "fixes" this. HOWEVER: Subversion itself broke compatibility here. In 1.4: $ svn info new; echo $? new: (Not a versioned resource) 0 Note the extra linebreak and successful exit. Current git-svn precisely matches this output. In 1.5, it's different: $ svn info new; echo $? svn: 'new' is not under version control 1 While it is of course up to you what you would like to do (and modulo test_must_fail, 2/2 can still be used to fix the tests if you decide to reject 1/2), I suggest changing to 1.5 behaviour. exit(1) is the sane thing to do in this case, and that is already breaking bit-for-bit compatibility with SVN 1.4, so we might as well adopt the new error message. Of course this prevents us from comparing the output literally in the tests, so I settled for a slightly weaker check: failure status and mention of the filename. Unfortunately this does raise the question whether the URL-encoding issue treated in the other series is in fact a similar incompatibility between 1.4 and 1.5, not a (minor but long-standing) bug in git-svn. - Thomas git-svn.perl | 4 +- t/t9119-git-svn-info.sh | 73 ++++++++++++++++------------------------------- 2 files changed, 27 insertions(+), 50 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
