Eric Blake <[EMAIL PROTECTED]> wrote: > According to James Youngman on 7/29/2007 2:22 PM: >> /bin/sh: line 8: ../../polzer/tests/test-canonicalize.sh: Permission denied >> >> I've noticed this happening for a couple of weeks now, and could >> believe that this has always been the case for this file, as it was >> only added to CVS on 2007-07-06. I guess the fix is to chmod the RCS >> file in the CVS repository (all the other .sh files in that directory >> are executable). Are you able to chmod the file, Jim? > > Bah, my fault again. I'll be glad when we make the switch to git, since > then anyone can fix these types of problems. Thanks, Jim, for fixing it. > > Hmm, it does raise a question, though - it looks like the current git > mirror did not pick up on that change; is that a bug in git-cvsimport? > (There are other changes not being picked up from the CVS gnulib master, > such as the rename of popcount->count-one-bytes - the git copy never > removed the popcount files). I also wonder if in the reverse direction, > if a change in file permissions in git will be properly reflected via the > git cvsserver?
It just so happens that I regenerated the whole gnulib.git repository from cvs and pushed the result a few minutes ago. Normally, I let an incremental process do the job automatically, but somehow there came to be a branch, and hence lots of ugly merge deltas. Now they're gone. That also removed the popcount files. Of course, just after pushing, I noticed that there is now too much hierarchy in the git tree: a "clone" gives you a top-level gnulib directory, with everything we're used to under that. It'll take a couple hours to regenerate it properly.
