From: Ivan Chernyavsky Date: Fri, 21 Aug 2015 21:29:59 GMT Subject: Which branch(es) contain certain commits? (was Re: (unknown)) Message-ID: <5570041440192599@web21j.yandex.ru> In-Reply-To: 17.08.2015, 20:49, "Junio C Hamano" : >  Duy Nguyen writes: > >>   On Wed, Aug 5, 2015 at 7:47 PM, Ivan Chernyavsky wrote: > >  That is a dangeous thought. I'd understand if it were internally >  two step process, i.e. (1) the first pass finds commits that hits >  the --grep criteria and then (2) the second pass does "--contains" >  for all the hits found in the first pass using existing code, but >  still, this operation is bound to dig all the way through the root >  of the history when asked to find something that does not exist. My intention was to use existing git-branch filters such as -a or -r and pattern to limit the scope, then apply --grep machinery. But now I had a look on the source and I can see that builtin/branch.c builds the list of references and prints them in a single place (print_ref_list()) so I will have to split that function into two in order to reuse existing functionality. Another problem is that builtin/branch.c currently does not use setup_revisions(), so I'll have to hook it there as well. Then, I assume, I'll need to use the initial ref_list (filled the same as for the current "list" case) to configure the rev_info structure after setup_revisions(), and start revision traversal. I'm not sure I've got it all right from the source in those few days, so I apologize in advance if it's stupid in some part or as a whole. That said, do you think the goal is worth such changes? Seems like git-branch currently has it's own way of doing things and I'm trying to teach it to use git-rev-list's. Maybe it is more reasonable to add an option to "git log --all --source --grep=..." to print *all* branch tips which are reachable from the found commits, not only the first one encountered? E.g. --decorate-all or --source-all? >  You should at least feed all --contains to a single invocation of >  "git branch". They are designed to be OR'ed together. Yep thanks I overlooked that. In my repository, this takes 40 seconds instead of 2 minutes 30 seconds for my construct. --   Ivan