From: Ævar Arnfjörð Bjarmason Date: Mon, 28 Nov 2022 12:43:08 GMT Subject: Re: making use of $sm_path @ git submodule foreach —recursive Message-ID: <221128.86r0xnkzyw.gmgdl@evledraar.gmail.com> In-Reply-To: <70ED2ECF-DAF1-4CCB-9E05-2599211F3569@dynax.at> On Sun, Nov 27 2022, signal@dynax.at wrote: > Hi, > > when using > git submodule foreach —recursive [..] > $sm_path contains only the relative part of the latest > recursion. Since the command has no generic way of knowing from which > recursion level it is executed any $sm_path of deeper recursion level > is useless unless there is a way for the command to find out from > which recursion it is called. > > I suggest $sm_path should be extended to be relative to the repo from where the submodule recursion started or another variable should contain the recursive part. > > Viele Grüße/Cheers, > Hagen. I think it might be sensible to have a $super_prefix_sm_path or something which does what you suggest here. But the current "$sm_path" is far from useless, it's just not useful for what you're trying to do. When we run a command in sub1/sub2 or whatever the "$sm_path" will be "sub1", then as we cd to "sub1" it'll be "sub2". You want "sub1/sub2" there, but a "sub2" is still useful, because we've chdir()'d to the "sub1" at that point. So you can e.g. run 'git -C "$sm_path" log' in your 'foreach', or another command that expects to get the *relative* submodule path. If we simply changed how "$sm_path" works that would break, and if we "fixed" that by not chdir()-ing from the super-project we'd break even more things, as e.g.: git submodule foreach 'git pull' Or whatever wouldn't behave as you'd expect.