How to get the absolute path of all files in a Git repository with submodules?

Question

I am trying to list all untracked files in all submodules recursively:

git submodule foreach --recursive git ls-files --others --exclude-standard --full-name

But this does not list the absolute path, it gives the relative path to the submodule's folder.

How should I modify my command?

Answer 1

Thanks to torek for the idea. This works:

git submodule foreach --quiet 'a=$(printf "$(git ls-files --others --exclude-standard --full-name|sed "s~^~$(pwd)/~")"; printf x); a=${a%x}; echo "$a"' | sed '/^\s*$/d'

Answer 2

(Note: this is a "tool" answer, not a "solution" answer. That is, this answer is meant to give you the tools you will need to construct the solution you want.)

Use git rev-parse --show-toplevel to get the absolute path to the current repository. Do this while in the superproject, of course. That gives you the first part of any path prefix—but now you'll need the remaining part.

As the git submodule documentation says, in a foreach, you have access to several variables, including $path:

$path is the name of the submodule directory relative to the superproject ...

What's not immediately clear (you will have to test this yourself) is whether $path is cumulative in the case of recursive submodules. That is, if superproject A has submodule B under path top/subB, $path will be top/subB when you're in submodule B. But suppose submodule B acts as a superproject to submodule C, with B's own path being contained/C. Will $path be top/subB/contained/C when inside submodule C, or will it be only contained/C?¹

In any case, once you have the appropriate prefix, just use something like sed to tack it on the front, e.g.,:

git ls-files --others --exclude-standard --full-name | sed "s,^,$prefix,"

where $prefix is set to the appropriate prefix, which might just be $path/. (This assumes none of your file names contain an embedded newline; if they do, your original command is in trouble anyway. As written, it also assumes none of your prefixes contains a comma—if they do, use a character other than ,, or make $prefix quote the commas with backslashes.)

---

¹Logic suggests it will be the complete path; the relative path is not entirely usable without knowing how many submodules have been recursed-into. However, the relative path is easier to produce.