Closed scunz closed 12 years ago
Hiya @scunz! I think the submodule rewrite that landed a few weeks ago resolves this issue. I wrote a test (in pull request #917) and it looks good to me. Can you confirm that this works for you now?
When I first read your new submodule, i thought it would not cover this issue... However, I just retested it - and the problem's gone, so issue solved. :metal: Great work!
On a side note: Retesting reminded me, that I have an assert in my code that shall remind me that the new submodule API has no getter to struct git_submodule
's fetch_recurse
.
D'oh! Put that on my TODO list.
A submodule which is only mentioned in .gitmodules and has no special path, but doesn't exist as a gitlink in the index yet, is never found from any
git_submodule_*
functions. (It falls through the grid in lines 171-190 ofsubmodule.c
)Reproduce with:
(Current implementations of git-submodule.sh do a clone when adding the submodule, thus the gitlink exists. But cloning might have failed. OTOH I know of repositories that just exist to manually populate a .gitmodules file like the example above; thus i feel like this use case should be supported)