WIP: respect index during checkout #3839
Conversation
Test that we can successfully force checkout a target when the workdir was updated (to match the target) but the index differs.
When performing a forced checkout, treat files as modified when the workdir is identical to the target, but the index differs. This ensures that we update the index to match, instead of assuming that since the working directory does not differ, there's nothing to do. Otherwise we would keep the index as-is and the staged change would propagate past the force checkout.
Test that we can successfully force checkout a target when the file contents are identical, but the mode has changed.
When performing a forced checkout, treat files as modified when the workdir or the index is identical except for the mode. This ensures that force checkout will update the mode to the target. (Apply this check for regular files only, if one of the items was a file and the other was another type of item then this would be a typechange and handled independently.)
When the target of a checkout has a deletion of an item, a modified index entry should cause a conflict even when the workdir has been reverted to the baseline.
When the target of a checkout has a modification of an item, a modified index entry should cause a conflict even when the workdir has been reverted to the baseline.
Introduce a new branch that changes the file `README` into a folder.
When the target of a checkout has a typechange of an item, a modified index entry should cause a conflict even when the workdir has been reverted to the baseline.
|
|
||
| cl_git_pass(git_status_list_new(&status, g_repo, NULL)); | ||
| cl_assert_equal_i(0, git_status_list_entrycount(status)); | ||
| git_status_list_free(status); |
There was a problem hiding this comment.
It's the third test now that I see this pattern, checking that our repository is actually in a clean state (that is, asserting the branch and then asserting that no modifications are present). Maybe put this into a function/macro? I bet it would be useful to other tests.
|
Are there any plans to continue this work? It seems like libgit2 does not properly manage file mode only changes. If I try to force checkout a file with only file mode changes nothing happens, as I believe libgit2 is only checking content, which has not changed. It looks like you fixed that in this WIP. |
|
Thanks @cjhoward92 for the ping. I'll bring forward those fixes in a separate PR. |
|
You rock! |
|
@ethomson: can this be closed? I feel like you've already committed most changes in here to master |
Our current checkout implementation looks to see if the working directory is dirty to determine whether it's safe to checkout a given file. It does not look at the index, however.
This is fine if we're checking out the repository's index itself, but if we're checking out a
git_treethen we could get into a place where the repository's index has modifications, but the workdir itself has been reverted. We would obliterate those staged changes.This adds checking for changes in the index before checking out (raising conflicts as appropriate otherwise) when a working directory file exists. This does not yet cover other cases (for example, when an entry exists in the index but not in the working directory.)
I think that we can simply change the
with_wdcase to awith_wd_or_indexcase and use the existingno_wdcase to beno_wd_or_index.