Repository navigation
Diff::a_path is wrong for unicode characters #1099
Description
Activity
Thanks a lot, I could reproduce the issue and even managed to sneak in the
-zflag like so:Repo('.').commit().diff('HEAD~', z=True)[0].a_path
This executes
git diff-tree -z c69d9db65234b59d43f17be40667bdbf0a2106f8 HEAD~1 -r --abbrev=40 --full-index -M --raw --no-colorand produces:100644 000000 257cc5642cb1a054f08cc83f2d943e56fd3ebe99 0000000000000000000000000000000000000000 Dšpatně.txtwhereas the version without
-zproduces:100644 000000 257cc5642cb1a054f08cc83f2d943e56fd3ebe99 0000000000000000000000000000000000000000 D"\305\241patn\304\233.txt"exactly as expected.
However, the resulting commit object was empty for some reason, maybe due to a silent parsing failure - even though it doesn't look like it.
This executes
git diff-tree -z c69d9db65234b59d43f17be40667bdbf0a2106f8 HEAD~1 -r --abbrev=40 --full-index -M --raw --no-colorand producesNote that separator characters is
\x00in this case. The following patch works for me:diff --git a/git/diff.py b/git/diff.py index 0fc30b9e..17ef15af 100644 --- a/git/diff.py +++ b/git/diff.py @@ -108,6 +108,7 @@ class Diffable(object): args.append("-p") else: args.append("--raw") + args.append("-z") # in any way, assure we don't see colored output, # fixes https://github.com/gitpython-developers/GitPython/issues/172 @@ -483,7 +484,7 @@ class Diff(object): if not line.startswith(":"): return - meta, _, path = line[1:].partition('\t') + meta, _, path = line[1:].partition('\x00') old_mode, new_mode, a_blob_id, b_blob_id, _change_type = meta.split(None, 4) # Change type can be R100 # R: status letter
Can you please upstream the patch?
- added a commit that references this issue
on Jan 5, 2021 @marxin could you have a look at the linked PR�? That's as far as I could get it, and it looks like there is more to it than the change above. Feel free to submit a new PR with your fixes and I will close mine. Thank you.
@marxin could you have a look at the linked PR�? That's as far as I could get it, and it looks like there is more to it than the change above. Feel free to submit a new PR with your fixes and I will close mine. Thank you.
Thank you for the start. I added one more commit and it seems it works. Please see PR #1102.
- added a commit that references this issue
on Jan 5, 2021 Awesome, thanks so much!
I thank you! See you.
The problem is slightly mentioned in #865.
Let's consider the following example:
Unless one uses
git config --global core.quotepath false, the file name is quoted and I don't see a simple solution how to decode it to unicode? One can use-zoption that can resolve the problem: