dvc and git does not behave the same with "!" and "**"

[Danial-Alh]

Consider the following project structure

• data
  • data1
    • file1
    • file1.dvc
  • data2
    • file2
    • file2.dvc
• .gitignore

.gitignore is as follows:

data/**
!data/*/
!*.dvc

git status gives:

while dvc push gives:

I expect to git and dvc behave the same with gitignore.

• dvc: 2.8.3
• python: 3.7

[pmrowla]

Your .dvc files are actually still ignored by those .gitignore patterns, the difference here is that you have explicitly staged them in git (presumably with git add -f). You can verify this yourself by doing

git check-ignore --no-index data/data1/file.dvc

DVC only checks the actual .gitignore patterns, and does not currently check if ignored files have been explicitly (force) staged in git (see #6291)

To make DVC work properly here, your .gitignore should contain:

/data/**
!/data/**/
!/data/**/*.dvc

[Danial-Alh]

Thanks for your quick response.

That .gitignore worked.
Thanks a lot.

However, I didn't add the files to git with --force option.
I checked the git check-ignore --no-index data/data1/file1.dvc. It gives no output.

[karajan1001]

$ cat .gitignore
data/**
!data/*/
!*.dvc
$ git check-ignore -v data/data1/a.dvc
.gitignore:3:!*.dvc	data/data1/a.dvc

And for dulwich

$ dulwich check-ignore data/data1
data/data1
$ dulwich check-ignore data/data1/file1.dvc
$

I think they both work correctly for this .gitignore file

[Danial-Alh]

If I have understood correctly, .dvc files are ignored in neither of my nor dulwich's format; but I don't get why dvc does not see .dvc files in my format.

[Danial-Alh]

for dulwich format I get this:

the folder is not ignored.

[Danial-Alh]

There is a problem with dulwich format which the file1 is not ignored by git.

I want all files in data folder to be ignored by git, except .dvc files.

[karajan1001]

There is a problem with dulwich format which the file1 is not ignored by git.

I want all files in data folder to be ignored by git, except .dvc files.

It is ignored in my computer

$ git check-ignore -v data/data1/a.dvc
.gitignore:3:!/data/**/*.dvc	data/data1/a.dvc
$ git check-ignore -v data/data1/a
.gitignore:1:/data/**	data/data1/a

Could you please share your .gitignore contents? it has at least 6 lines from your picture.

[Danial-Alh]

Yes I have both formats and commented mine :)

[Danial-Alh]

can it be a problem with my git version?
mine: git version 2.34.0

[karajan1001]

Quite weird here.

let me try to upgrade to a newer version.

[karajan1001]

I guess it is a bug of Git

Didn't see any thing related to gitignore algorithm in both 2.34.0 and 2.33.1

[Danial-Alh]

Not yet. However, I've just sent an email to git mailing list, describing the issue.

[Danial-Alh]

Here is the thread.

[karajan1001]

Sounds like the previous behavior is actually a bug, and had been fixed in some recent release?