Wrong gitattributes documentation?

Wrong gitattributes documentation?

[Rene Pasing]

Hi all,

I have noticed a strange behaviour when using git-lfs.

If I understood correctly, git-lfs adds patterns to .gitattributes, for 
which git then calls lfs for any matches with this pattern.

The problem is, the documentation[1] says: "The rules how the pattern 
matches paths are the same as in .gitignore files; see gitignore[5].", 
so when I have a pattern like '/images/', it should match on all 
files+folders under /images, even the directory itself, right? Or when 
I'd use '/images/*', it should match on all files+folders under /images, 
right?

This does not seem to be the case. Nothing is done by lfs when using 
'/images/', and when using '/images/*', it only takes into account files 
which are directly in /images (so it doesnt take into account 
subdirectories).

The only solution to this (until now) is to use '/images/**/*', which 
seems to do exactly what I want (lfs gets executed for all files+folders 
under /images).

Is this expected behaviour? If yes, then the documentation of 
gitattributes seems incorrect to me. If not, can this be considered as a 
bug?

I found several complaints about this to git-lfs (or other projects 
relying on gitattributes), e.g. [2], but I didn't find anything about 
this regarding gitattributes itself.

Thanks for answers
Regards
Rene

PS: $ git --version
git version 2.1.4

[1] = https://git-scm.com/docs/gitattributes
[2] = https://github.com/git-lfs/git-lfs/issues/2068

Re: Wrong gitattributes documentation?

[Junio C Hamano]

Rene Pasing <rene@pasing.net> writes:

> The problem is, the documentation[1] says: "The rules how the pattern
> matches paths are the same as in .gitignore files; see gitignore[5].",
> so when I have a pattern like '/images/', it should match on all
> files+folders under /images, even the directory itself, right?
>
> Or when
> I'd use '/images/*', it should match on all files+folders under
> /images, right?

My reading of "PATTERN FORMAT" section in "git help ignore" says it
shouldn't.

 - If the pattern does not contain a slash '/', Git treats it as
   a shell glob pattern and checks for a match against the
   pathname relative to the location of the `.gitignore` file
   (relative to the toplevel of the work tree if not from a
   `.gitignore` file).

 - Otherwise, Git treats the pattern as a shell glob suitable
   for consumption by fnmatch(3) with the FNM_PATHNAME flag:
   wildcards in the pattern will not match a / in the pathname.
   For example, "Documentation/{asterisk}.html" matches
   "Documentation/git.html" but not "Documentation/ppc/ppc.html"
   or "tools/perf/Documentation/perf.html".

Your "/images/*" is the "Otherwise" case, isn't it?

Re: Wrong gitattributes documentation?

[Rene Pasing]

On 06/03/2017 02:01 AM, Junio C Hamano wrote:
> Your "/images/*" is the "Otherwise" case, isn't it?

Yes, true. I forgot to mention in my first mail that I also tried
'/images/', which had the same (=none) effect as '/images/*', sorry
about the confusion.

Re: Wrong gitattributes documentation?

[Rene Pasing]

On 06/03/2017 02:01 AM, Junio C Hamano wrote:
> Your "/images/*" is the "Otherwise" case, isn't it?

Ok, sorry, didn't read your answer thoroughly enough the first time.

The problem is, when I have an entry in .gitignore like this:
/images

Then git will ignore that whole directory (and all of its subfolders
etc.). Git will just ignore everything in the tree starting with that
pathname.

But when I use the same pattern ('/images') in .gitattributes I would
expect (due to the documentation) that git-lfs will be called for all
files+subfolders+subfolderfiles and even the folder itself (this is what
I understand from 'The rules how the pattern matches paths are the same
as in .gitignore files').

This does not seem to be the case!

Sorry again for the confusion.