threads / bug / 62976

Possible bug: Empty magic word list in pathspec is handled differently in short vs. long form

Subject: Possible bug: Empty magic word list in pathspec is handled differently in short vs. long form

## tl;dr

4 messages between Feb 19, 2025 and Feb 24, 2025.

replies: 3people: 3as markdown or json

Anselm Schüler· Feb 19, 2025, 10:49 UTC · lore
If you run git diff for a pathspec with an empty magic word/symbol list 
in short form (prefixed by ::) and in long form (prefixed by :():), you 
get different results.
Here’s an example use of git that illustrates this:
~☭ mkdir gittest
~☭ y gittest/
~/gittest☭ git init
Initialized empty Git repository in /home/anselmschueler/gittest/.git/
~/gittest (b0)☭ uuidgen >file
~/gittest (b0)☭ git add file && git commit -m "$(read)"
read> initial commit
[b0 (root-commit) cd3bd43] initial commit
  1 file changed, 1 insertion(+)
  create mode 100644 file
~/gittest (b0)☭ uuidgen >file
~/gittest (b0)☭ git diff -- "::file"

file ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

───┐ 1: │ ───┘ 66d18ffa-1128-42a5-8d4e-f9d75eb86b92 9acea2b5-50a6-4f87-ae1a-64019a93e55e ~/gittest (b0)☭ git diff -- ":():file" ~/gittest (b0)☭

Is there a reason for this? I couldn’t find it documented in the git glossary <https://git-scm.com/docs/gitglossary>.

I’m running git 2.47.2, built with Nix (nixpkgs commit 2ff53fe) from the kernel.org source tarball.

~Anselm Schüler ⚙️
Lucas Oshiro· Feb 19, 2025, 14:28 UTC · re: Anselm Schüler · lore

Re: Possible bug: Empty magic word list in pathspec is handled differently in short vs. long form

> If you run git diff for a pathspec with an empty magic word/symbol list in short form (prefixed by ::) and in long form (prefixed by :():), you get different results.
> 
From the gitglossary, pathspec section:

""" In the short form, the leading colon `:` is followed by zero or more "magic signature" letters (which optionally is terminated by another colon :), and the remainder is the pattern to match against the path. """

and:

""" In the long form, the leading colon : is followed by an open parenthesis `(`, a comma-separated list of zero or more "magic words", and a close parentheses `)` , and the remainder is the pattern to match against the path. """

Note that the long form doesn't mention that a second colon has the same meaning as in the short form. In the short form, the second colon is optionally used to separate the magic signatures from the file pattern. It wouldn't be necessary in the long form, as the close parentheses acts just like that.

This way, "::" and ":():" doesn't have the same meaning, just like, for example, ":(exclude):" doesn't mean ":!" (but ":(exclude)" does).

Junio C Hamano· Feb 19, 2025, 17:11 UTC · re: Anselm Schüler · lore

Re: Possible bug: Empty magic word list in pathspec is handled differently in short vs. long form

Anselm Schüler <mail@anselmschueler.com> writes:
> If you run git diff for a pathspec with an empty magic word/symbol
> list in short form (prefixed by ::) and in long form (prefixed by
> :():), you get different results.

The short form is prefixed by a ":" and optionally ends by a ":", but the optional termination may not be required when there is no ambiguity (e.g. ":/!path" is "from the top, not matching "path", as "p" cannot be a magic signature; ":/:!path" is "from the top, matching "!path", as the set of magic signatures ends with the second disambiguating colon).

The long form is prefixed by a ":(" and ends by a ")".
Show 25 quoted lines
> Here’s an example use of git that illustrates this:
>
> ~☭ mkdir gittest
> ~☭ y gittest/
> ~/gittest☭ git init
> Initialized empty Git repository in /home/anselmschueler/gittest/.git/
> ~/gittest (b0)☭ uuidgen >file
> ~/gittest (b0)☭ git add file && git commit -m "$(read)"
> read> initial commit
> [b0 (root-commit) cd3bd43] initial commit
>  1 file changed, 1 insertion(+)
>  create mode 100644 file
> ~/gittest (b0)☭ uuidgen >file
> ~/gittest (b0)☭ git diff -- "::file"
>
> file
> ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
>
> ───┐
> 1: │
> ───┘
> 66d18ffa-1128-42a5-8d4e-f9d75eb86b92
> 9acea2b5-50a6-4f87-ae1a-64019a93e55e
> ~/gittest (b0)☭ git diff -- ":():file"
> ~/gittest (b0)☭

Doesn't ":():file" ask for a match for the pattern ":file", with a leading colon?

Anselm Schüler· Feb 24, 2025, 14:11 UTC · re: Junio C Hamano · lore

Re: Possible bug: Empty magic word list in pathspec is handled differently in short vs. long form

Thank you and Lucas Oshiro <lucasseikioshiro@gmail.com> for the clarification

On 19/02/2025 18:11, Junio C Hamano wrote:
Show 42 quoted lines
> Anselm Schüler <mail@anselmschueler.com> writes:
>
>> If you run git diff for a pathspec with an empty magic word/symbol
>> list in short form (prefixed by ::) and in long form (prefixed by
>> :():), you get different results.
> The short form is prefixed by a ":" and optionally ends by a ":",
> but the optional termination may not be required when there is no
> ambiguity (e.g. ":/!path" is "from the top, not matching "path", as
> "p" cannot be a magic signature; ":/:!path" is "from the top,
> matching "!path", as the set of magic signatures ends with the
> second disambiguating colon).
>
> The long form is prefixed by a ":(" and ends by a ")".
>
>
>> Here’s an example use of git that illustrates this:
>>
>> ~☭ mkdir gittest
>> ~☭ y gittest/
>> ~/gittest☭ git init
>> Initialized empty Git repository in /home/anselmschueler/gittest/.git/
>> ~/gittest (b0)☭ uuidgen >file
>> ~/gittest (b0)☭ git add file && git commit -m "$(read)"
>> read> initial commit
>> [b0 (root-commit) cd3bd43] initial commit
>>   1 file changed, 1 insertion(+)
>>   create mode 100644 file
>> ~/gittest (b0)☭ uuidgen >file
>> ~/gittest (b0)☭ git diff -- "::file"
>>
>> file
>> ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
>>
>> ───┐
>> 1: │
>> ───┘
>> 66d18ffa-1128-42a5-8d4e-f9d75eb86b92
>> 9acea2b5-50a6-4f87-ae1a-64019a93e55e
>> ~/gittest (b0)☭ git diff -- ":():file"
>> ~/gittest (b0)☭
> Doesn't ":():file" ask for a match for the pattern ":file", with a
> leading colon?

← back to recent threads