git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v3 1/3] worktree: remove "the_repository" from is_current_worktree()

From
Junio C Hamano <gitster@pobox.com>
Date
Mar 27, 2026, 17:07 UTC
Message-ID
<xmqq1ph5kxpx.fsf@gitster.g>
In-Reply-To
<317ed4f7-f88d-4415-bd25-b62b4a076728@gmail.com>
Phillip Wood <phillip.wood123@gmail.com> writes:
Show 18 quoted lines
> On 26/03/2026 15:48, Junio C Hamano wrote:
>> Phillip Wood <phillip.wood123@gmail.com> writes:
>> 
>>> ... it calls get_workree_git_dir() which also depends on
>>> "the_repository". This means that even if "wt->path" matches
>>> "wt->repo->worktree" is_current_worktree(wt) will return false when
>>> "wt->repo" is not "the_repository". Consequently die_if_checked_out()
>>> will fail to skip such a worktree when checking if a branch is already
>>> checked out and may die errounously. Fix this by using the worktree's
>>> repository instance instead of "the_repository" when comparing gitdirs.
>> 
>> It sounds like the above is something we can write in a test to make
>> sure the code with this fix won't regress in the future.  Can we add
>> one?
>
> We'd need to create a struct worktree instance with a repository 
> instance that is not "the_repository" - I'm not sure we can do that in a 
> script.

Ah, so this is more of "futureproofing" than "fix"? That is fine. Thanks for clarifying.

Show 22 quoted lines
>>> -	wt->is_current = is_current_worktree(wt);
>>> +	wt->is_current = true;
>>>   	add_head_info(wt);
>>>   
>>>   	free(gitdir);
>> 
>> I think I found the semantics of get_worktree_from_repository()
>> unclear when we discussed a different patch series, so I may have
>> asked the same question already, but what exactly does it mean to
>> "get worktree from repository", i.e., this function does?  A
>> repository can have one or more worktrees attached to it (that was
>> the whole point of introducing the worktree mechanism), so "I have
>> this repository, give me the worktree for it" is not a sensible
>> request.
>
> A repository can have more that one worktree but a "struct repository" 
> instance has "gitdir" and "worktree" members that point to a specific 
> worktree within that repository. For example
>
> 	repo_get_oid(repo, "HEAD", &oid);
>
> reads "HEAD" from a specific worktree within the repository.
And...?

I _think_ what I am frustrated about is the lack of description on "what it means to be the worktree among many that is pointed by via the .worktree member in the repository struct". Does it correspond to the worktree being "the current worktree the codepath is working on?"

Show 9 quoted lines
>> The function's comment in <worktree.h> talks only at the
>> implementation level "construct a worktree struct from repo->gitdir
>> and repo->worktree" as if it is so obvious what the resulting
>> worktree struct means at a higher layer's point of view, which does
>> not help, either.
>
> That comes from me thinking of a struct repository as referring to a 
> specific worktree - would calling it 
> "get_worktree_from_repository_instance" be clearer?

It does not change the descriptive value of the name in any meaningful way, so let's not do that. If the answer to my "what frustrates me" comment above is "yeah, we are getting the current worktree", then renaming the function to include "current" in its name would add descriptive value vastly, though.

Show 15 quoted lines
> I feel I'm struggling to explain this clearly - I find this whole 
> discussion gets confusing because we have "struct worktree" and also a 
> "worktree" member of "struct repository" which means a "struct 
> repository" instance is tied to a specific worktree within the 
> repository. If "struct repository" only had a "commondir" member and no 
> "gitdir" or "worktree" members and we instead used "sturct worktree" to 
> refer to a specific worktree within a repository with functions like
>
> 	worktree_get_oid(wt, "HEAD", &oid);
>
> instead of
>
> 	repo_get_oid(repo, "HEAD", &oid);
>
> it might be clearer but that would be a very big change.

In short, am I hearing the worktree subsystem is not conceptually clean and it would be a huge undertaking to clean it up?

Previous: Phillip WoodNext: Phillip Wood
Message 24 of 27 in “worktree: stop using "the_repository" in is_current_worktree()”
  1. 0/3 worktree: stop using "the_repository" in is_current_worktree()Phillip Wood, Mar 13, 2026
  2. 1/3 worktree: remove "the_repository" from is_current_worktree()Phillip Wood, Mar 13, 2026
  3. 2/3 worktree add: stop reading ".git/HEAD"Phillip Wood, Mar 13, 2026
  4. Junio C HamanoMar 13, 2026
  5. 3/3 worktree: reject NULL worktree in get_worktree_git_dir()Phillip Wood, Mar 13, 2026
  6. Junio C HamanoMar 13, 2026
  7. Phillip WoodMar 14, 2026
  8. 0/3 worktree: stop using "the_repository" in is_current_worktree()Phillip Wood, Mar 15, 2026
  9. 1/3 worktree: remove "the_repository" from is_current_worktree()Phillip Wood, Mar 15, 2026
  10. Patrick SteinhardtMar 16, 2026
  11. Phillip WoodMar 16, 2026
  12. Phillip WoodMar 17, 2026
  13. Shreyansh PaliwalMar 23, 2026
  14. Phillip WoodMar 23, 2026
  15. Shreyansh PaliwalMar 23, 2026
  16. 2/3 worktree add: stop reading ".git/HEAD"Phillip Wood, Mar 15, 2026
  17. Patrick SteinhardtMar 16, 2026
  18. 3/3 worktree: reject NULL worktree in get_worktree_git_dir()Phillip Wood, Mar 15, 2026
  19. Junio C HamanoMar 15, 2026
  20. 0/3 worktree: stop using "the_repository" in is_current_worktree()Phillip Wood, Mar 26, 2026
  21. 1/3 worktree: remove "the_repository" from is_current_worktree()Phillip Wood, Mar 26, 2026
  22. Junio C HamanoMar 26, 2026
  23. Phillip WoodMar 27, 2026
  24. Junio C HamanoMar 27, 2026
  25. Phillip WoodApr 2, 2026
  26. 2/3 worktree add: stop reading ".git/HEAD"Phillip Wood, Mar 26, 2026
  27. 3/3 worktree: reject NULL worktree in get_worktree_git_dir()Phillip Wood, Mar 26, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.