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

Re: [PATCH] read-cache: use istate->repo in overlay_tree_on_index()

From
Junio C Hamano <gitster@pobox.com>
Date
Nov 16, 2026, 16:51 UTC
Message-ID
<xmqqtslglnxk.fsf@gitster.g>
In-Reply-To
<b7abaeba-1dd9-45f3-ad63-e43c49ee48d4@gmail.com>
Derrick Stolee <stolee@gmail.com> writes:
Show 11 quoted lines
> On 11/16/2026 5:34 AM, Srujith A Poojary via GitGitGadget wrote:
>> From: Srujith A Poojary <poojarysrujith@gmail.com>
>> 
>> Use the repository associated with the index state when
>> resolving the tree name, parsing the tree, and reading
>> its entries.
>> 
>> This avoids relying on the global repository when the
>> function already has access to its associated repository.
>
> Cleanups are good.
At least this one is not making things any worse.

But there are other functions in this file, like add_to_index() and refresh_index(), that take istate but still mention the_repository. Cleaning these functions that share the same issue together in a single patch would help us reduce the review load and patch churn.

On the other hand, I do not think it is a good idea to try losing USE_THE_REPOSITORY from this file all at once. There are many lower-level functions that do not take istate (or repo), and it needs a design based on deeper thinking (which has not been made here for the obvious reason of not needing one yet) to decide what to do with them [*]. We do not want to see such changes in the same series of patches as a mechanical replacement patch like this one.

Thanks.
[Footnote]
 * Passing an extra parameter, either repo or istate, to these
   functions may be an easy way, but the overhead may make it a
   less-than-optimal solution, for example.  Teaching ce which
   istate it belongs to is an absolute no-no, as these things are
   born without their own istate, and are transfered between
   different istates; besides they are so numerous.  Those are
   merely one or two examples that need to be considered before we
   know what the right approach is.
Show 16 quoted lines
>> +	struct repository *repo = istate->repo;
>
>> -	if (repo_get_oid(the_repository, tree_name, &oid))
>> +	if (repo_get_oid(repo, tree_name, &oid))
>
>> -	tree = repo_parse_tree_indirect(the_repository, &oid);
>> +	tree = repo_parse_tree_indirect(repo, &oid);
>
>> -	err = read_tree(the_repository, tree, &pathspec, fn, istate);
>> +	err = read_tree(repo, tree, &pathspec, fn, istate);
> The only other way to do this would be to replace the_repository
> with 'istate->repo' directly, but having a 'repo' variable is
> just as good.
>
> LGTM.
> -Stolee
Previous: Derrick StoleeNext: Srujith A Poojary via GitGitGadget
Message 3 of 4 in “read-cache: use istate->repo in overlay_tree_on_index()”
  1. read-cache: use istate->repo in overlay_tree_on_index()Srujith A Poojary via GitGitGadget, Nov 16, 2026
  2. Derrick StoleeNov 16, 2026
  3. Junio C HamanoNov 16, 2026
  4. read-cache: use istate->repo in index operationsSrujith A Poojary via GitGitGadget, Nov 18, 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.