Re: [PATCH 0/4] environment: move line ending config variables into repo_config_values
- From
Patrick Steinhardt <ps@pks.im>
- Date
- Oct 26, 2026, 12:27 UTC
- Message-ID
- <at9HOXZhVcl53CGX@pks.im>
- In-Reply-To
- <20261024220012.486-1-aniketgupta3001@gmail.com>
Hi,
On Sat, Oct 24, 2026 at 10:00:08PM +0000, Aniket Gupta wrote:
Show 13 quoted lines
> Hi, > > This series moves the global variables behind core.autocrlf, core.eol > and core.safecrlf into 'struct repo_config_values' to continue the > libification effort. > > The first patch passes 'repo' to the working tree conversion > functions which need it to read the values. The other three move one > variable each just like the earlier conversions. > > An alternative to passing 'repo' would be to store the resolved values > in 'struct conv_attrs'. I went with passing it since that is what the > earlier conversions did, but I can change it if you prefer.
This isn't really your fault, but we've had the discussion multiple times by now that we need to stop moving more config values into `repo_config_values` until we have a way for how to decouple it from `the_repository`. The problem here is that the only way to populate those settings is via `repo_config(git_default_config)`, and that function will populate the repo config values of `the_repository`. Consequently, `repo_config_values()` BUGs when called with a different repostiory.
While these patches _look_ like they libify Git more, they actually do the exact opposite right now and tie us more and more to `the_repository`. So I'm inclined to say that we should hold off merging these patches until we fix that issue, or otherwise we paint ourselves into a corner.
Patrick