force current stage when --stage is expilicitly used

[onur-ozkan]

Allows users to build std with stage2 compiler without using full-bootstrap.

Fixes #118233
Fixes #118918

[rustbot]

r? @clubby789

(rustbot has picked a reviewer for you, use r? to override)

[rustbot]

This PR modifies src/bootstrap/src/core/config.

If appropriate, please update CONFIG_CHANGE_HISTORY in src/bootstrap/src/utils/change_tracker.rs and change-id in config.example.toml.

[onur-ozkan]

Do you have any concerns on this? @Mark-Simulacrum

[onur-ozkan]

r? Mark-Simulacrum

[Mark-Simulacrum]

This condition feels iffy to me as being first -- if rustbuild wants to force some stage, shouldn't we respect that? Or should this condition be made part of force_use_stage1/2?

It seems like perhaps those functions should just be no-ops with explicit stage set?

Or perhaps more/different work is warranted; it seems odd to me that --stage opts-in to behavior different than what the default is. If we think the right thing to do is to build stage 2 when it's requested on the CLI, shouldn't we do the same when it's the implicit default? (IIRC, we made a change to make stage 1 the default in most modes, so presumably this wouldn't be a change in default behavior for most people?)

[onur-ozkan]

This condition feels iffy to me as being first -- if rustbuild wants to force some stage, shouldn't we respect that?

Not if the stage was explicitly requested by the user, in my opinion.

From #118233 (comment)

I don't want to compile the compiler three times. I just want the standard library built with the stage that I explicitly specified.

This seems like a very reasonable request for adding an interface that allows users to use their desired stage when they want. I am aware of that this use case is quite rare, but we should still provide a way for users to enforce their desired stage.

If we think the right thing to do is to build stage 2 when it's requested on the CLI, shouldn't we do the same when it's the implicit default?

This is very true. Maybe we should add another flag to bootstrap something like --force-stage with explaining how its different than --stage one?

[Mark-Simulacrum]

I guess the question I'm asking is why do we need force-stage separate from just stage? I.e., can we just make the behavior of adjusting the stage (--stage N or default stage N) the same?

[onur-ozkan]

There are several cases where force_use_stage1/2 make sense. By default they should be enabled as they can significantly reduce the build time. But for certain cases we should also allow developers to disable/ignore them.

[Mark-Simulacrum]

@onur-ozkan is this still relevant, or has subsequent work replaced it? I'll need to page it back into cache but I think the open comment thread is still something I haven't had my mind changed on yet...

[onur-ozkan]

I need to check this again and see if we can improve the stage handling logic without adding explicit_stage field.

@rustbot author

[bors]

☔ The latest upstream changes (presumably #131988) made this pull request unmergeable. Please resolve the merge conflicts.

[onur-ozkan]

#t-infra/bootstrap > stepping down from T-Bootstrap