bootstrap: use git merge-base for LLVM CI download logic#113588
bootstrap: use git merge-base for LLVM CI download logic#113588bors merged 2 commits intorust-lang:masterfrom
Conversation
|
This PR changes how LLVM is built. Consider updating src/bootstrap/download-ci-llvm-stamp. |
This comment has been minimized.
This comment has been minimized.
|
Hm, this doesn't seem to work on CI already... Does that checkout have the right remote under some other name? Or should we fall back to just |
8babafe to
28ee7ba
Compare
This comment has been minimized.
This comment has been minimized.
|
don't have time for reviews r? bootstrap |
|
Looks like CI doesn't have any But I am confused that the logic even gets called there, doesn't CI need to always build LLVM itself anyway? The old logic would have returned the just-created fresh bors merge commit, and then it would have failed to find an LLVM build for that commit -- right? |
see the logic in cc #113250 |
|
@rustbot author |
|
☔ The latest upstream changes (presumably #113637) made this pull request unmergeable. Please resolve the merge conflicts. |
8df83e4 to
a8d6456
Compare
|
CI only runs when there are no merge conflicts, because it runs on the merge result. |
a8d6456 to
49d29c0
Compare
This comment has been minimized.
This comment has been minimized.
|
Interesting, the rev-list path filtering seems to be doing absolutely nothing on CI. I guess this is because of the shallow clone. I see two options for what we could do here:
@albertlarsan68 any preferences? |
|
If you are able to make option 1 work, the go ahead. Otherwise, I think the option 2 is a good compromise. |
a4ca940 to
a4c9ca4
Compare
|
I couldn't quite figure out the best way to do this, so I went with a different approach: we just use |
|
@rustbot ready |
a4c9ca4 to
bdc3ed5
Compare
albertlarsan68
left a comment
There was a problem hiding this comment.
LGTM, r=me with or without the change
|
@bors r=albertlarsan68 |
…iaskrgr Rollup of 7 pull requests Successful merges: - rust-lang#111081 (impl SliceIndex<str> for (Bound<usize>, Bound<usize>)) - rust-lang#113394 (style-guide: Document style editions, start 2024 style edition) - rust-lang#113588 (bootstrap: use git merge-base for LLVM CI download logic) - rust-lang#113743 (Directly link more target docs) - rust-lang#114262 (Improve the rust style guide doc) - rust-lang#114309 (Update books) - rust-lang#114313 ([rustc_data_structures] Simplify SortedMap::insert.) r? `@ghost` `@rustbot` modify labels: rollup
Revert rust-lang#113588 to fix bootstrap timings This reverts rust-lang#113588 which seems to have broken perf's bootstrap timings via some git issue rust-lang#114318 (comment) show a newly broken benchmark, the error at the time was ``` fatal: Path 'src/ci/channel' exists on disk, but not in 'e62323df22ecf9c163023132d17b7114f68b72e8'. thread 'main' panicked at 'command did not execute successfully: cd "/home/collector/rustc-perf/rust" && "git" "show" "e62323df22ecf9c163023132d17b7114f68b72e8:src/ci/channel" expected success, got: exit status: 128', config.rs:1786:27 ``` If this lands, it will reopen rust-lang#101907 and annoy miri, but it could actually be an issue that would appear during the next bootstrap bump, not just rustc-perf today. r? `@ghost`
bootstrap: use git merge-base for LLVM CI download logic This re-lands rust-lang#113588, now that the perf issues are hopefully fixed by rust-lang/rustc-perf#1684. r? `@lqd` `@Mark-Simulacrum` Fixes rust-lang#101907
bootstrap: use git merge-base for LLVM CI download logic This re-lands rust-lang/rust#113588, now that the perf issues are hopefully fixed by rust-lang/rustc-perf#1684. r? `@lqd` `@Mark-Simulacrum` Fixes rust-lang/rust#101907
Fixes #101907
I tested this with a local branch that has extra merge commits due to Miri, and it worked fine there. But I am sure there are tons of other situations I did not think of...
r? @jyn514