Stream: t-compiler/rust-analyzer

Topic: Did I break it?


Laurențiu (Dec 21 2020 at 15:03, on Zulip):

Can anyone test that a master-built RA is still working? Mine just sits there saying "Rust Analyzer is still loading"

matklad (Dec 21 2020 at 15:04, on Zulip):

@Laurențiu Nicola vscode version?

Laurențiu (Dec 21 2020 at 15:04, on Zulip):

1.52.1

Laurențiu (Dec 21 2020 at 15:04, on Zulip):

Like.. it was working earlier and I don't have any local changes

Florian Diebold (Dec 21 2020 at 15:07, on Zulip):

mine is working in Emacs

Lukas Wirth (Dec 21 2020 at 15:07, on Zulip):

working for me in vscode as well with 1.52.1

Laurențiu (Dec 21 2020 at 15:10, on Zulip):

:scream:

Laurențiu (Dec 21 2020 at 15:16, on Zulip):

So it works after a reboot... :confused:

matklad (Dec 21 2020 at 15:17, on Zulip):

@Laurențiu Nicola no worries, I have your back, I broke it: https://rust-lang.zulipchat.com/#narrow/stream/185405-t-compiler.2Fwg-rls-2.2E0/topic/VS.20Code.20requirement.20is.20too.20high/near/220592392 :D

Laurențiu (Dec 22 2020 at 12:16, on Zulip):

Oof, I permanently broke CI :(. I merged that release PR manually and xtask tidy complains about a non-bors commit in the history.

Laurențiu (Dec 22 2020 at 12:20, on Zulip):
$ git pr
Current branch master is up to date.
$ git rev-list --merges --invert-grep --author 'bors\[bot\]' HEAD~19..
$
Laurențiu (Dec 22 2020 at 12:20, on Zulip):

Hmm

Laurențiu (Dec 22 2020 at 12:21, on Zulip):

image.png

Laurențiu (Dec 22 2020 at 12:21, on Zulip):

I don't get it

Laurențiu (Dec 22 2020 at 12:24, on Zulip):

image.png

Lukas Wirth (Dec 22 2020 at 12:49, on Zulip):
rust-analyzer on  master [⇡$] via  v1.48.0-nightly
❯ git rev-list --merges --invert-grep --author 'bors\[bot\]' HEAD~19..
02d8cf82ca3ae8f78a642e5c5afc3ade02caa300

I do get a commit when invoking the command after pulling latest master

Lukas Wirth (Dec 22 2020 at 12:49, on Zulip):

xtask also fails for me

Laurențiu (Dec 22 2020 at 13:02, on Zulip):

Ah, my master branch was tracking my fork. But I still don't know how to fix it.

Lukas Wirth (Dec 22 2020 at 13:11, on Zulip):

The only way I see to fix this would be to remove the merge via force push, which sounds less than ideal

Lukas Wirth (Dec 22 2020 at 13:11, on Zulip):

This also means CI will always break if bors isnt the one that merged doesn't it

Laurențiu (Dec 22 2020 at 13:11, on Zulip):

Or to quickly merge 20 pull requests

Lukas Wirth (Dec 22 2020 at 13:12, on Zulip):

But that would require bors to merge them all, and since CI fails bors won't merge them will it :sweat_smile:

Jonas Schievink [he/him] (Dec 22 2020 at 13:12, on Zulip):

force-pushed

Jonas Schievink [he/him] (Dec 22 2020 at 13:12, on Zulip):

If we force-push quickly it shouldn't cause issues

Laurențiu (Dec 22 2020 at 13:14, on Zulip):

Filed another PR

Laurențiu (Dec 22 2020 at 13:14, on Zulip):

So is there a way to merge --ff-only it from the GitHub UI? Does "Rebase and merge" do that?

Jonas Schievink [he/him] (Dec 22 2020 at 13:15, on Zulip):

No, I don't think there is

Jonas Schievink [he/him] (Dec 22 2020 at 13:15, on Zulip):

Why doesn't bors work?

Laurențiu (Dec 22 2020 at 13:15, on Zulip):

Wait, bors just started with https://github.com/rust-analyzer/rust-analyzer/pull/6989

Jonas Schievink [he/him] (Dec 22 2020 at 13:16, on Zulip):

Hah, yeah

Laurențiu (Dec 22 2020 at 13:17, on Zulip):

Jonas Schievink said:

Why doesn't bors work?

It used to not work on PRs with workflow changes or that would merge over other workflow changes (e.g. you filed a PR, then the workflows changed on master). It would time out or not run at all.

Jonas Schievink [he/him] (Dec 22 2020 at 13:18, on Zulip):

oh, weird. I'd expect that to work fine.

Jonas Schievink [he/him] (Dec 22 2020 at 13:18, on Zulip):

bors doesn't even look at workflow files

Laurențiu (Dec 22 2020 at 13:18, on Zulip):

I think it was a permissions issue (integrations not being able to modify workflows)

Laurențiu (Dec 22 2020 at 13:18, on Zulip):

But it might be working fine now, I remember seeing it merge (only) one PR

Laurențiu (Dec 22 2020 at 13:20, on Zulip):

image.png weee

matklad (Dec 22 2020 at 15:23, on Zulip):

Regarding force pushes -- imo, they are totally OK and overly-demonized. It's like usnafe -- ok, if you know what you are doing

matklad (Dec 22 2020 at 15:23, on Zulip):

So feel free to push --force-with-lease if you need to (but you should understand the difference between --force & --force-with-lease ;) )

bjorn3 (Dec 22 2020 at 15:33, on Zulip):

My university completely disabled force pushes even when you disable branch protection by using a push hook. I once had to switch the default branch, delete the original default branch and then recreate it without the commit I wanted to remove.

bjorn3 (Dec 22 2020 at 15:35, on Zulip):

And yes, --force-with-lease is much better than --force. --force-with-lease should really have been called --force and --force --really-really-force-this-action.

matklad (Dec 22 2020 at 15:39, on Zulip):

To be fair, this used to be reasonable-ish: previously, git defaulted to pushing all branches, and not only the current one, and using --force with that was problematic

matklad (Dec 22 2020 at 15:39, on Zulip):

IIRC, --force-with-lease will become default in some future git version

Laurențiu (Dec 22 2020 at 15:40, on Zulip):

Also:

Some editors like Microsoft's VSC have a feature to auto-fetch in the background, this bypasses the protections offered by --force-with-lease

bjorn3 (Dec 22 2020 at 15:41, on Zulip):

I have that disabled.

Last update: Jul 29 2021 at 08:15UTC