Stream: t-compiler

Topic: Can crater check warnings change?


Esteban Küber (Oct 29 2018 at 17:41, on Zulip):

Can crater check warnings change?
I'd like to check how much noise https://github.com/rust-lang/rust/pull/55373 would cause in the ecosystem at large. Tried just keeping it as an error, but it breaks so many things (namely in this case, crossbeam) that cargo doesn't even get to compile.

Esteban Küber (Oct 29 2018 at 17:41, on Zulip):

If we can monitor output change, not just breakage then I can change it from a hard error to a warning.

Esteban Küber (Oct 29 2018 at 17:42, on Zulip):

otherwise we're flying blind as to how much this change will break havoc

Pietro Albini (Oct 29 2018 at 17:42, on Zulip):

I don't think we can check warnings directly on crater

Pietro Albini (Oct 29 2018 at 17:43, on Zulip):

we can do a run with -Dlint_name though

Esteban Küber (Oct 29 2018 at 18:13, on Zulip):

Fair enough... That could work well enough...

Esteban Küber (Oct 29 2018 at 18:13, on Zulip):

but I will need to change the PR to be a lint

Esteban Küber (Oct 29 2018 at 18:13, on Zulip):

that might take me a while...

Esteban Küber (Oct 29 2018 at 18:13, on Zulip):

right now the check's done where it is easier, at the macro parsing stage

Esteban Küber (Oct 29 2018 at 18:14, on Zulip):

thanks @Pietro Albini, I'll go back to the drawing board and turn it into lint. It's probably what it should be regardless.

Pietro Albini (Oct 29 2018 at 18:15, on Zulip):

another option is to just execute a run with the new PR, download all the logs and analyze it locally

Esteban Küber (Oct 29 2018 at 18:25, on Zulip):

Do you have a link to the docs in order to do that?

Esteban Küber (Oct 29 2018 at 18:25, on Zulip):

I can very easily change the PR from err to warn, turning it into a lint would take me much longer

Esteban Küber (Oct 29 2018 at 18:26, on Zulip):

doing a quick grep on the results would give us actionable info the quickest

Pietro Albini (Oct 29 2018 at 18:26, on Zulip):

just start a normal run with the try toolchain

Pietro Albini (Oct 29 2018 at 18:27, on Zulip):

then, in the report there is the "downloads" section, with a tar.gz with all the logs

Esteban Küber (Oct 29 2018 at 18:28, on Zulip):

Cool!

nikomatsakis (Oct 30 2018 at 14:53, on Zulip):

@Esteban Küber wait that PR confuses me a bit

It claims that "Macro pattern ($a:expr, $b:expr) is invalid, for good reason." but I think that is not true.

Perhaps you meant $a:expr $b:expr?

Esteban Küber (Oct 30 2018 at 16:38, on Zulip):

@nikomatsakis yep, that's indeed what I meant :)

Last update: Nov 16 2019 at 02:20UTC