Stream: t-compiler/help

Topic: Test errors on master branch


Bram van den Heuvel (Jul 23 2020 at 12:46, on Zulip):

Wannabe contributor here. I'm trying to run tests on current master (2bbfa02b1b15974d5772b520aa027bf79f8c248e, ./x.py test --stage 1) but I get 5 failures. Are these supposed to fail, or did I make a mistake in my setup (e.g. using my own llvm as described in the book)? The errors

If this is not the right place to ask, please let me know.

lzutao (Jul 23 2020 at 12:48, on Zulip):

Should have the error message before this summary. Can you post them?

Bram van den Heuvel (Jul 23 2020 at 12:49, on Zulip):

https://gist.github.com/vandenheuvel/543804266f1f1924f0b834c9b1dd963b

lzutao (Jul 23 2020 at 12:54, on Zulip):

trying to run the fail tests ...

Bram van den Heuvel (Jul 23 2020 at 12:55, on Zulip):

Thanks for taking a look. Updating all submodules didn't help.

Bram van den Heuvel (Jul 23 2020 at 12:56, on Zulip):

Looks like they're all ui errors. It looks like someone forgot to --bless, but that couldn't get past CI, I assume?

lzutao (Jul 23 2020 at 13:04, on Zulip):

All failing tests passes on my machine.
Did you change Rust source somewhere?

Bram van den Heuvel (Jul 23 2020 at 13:04, on Zulip):

No, it's clean

Bram van den Heuvel (Jul 23 2020 at 13:04, on Zulip):

I only edited the configuration file, following the book

Bram van den Heuvel (Jul 23 2020 at 13:05, on Zulip):

Any other tips on how I could diagnose this?

simulacrum (Jul 23 2020 at 13:06, on Zulip):

do you by chance have parallel-compiler = true on?

lzutao (Jul 23 2020 at 13:07, on Zulip):

Can you post a diff config.toml.example config.toml?

Bram van den Heuvel (Jul 23 2020 at 13:08, on Zulip):

@simulacrum I do, that's probably it then. If I don't write in this thread in a couple of minutes, consider this fixed.

simulacrum (Jul 23 2020 at 13:08, on Zulip):

yeah I would leave it off, it's not tested in CI beyond x.py check working and isn't likely to bring too many wins at this point if any

Bram van den Heuvel (Jul 23 2020 at 13:10, on Zulip):

I hate to be cynical, but isn't rust supposed to make writing parallel code easier? Minimal gains I can understand, but different compilation results without any notification... Hm.

simulacrum (Jul 23 2020 at 13:10, on Zulip):

we haven't spent a lot of time on it

simulacrum (Jul 23 2020 at 13:11, on Zulip):

well, those tests are "known to have different results"

simulacrum (Jul 23 2020 at 13:11, on Zulip):

because all of them are cycle errors

simulacrum (Jul 23 2020 at 13:11, on Zulip):

so it's somewhat expected that parallel execution will lead to slightly different evaluation order and such

Bram van den Heuvel (Jul 23 2020 at 13:12, on Zulip):

Okay, thanks for clarifying a bit what's at play!

Last update: Sep 27 2020 at 13:45UTC