Stream: t-compiler

Topic: test suite org, abi tests


nikomatsakis (Aug 14 2019 at 15:34, on Zulip):

I'm inclined to r+ https://github.com/rust-lang/rust/pull/62593 -- anybody have strong opinions? It moves abi related tests from src/test/run-pass into src/test/run-pass/abi. In some cases, also from directiories like src/test/run-pass/struct to src/test/run-pass/abi/struct. I'm not sure what the ideal org is but it seems like an improvement. @centril, you may care about this...

centril (Aug 14 2019 at 15:35, on Zulip):

@nikomatsakis agree that it is better organization... but something weird is going on; it seems to be moving tests in the run-pass directory... but we removed it?

centril (Aug 14 2019 at 15:36, on Zulip):

(i.e. why is there not a merge conflict... will investigate)

centril (Aug 14 2019 at 15:37, on Zulip):

(oh no, we have a test in https://github.com/rust-lang/rust/blob/master/src/test/run-pass/generator/niche-in-generator.rs that isn't being tested)

centril (Aug 14 2019 at 15:38, on Zulip):

(see https://github.com/rust-lang/rust/tree/master/src/test/run-pass -- there's just a single test ^--- there)

centril (Aug 14 2019 at 15:40, on Zulip):

cc @Vadim Petrochenkov ^---

Vadim Petrochenkov (Aug 14 2019 at 15:43, on Zulip):

It seems to always happen when test suites are removed.

centril (Aug 14 2019 at 15:46, on Zulip):

@Vadim Petrochenkov any idea why the PR doesn't have a merge conflict tho?

centril (Aug 14 2019 at 15:47, on Zulip):

(I'll go move the generator test)

RalfJ (Aug 14 2019 at 15:55, on Zulip):

Vadim Petrochenkov any idea why the PR doesn't have a merge conflict tho?

git is file-based, not dir-based

RalfJ (Aug 14 2019 at 15:55, on Zulip):

so it doesnt care if you removed the dir

RalfJ (Aug 14 2019 at 15:55, on Zulip):

and there's no conflict on the file level

centril (Aug 14 2019 at 15:56, on Zulip):

@RalfJ oh, but so what happens if we merge the PR? does it get moved back to run-pass?

RalfJ (Aug 14 2019 at 15:57, on Zulip):

nope

RalfJ (Aug 14 2019 at 15:57, on Zulip):

git follows renames of files, but not of dirs -- it doesnt even have that concept

RalfJ (Aug 14 2019 at 15:57, on Zulip):

wait, what is "it"?

RalfJ (Aug 14 2019 at 15:58, on Zulip):

that old generator thing was likely a PR adding a new test getting rebased against the test suite rename

RalfJ (Aug 14 2019 at 15:58, on Zulip):

new file being created in a renamed dir, git doesnt even notice anything weird is going on

centril (Aug 14 2019 at 15:58, on Zulip):

@RalfJ https://github.com/rust-lang/rust/pull/62593 reorganizes files in the run-pass dir, but that dir does not exist anymore (well... it does... but those files don't)

RalfJ (Aug 14 2019 at 15:59, on Zulip):

git could be "smart" enough that if one PR renames A to B, and then another one renames A to C, when the former gets rebased it results in a B -> C rename

RalfJ (Aug 14 2019 at 15:59, on Zulip):

but I haven't tried. but that seems consistent with following renames.

centril (Aug 14 2019 at 15:59, on Zulip):

(the github UI shows src/test/run-pass/abi-sysv64-arg-passing.rs → src/test/run-pass/abi/abi-sysv64-arg-passing.rs tho)

RalfJ (Aug 14 2019 at 16:00, on Zulip):

(I recently rebased something over some rename and was super-happy that that just worked. would have been extremely painful.)

RalfJ (Aug 14 2019 at 16:00, on Zulip):

(the github UI shows src/test/run-pass/abi-sysv64-arg-passing.rs → src/test/run-pass/abi/abi-sysv64-arg-passing.rs tho)

GH doesnt re-render when the target gets updated

RalfJ (Aug 14 2019 at 16:00, on Zulip):

author needs to rebase and force-push for an updated diff

centril (Aug 14 2019 at 16:00, on Zulip):

ah

centril (Aug 14 2019 at 16:01, on Zulip):

I'll note that as a comment then

nikomatsakis (Aug 14 2019 at 19:32, on Zulip):

good catch @centril <3

Last update: Nov 22 2019 at 05:35UTC