Move compile-fail tests with NOTE/HELP annotations to UI

[petrochenkov]

Remove NOTE/HELP annotations from UI tests

cc #44844 @oli-obk @est31
r? @nikomatsakis

[nikomatsakis]

This looks good. We should however document the expected policy, I think, to the compiler test documentation.

I guess the idea is:

• Use ui tests and the stderr files to document notes etc.

I will note one thing (which should not hold up this PR). I've been very much enjoying the fact that UI tests check for ERRORs. I like being able to add minimal annotations to the test that give the kind of "overall parameters" for what the test is testing (that errors occur here, here, and here) and then having the stderr files capture the full behavior.

But there are times when I want to document a subset of the NOTEs that are present. i.e., so that I can outline "other variations in the stderr output may be acceptable, but if this NOTE disappears, that's a big problem". Note that I say subset -- often there are many notes, but it's just one piece I want to guarantee is present.

Before, we required that if you document any note, you must document all notes. The result is that it discourages me from annotating these notes that I want to be present.

(On the other hand, if you can't declare that the set of notes is exhaustive, you can't document that a note should not appear. I don't find this to be something I want to document very often, but it's certainly possible. But we could easily have a "NO_NOTE" annotation or something then.)

What I do for now is to leave comments, but I fear that future people will not read them.

[oli-obk]

What I do for now is to leave comments, but I fear that future people will not read them.

Maybe it would be better suited for an issue with the appropriate tags? Do we have testsuite tags?

[nikomatsakis]

@bors r+

[bors]

📌 Commit 8f58682 has been approved by nikomatsakis

[nikomatsakis]

@oli-obk

Maybe it would be better suited for an issue with the appropriate tags? Do we have testsuite tags?

I don't follow. Maye what would be better suited?

[oli-obk]

Sorry, I kind of skipped ahead.

Adding a way to make these comments annotations that can't be ignored is certainly something that makes sense. It'll prevent accidental changes down the road. What I meant was make this comment an issue so it's not forgotten.

[bors]

☔ The latest upstream changes (presumably #46558) made this pull request unmergeable. Please resolve the merge conflicts.

[nikomatsakis]

@oli-obk

Adding a way to make these comments annotations that can't be ignored is certainly something that makes sense. It'll prevent accidental changes down the road. What I meant was make this comment an issue so it's not forgotten.

yep =) #46667

[petrochenkov]

@bors r=nikomatsakis

[bors]

📌 Commit 2820616 has been approved by nikomatsakis

[bors]

⌛ Testing commit 28206165f20f9aab2eb1bdc9e27e93e64b266369 with merge 634341d84dbfa933c5ef2d6b41b9c9e554294d25...

[bors]

💔 Test failed - status-travis

[petrochenkov]

Curl failure, looks spurious.
@bors retry

[bors]

⌛ Testing commit 28206165f20f9aab2eb1bdc9e27e93e64b266369 with merge dc02dfe1058a0af48ebd811a8f744e6ef26500c9...

[bors]

💔 Test failed - status-travis

[kennytm]

@bors retry — http://crosstool-ng.org was down.

[bors]

☔ The latest upstream changes (presumably #46633) made this pull request unmergeable. Please resolve the merge conflicts.

[petrochenkov]

@bors r=nikomatsakis

[bors]

📌 Commit 66bd53a has been approved by nikomatsakis

[bors]

⌛ Testing commit 66bd53a with merge 2f35141...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: nikomatsakis
Pushing 2f35141 to master...