1603726 - Spidermonkey bustages at js/src/jit-test/tests/wasm/passive-segs-nonboundary.js - when Gecko 73 merges to Beta on 2020-01-06

Status: VERIFIED FIXED

(Core :: JavaScript: WebAssembly, defect, P2)

Description

[Oana Pop-Rus]

[Tracking Requested - why for this release]:

Central as beta simulation: https://treeherder.mozilla.org/#/jobs?repo=try&resultStatus=testfailed%2Cbusted%2Cexception%2Cretry%2Cusercancel%2Crunnable&revision=b9ce17db3ec31fca39cc450f9ce4cc73750462cd&selectedJob=281038041

Log link: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=281038041&repo=try&lineNumber=40363

Lars: this seems to be a regression from Bug 1599125, can you please take a look?

[task 2019-12-13T10:49:07.032Z] TEST-PASS | js/src/jit-test/tests/wasm/passive-segs-boundary.js | Success (code 0, args "--test-wasm-await-tier2") [0.6 s]
[task 2019-12-13T10:49:07.272Z] /builds/worker/workspace/build/src/js/src/jit-test/lib/wasm.js:12:13 CompileError: at offset 100: initializer type must be subtype of element type
[task 2019-12-13T10:49:07.272Z] Stack:
[task 2019-12-13T10:49:07.272Z] wasmEvalText@/builds/worker/workspace/build/src/js/src/jit-test/lib/wasm.js:17:9
[task 2019-12-13T10:49:07.272Z] @/builds/worker/workspace/build/src/js/src/jit-test/tests/wasm/passive-segs-nonboundary.js:358:27
[task 2019-12-13T10:49:07.272Z] Exit code: 3
[task 2019-12-13T10:49:07.272Z] FAIL - wasm/passive-segs-nonboundary.js
[task 2019-12-13T10:49:07.273Z] TEST-UNEXPECTED-FAIL | js/src/jit-test/tests/wasm/passive-segs-nonboundary.js | /builds/worker/workspace/build/src/js/src/jit-test/lib/wasm.js:12:13 CompileError: at offset 100: initializer type must be subtype of element type (code 3, args "") [0.5 s]
[task 2019-12-13T10:49:07.273Z] INFO exit-status : 3
[task 2019-12-13T10:49:07.273Z] INFO timed-out : False
[task 2019-12-13T10:49:07.273Z] INFO stderr 2> /builds/worker/workspace/build/src/js/src/jit-test/lib/wasm.js:12:13 CompileError: at offset 100: initializer type must be subtype of element type
[task 2019-12-13T10:49:07.273Z] INFO stderr 2> Stack:
[task 2019-12-13T10:49:07.273Z] INFO stderr 2> wasmEvalText@/builds/worker/workspace/build/src/js/src/jit-test/lib/wasm.js:17:9
[task 2019-12-13T10:49:07.273Z] INFO stderr 2> @/builds/worker/workspace/build/src/js/src/jit-test/tests/wasm/passive-segs-nonboundary.js:358:27
[task 2019-12-13T10:49:07.284Z] /builds/worker/workspace/build/src/js/src/jit-test/lib/wasm.js:12:13 CompileError: at offset 100: initializer type must be subtype of element type
[task 2019-12-13T10:49:07.284Z] Stack:
[task 2019-12-13T10:49:07.284Z] wasmEvalText@/builds/worker/workspace/build/src/js/src/jit-test/lib/wasm.js:17:9
[task 2019-12-13T10:49:07.284Z] @/builds/worker/workspace/build/src/js/src/jit-test/tests/wasm/passive-segs-nonboundary.js:358:27
[task 2019-12-13T10:49:07.284Z] Exit code: 3

Comment 1

[Lars T Hansen [:lth]]

yeah, will look today. thanks.

Comment 2

[Lars T Hansen [:lth]]

Oy, what a boondoggle.

There are for sure test cases in wasm/passive-segs-nonboundary.js that should not be there, because they test reference-type functionality, and such tests should be in wasm/gc.

But there are also failures in wasm/spec/binary.wast.js, wasm/spec/bulk.wast.js, and wasm/spec/elem.wast.js. It is possible these have the same problem; TBD.

Comment 3

[Lars T Hansen [:lth]]

It looks like, for the wasm/spec tests, they are broken because the latest update was pulled from the bulk memory proposal repo. What I don't understand is why these tests don't break Beta, since the update landed before the last merge day.

Comment 4

[Lars T Hansen [:lth]]

Attachment: Bug 1603726 - Move tests to wasm/gc. r=rhunt

Comment 5

[Ryan Hunt [:rhunt]]

From discussing on irc, I believe that the issue is more fallout from bug 1598377 [1]. It seems that we were mistakenly allowing the spec tests of non-default features to pass, and that commit stopped this. The solution for this is going to be to update the spec tests to only be from the main spec repo (no proposals).

I put together a quick try run here [2]. Unfortunately there's some test failures... The issue is that one of the bulk-memory changes is technically a breaking change. I'm going to file an issue on the repo to try to resolve this upstream.

[1] https://hg.mozilla.org/mozilla-central/rev/9d52a8119f41
[2] https://treeherder.mozilla.org/#/jobs?repo=try&revision=2cec5d520c3bac55bad2df52133355628384e811

Comment 6

[Ryan Hunt [:rhunt]]

https://github.com/WebAssembly/bulk-memory-operations/issues/133

Comment 7

[Arthur Iakab [arthur_iakab]]

The spidermonkey bustages are still occurring with the patch applied.
https://treeherder.mozilla.org/#/jobs?repo=try&resultStatus=testfailed%2Cbusted%2Cexception&classifiedState=unclassified&revision=d39d7e751ba1bb824e34d9f2435f8a366c182cb1&searchStr=spidermonkey

Failure log https://treeherder.mozilla.org/logviewer.html#?job_id=281249440&repo=try

After backingout Bug 1599125 the bustages stopped occurring .

Comment 8

[Lars T Hansen [:lth]]

(In reply to Arthur Iakab [arthur_iakab] from comment #7)

The spidermonkey bustages are still occurring with the patch applied.

Yes, because the patch only fixes some of the errors, namely the first set of errors diagnosed by comment 2. The second set of errors needs a different patch, and since the tests that fail are imported from another repo and need to be autogenerated from that repo, that repo needs to be patched first, see comment 6 for details.

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/572a5cc1d0b5
Move tests to wasm/gc. r=rhunt

Comment 10

[Ryan Hunt [:rhunt]]

I'm working on a spec-repo commit we can use to work around this issue while it is figured out upstream.

Comment 11

[Ryan Hunt [:rhunt]]

Commit passed locally for me, but here's a try run to be sure.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=e999e1b78dbb7d8939337ab856ca95b91fb5fc6c

Comment 12

[Ryan Hunt [:rhunt]]

Attachment: Bug 1603726 part 1 - Switch back to main repo for wasm spec-tests. r?lth

This commit updates the spec tests to be based off the main repo with no
proposals.

Two commits are grafted on the branch. The first is a pre-existing commit to
drop a test that isn't valid with a bottom type. The second is a new commit to
allow segment initialization failures to be reported as a runtime or link error.

[1] https://github.com/eqrion/wasm-spec/tree/spidermonkey-tree-tests
[2] https://github.com/eqrion/wasm-spec/commit/f8044be9f7bc2b3c7add51251efc6276c6ce0ce4
[3] https://github.com/eqrion/wasm-spec/commit/b8836785dd82bdcb9fd0f75406fa7418ce7417d2

Comment 13

[Ryan Hunt [:rhunt]]

Attachment: Bug 1603726 part 2 - Switch back to main repo for wasm spec-tests. r?lth

Depends on D57335

Comment 14

[Natalia Csoregi [:nataliaCs]]

https://hg.mozilla.org/mozilla-central/rev/572a5cc1d0b5

Comment 15

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/6ed5d68689f1
part 1 - Switch back to main repo for wasm spec-tests. r=lth
https://hg.mozilla.org/integration/autoland/rev/c42b5a8f9bf9
part 2 - Switch back to main repo for wasm spec-tests. r=lth

Comment 16

[Andreea Pavel [:apavel]]

https://hg.mozilla.org/mozilla-central/rev/6ed5d68689f1
https://hg.mozilla.org/mozilla-central/rev/c42b5a8f9bf9

Comment 17

[Ryan VanderMeulen [:RyanVM]]

Does this bug need to be open still?

Comment 18

[Lars T Hansen [:lth]]

It does not.

Comment 19

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

Verified fixed in latest beta simulation:

https://treeherder.mozilla.org/#/jobs?repo=try&resultStatus=testfailed%2Cbusted%2Cexception%2Cretry%2Cusercancel%2Crunnable&revision=6877759d8c66c68181fde8a49d327b14119437ae