1566434 - Assert there's a unique entry for pc + kind in BaselineScript::retAddrEntryFromPCOffset

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, task, P1)

Description

[Jan de Mooij [:jandem]]

We really don't want to find duplicates there and should assert against this.

Comment 1

[Jan de Mooij [:jandem]]

Actually it's probably simpler and better to assert this in the compiler; the sooner we fail the better.

Comment 2

[Jan de Mooij [:jandem]]

Attachment: Bug 1566434 part 1 - Remove markLastRetAddrKind in BaselineCodeGen, use the correct Kind from the start. r?iain!

This is nicer and will make it easier to assert RetAddrEntry invariants in
appendRetAddrEntry in the next patch.

Comment 3

[Jan de Mooij [:jandem]]

(In reply to Jan de Mooij [:jandem] from comment #1)

Actually it's probably simpler and better to assert this in the compiler; the sooner we fail the better.

While biking last night it occurred to me this doesn't work because the invariant only holds for the callVMs (or bytecode ops) that can trigger DebugOSR. Oh well.

Comment 4

[Jan de Mooij [:jandem]]

Attachment: Bug 1566434 part 2 - Add more RetAddrEntry assertions. r?iain!

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/bdb913e6361e
part 1 - Remove markLastRetAddrKind in BaselineCodeGen, use the correct Kind from the start. r=iain
https://hg.mozilla.org/integration/autoland/rev/08fa2abfae88
part 2 - Add more RetAddrEntry assertions. r=iain

Comment 6

[Oana Pop-Rus]

https://hg.mozilla.org/mozilla-central/rev/bdb913e6361e
https://hg.mozilla.org/mozilla-central/rev/08fa2abfae88