1192304 - Common up the checks we do in abortGC and collect

Status: RESOLVED FIXED

(Core :: JavaScript: GC, defect)

Description

[Terrence Cole [:terrence]]

Attachment: 1_split_out_per_gc_event_logic-v0.diff

At the start of every collection event, we check that the runtime is in a valid place for any sort of GC event to have occurred. These checks appear to have diverged between abortGC and collect. I also split off the is-gc-allowed check because somehow the TraceLogger event was getting fired before we'd finished these checks: probably doesn't matter because we're unlikely to run with both trace logging and deterministic mode, but still worth fixing.

Comment 1

[Jon Coppeard (:jonco) (PTO until 2nd March)]

Comment on attachment 8645025 [details] [diff] [review]
1_split_out_per_gc_event_logic-v0.diff

Review of attachment 8645025 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/gc/GCRuntime.h
@@ +897,5 @@
> +    void resetIncrementalGC(const char* reason);
> +
> +    // Assert if the system state is such that we should never
> +    // receive a request to do GC work.
> +    void sanityCheckGCEventRequest();

I can't help thinking there's a better name for this... checkCanCallAPI() maybe?

Comment 2

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/9c9f970f1fff

Comment 3

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/9c9f970f1fff