1167795 - Remove DumpHeapVisitChild

Status: RESOLVED FIXED

(Core :: JavaScript: GC, defect)

Description

[Terrence Cole [:terrence]]

Attachment: remove_DumpHeapVisitChild-v0.diff

The only difference from DumpHeapVisitRoot is a prefix on the string. If we just pass an explicit prefix to print, we can skip having a second almost-identical function here and, more importantly, kill the last user of setTraceCallback.

We can't move to fully static dispatch because the tracing has to be out-of-line, but switching to virtual dispatch here would be nicer on the analysis and potentially faster in some cases.

Comment 1

[Andrew McCreight [:mccr8]]

Comment on attachment 8609663 [details] [diff] [review]
remove_DumpHeapVisitChild-v0.diff

Review of attachment 8609663 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/public/TracingAPI.h
@@ -133,5 @@
>          contextName_(nullptr), contextIndex_(InvalidIndex), contextFunctor_(nullptr)
>      {}
>  
> -    // Update the trace callback.
> -    void setTraceCallback(JSTraceCallback traceCallback) {

Nice!

::: js/src/jsfriendapi.cpp
@@ +892,1 @@
>      FILE*  output;

Maybe get rid of this weird extra space before output while you are here.

@@ +968,1 @@
>  DumpHeapVisitRoot(JS::CallbackTracer* trc, void** thingp, JS::TraceKind kind)

Please rename DumpHeapVisitRoot to something more generic.  DumpHeapVisitGCThing or something?

Comment 2

[Terrence Cole [:terrence]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=483b0d73894e

Comment 3

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/29b70379883a

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/29b70379883a