1238786 - Make public tracing functions allow nullable pointers by default

Status: RESOLVED FIXED

(Core :: JavaScript: GC, defect)

Description

[Terrence Cole [:terrence]]

Attachment: 1_implicit_nullable_public_tracing-sm-v0.diff

When I moved everything over to JS::TraceEdge<T> and friends last week, I kept the same null-handling semantics as we currently expose. I'd like to change these to always null-check to simplify callers, but put it off to a later change in the off chance that one of these callers is actually hot on some benchmark we care about. I don't think that's too likely though, so let's make it happen.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f5188f30e186

Comment 1

[Terrence Cole [:terrence]]

Attachment: 2_implicit_nullable_public_tracing-gk-v0.diff

Comment 2

[Steve Fink [:sfink] [:s:]]

Comment on attachment 8706667 [details] [diff] [review]
1_implicit_nullable_public_tracing-sm-v0.diff

Review of attachment 8706667 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/public/TracingAPI.h
@@ +295,5 @@
>  // is used as a key in a hashtable, then the object must be removed and
>  // re-inserted with the correct hash.
> +//
> +// Note that while |edgep| must never be null, it fine to call this on nullable
> +// pointers without checking for nullptr first.

"it fine"

Though I wonder if it reads better as:

  Note that while |edgep| must never be null, it is fine for *edgep to be nullptr.

Comment 3

[Terrence Cole [:terrence]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/560e1ee498a5662d02806b21db11107359f95d13
Bug 1238786 - Part 1: Allow null pointers in public tracing APIs; r=sfink

https://hg.mozilla.org/integration/mozilla-inbound/rev/a9935197c5b73c8f8fe2e736587393379ec38ffc
Bug 1238786 - Part 2: We no longer need to null check before using trace APIs; r=smaug

Comment 4

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/560e1ee498a5
https://hg.mozilla.org/mozilla-central/rev/a9935197c5b7