Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(327)

Issue 98973002: OilPan related infrastructure on the V8 side. (Closed)

Created:
7 years ago by Vyacheslav Egorov (Chromium)
Modified:
6 years, 10 months ago
CC:
v8-dev, Paweł Hajdan Jr.
Visibility:
Public.

Description

OilPan related infrastructure on the V8 side. Provides two API extensions: - Access to internal interruption mechanism to implement OilPan safepoint inside long running JavaScript code (v8::Isolate::RequestInterrupt, v8::Isolate::ClearInterrupt); - GC callback flag to distinguish forced GC from regular GC for testing purposes. NOT FOR COMMIT BUG=

Patch Set 1 #

Patch Set 2 : Add api comments. #

Unified diffs Side-by-side diffs Delta from patch set Stats (+135 lines, -27 lines) Patch
M include/v8.h View 1 2 chunks +20 lines, -1 line 0 comments Download
M src/api.cc View 1 chunk +11 lines, -0 lines 0 comments Download
M src/execution.h View 3 chunks +10 lines, -1 line 0 comments Download
M src/execution.cc View 1 4 chunks +49 lines, -0 lines 0 comments Download
M src/extensions/gc-extension.cc View 1 chunk +4 lines, -2 lines 0 comments Download
M src/heap.h View 4 chunks +19 lines, -10 lines 0 comments Download
M src/heap.cc View 6 chunks +16 lines, -10 lines 0 comments Download
M src/heap-inl.h View 1 chunk +5 lines, -2 lines 0 comments Download
M src/heap-snapshot-generator.cc View 1 chunk +1 line, -1 line 0 comments Download

Messages

Total messages: 3 (0 generated)
Vyacheslav Egorov (Chromium)
PTAL. This is not a for commit. This a rebase of the patch that we ...
7 years ago (2013-12-02 12:35:20 UTC) #1
Michael Starzinger
I am fine with the stack-check based interrupt handler. If the stack-check is good enough ...
7 years ago (2013-12-02 17:29:55 UTC) #2
Mads Ager (chromium)
7 years ago (2013-12-02 18:27:18 UTC) #3
On 2013/12/02 17:29:55, Michael Starzinger wrote:
> I am fine with the stack-check based interrupt handler. If the stack-check is
> good enough for the OilPan thread suspension at safepoints then that's cool.
> 
> As for the pushing the additional kGCCallbackForced flag through to the
> callback, I would like to understand the underlying use-case a little bit
> better, but can certainly be convinced. Let's discuss that tomorrow.

Yes, let's talk about it more tomorrow. To have it recorded here as well, this
is only for testing. The issue is that WebKit layout tests expect to be able to
force a V8 GC and then test that certain DOM nodes are dead. However, we are not
coupling V8 and Blink GC tightly so forcing a V8 GC does not necessarily lead to
a Blink GC being forced and therefore the DOM nodes will not necessarily die at
the expected time. We are using this flag in the Blink V8GCController to force a
Blink GC after V8 GCs that were forced for testing purposes through the GC
extension. It is nice to leave these test in there in order to be able to check
that things actually are unreachable when they should be.

I'm also very open to other suggestions if you guys think there is a cleaner way
to do this.

Powered by Google App Engine
This is Rietveld 408576698