Make V8 5.5 ABI compatible with 5.4 #2
Comments
|
The ABI patch for 5.0/5.1 fixed all incompatibilities in the |
|
@nodejs/ctc we need to have further discussion about this because of async/await. @ofrobots raised this at the last meeting - V8 doesn't yet have hooks for proper Promise inspection with AsyncHooks et. al. but exposing async/await is going to drive much heavier adoption of native Promises. Is this a situation we are comfortable with in the short term while we wait for deeper hooks or do we want to postpone until we have the tools we need to provide proper runtime inspection? FYI if you didn't realise already, 5.5 is the one that unflags async/await, it's really the only major JS feature. |
IndexedPropertyHandlerConfiguration and NamedPropertyHandlerConfiguration picked up extra fields. EmbedderHeapTracer has new virtual methods. I would expect that to change the vtable of derived classes but abi-compliance-checker doesn't seem to think it's an issue. |
What would be the way to handle it? Re async/await: We could float a patch to keep it behind a flag until we decide it's ready for production. |
|
You could try reverting it to the old API but in the specific case of EmbedderHeapTracer I'm not sure if it's worth the hassle. It's only useful to embedders (i.e., node.js, but we don't use it), it's not something add-ons would or should use. If a tree falls and no one is around to hear it... |
|
Thanks. I'll move the discussion to a PR where the changes can be more easily seen |
Revision: 8f22fce BUG=chromium:662674 LOG=N NOTRY=true NOPRESUBMIT=true NOTREECHECKS=true R=yangguo@chromium.org Review URL: https://codereview.chromium.org/2490023004 . Cr-Commit-Position: refs/branch-heads/5.4@{v8#81} Cr-Branched-From: 5ce2827-refs/heads/5.4.500@{#2} Cr-Branched-From: ad07b49-refs/heads/master@{#38841}
Merged: Add test for making private symbols non-enumerable Revision: 942604d Merged: Make private symbols non-enumerable Revision: 135b9f9 BUG=chromium:664411,chromium:664411 LOG=N NOTRY=true NOPRESUBMIT=true NOTREECHECKS=true R=cbruni@chromium.org Review URL: https://codereview.chromium.org/2497213003 . Cr-Commit-Position: refs/branch-heads/5.4@{v8#83} Cr-Branched-From: 5ce2827-refs/heads/5.4.500@{#2} Cr-Branched-From: ad07b49-refs/heads/master@{#38841}
#2 id:20001 of https://codereview.chromium.org/2480223002/ ) Reason for revert: http://build.chromium.org/p/client.v8/builders/V8%20Linux%20gcc%204.8/builds/9724 Original issue's description: > [debugger] Migrate more debugger tests to inspector > > This moves all tests currently working with the inspector debugger wrapper to > test/debugger. > > BUG=v8:5530 TBR=yangguo@chromium.org # Skipping CQ checks because original CL landed less than 1 days ago. NOPRESUBMIT=true NOTREECHECKS=true NOTRY=true BUG=v8:5530 Review-Url: https://codereview.chromium.org/2480283002 Cr-Commit-Position: refs/heads/master@{#40805}
…set #2 id:20001 of https://codereview.chromium.org/2495213003/ ) Reason for revert: Breaks https://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20vtunejit/builds/14690 compile Original issue's description: > [serializer] print use count of external references. > > R=peria@chromium.org, vogelheim@chromium.org > BUG=chromium:617892 > NOPRESUBMIT=true TBR=peria@chromium.org,vogelheim@chromium.org,yangguo@chromium.org # Skipping CQ checks because original CL landed less than 1 days ago. NOPRESUBMIT=true NOTREECHECKS=true NOTRY=true BUG=chromium:617892 Review-Url: https://codereview.chromium.org/2498163003 Cr-Commit-Position: refs/heads/master@{#40980}
It seems that there are not a lot of things to do to achieve this goal.
I compared the headers of 5.5 and 5.4 with abi-compliance-checker and the report can be viewed at https://direct.lactame.com/files/nodejs/v8/compat_report.html.
I already fixed the
kJSApiObjectTypeandkJSObjectTypevalues (see this commit) and now I wonder if anything else is needed.Also, do we only care about the changes in
v8.h?/cc @nodejs/v8
The text was updated successfully, but these errors were encountered: