[v8.x] deps: update V8 to 6.2

[targos]

/cc @nodejs/v8

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

Affected core subsystem(s)

V8

[targos]

@addaleax

Fwiw, for reasonable API/ABI compat to 6.1 I think all we need is:

• addaleax/node@338b581 – Move new PARSER enum value to end of enum
• addaleax/node@82807a8 – Revert removal of deprecated method
• Fix up the internal values of JS_API_OBJECT_TYPE to match what’s declared in v8.h. @targos I think you had a better solution than me the last time we needed to do that, do you remember what you did for it?

Thanks. I cherry-picked your commits. For the internal values, see cb3dfe9. Usually it's because a new value is added to the enum before the ones exposed in v8.h. I just moved it to a lower place.

I wonder if we also need to revert the change to the StateTag enum (or reorder it too?).

[addaleax]

I wonder if we also need to revert the change to the StateTag enum (or reorder it too?).

Doesn’t addaleax@338b581 catch that? It’s reordering rather than reverting but I think that should be fine for defensively coded applications (if there are any)

[targos]

@addaleax ah sorry I didn't see it was about this enum. The current diff looks like this though:

It looks like BYTECODE_COMPILER should also be moved to the end?

[addaleax]

@targos Hm, right – guess I didn’t catch that? Do you want to update the commit?

[targos]

Yes, I can update the commit tomorrow.

I will also have to revert the move of JS_MODULE_NAMESPACE_TYPE from v8/v8@ddec1c4

[targos]

Updated. Do we care about tools/v8heapconst.py ? In case we do, I don't know how to regenerate that file.

Edit: well, the heap-profiler addon test is broken now, so I guess we do care.

=== release test ===                                                           
Path: addons/heap-profiler/test
Command: out/Release/node /home/mzasso/git/nodejs/node-v8/test/addons/heap-profiler/test.js
--- CRASHED (Signal: 11) ---
[02:25|% 100|+ 2006|-   1]: Done                                               
make: *** [Makefile:213: test] Error 1

[hashseed]

I think it was through make postmortem-data or something...

[bnoordhuis]

Some things I noticed:

1. Missing HeapProfiler::GetProfilerMemorySize() - could be reimplemented in api.cc as a method returning 0.

2. Replacing TransferrableModule move constructor and assignment operator with default implementations. The operator lived in api.cc so that's certainly an ABI change; the constructor was inline but it probably still qualifies as an ABI change.

3. New default arguments in Context::SetDefaultContext() and Context::New().

Apropos v8heapconst: AFAIK, that's created by grokdump, which we don't use.

[addaleax]

@bnoordhuis re: TransferrableModule … for the previous upgrades in 8.x I think we pretty much agreed to leave the ES Modules/WASM stuff aside.

[psmarshall]

Yep, v8heapconst.py can be generated with (from v8 directory):

out/x64.release/mkgrokdump > tools/v8heapconst.py

[gibfahn]

Landed in 072902a...2a2c881