deps: backport 819b40a from V8 upstream

[targos]

Original commit message:

Use baseline code to compute message locations.
This switches Isolate::ComputeLocation to use baseline code when
computing message locations. This unifies locations between optimized
and non-optimized code by always going through the FrameSummary for
location computation.
R=bmeurer@chromium.org
TEST=message/regress/regress-4266
BUG=v8:4266
LOG=n
Review URL: https://codereview.chromium.org/1331603002
Cr-Commit-Position: refs/heads/master@{#30635}

Fixes: #3934

cc @nodejs/v8

[cjihrig]

Rubber stamp LGTM

[indutny]

LGTM

[targos]

CI: https://ci.nodejs.org/job/node-test-pull-request/812/

Edit: retrying... https://ci.nodejs.org/job/node-test-pull-request/814/

[bnoordhuis]

LGTM although I'd be happier if we could run the V8 test suite on this.

[bnoordhuis]

@targos Have you tried getting this accepted in the upstream 4.6 branch?

[targos]

I'd be happier if we could run the V8 test suite on this.

I'll try to run it locally

Have you tried getting this accepted in the upstream 4.6 branch?

I haven't. Should I follow this guide to do it ? https://github.com/v8/v8/wiki/Merging%20&%20Patching

[bnoordhuis]

Yes, that's the one.

[ofrobots]

I asked the V8 team already about this; it is unlikely to be merge-accepted into the 4.6 branch at this point. (But don't let me discourage you from requesting a merge.)

[ofrobots]

For more transparency:

4.6 is a tricky beast because we have no longer test coverage for it. We branched a week ago and all our testers have switched to 4.8 and 4.7.

[targos]

I have one failing test on the 4.6 branch with this change:

/home/mzasso/git/chromium/v8/out/x64.release/cctest --random-seed=-1779142066 --turbo --always-opt test-strings/RobustSubStringStub --nohard-abort --nodead-code-elimination --nofold-constants --testing_serialization_file=/home/mzasso/git/chromium/v8/out/.serdes/serdes_RobustSubStringStub__turbo__always_opt
#
# Fatal error in , line 0
# Unexpected translation type
#
==== C stack trace ===============================
1: 0x6e2583
2: 0x7f939a
3: 0x82d4f7
4: 0x934f89
5: 0x934854
6: 0xb77471
7: 0xedd3510839b
[1] 9019 abort (core dumped) /home/mzasso/git/chromium/v8/out/x64.release/cctest --random-seed=-1779142066

[bnoordhuis]

Do you mean the test passes without the change from this PR? Apparently it's corrupting optimized stack frames, going by the error message. You may get a more helpful error with a make x64.debug extrachecks=on slowdchecks=on build.

[targos]

You may get a more helpful error with a make x64.debug extrachecks=on slowdchecks=on build.

That helped a lot, thanks.

I just pushed an update. Sorry I missed this when resolving the cherry-pick conflict. Since it's a turbofan part, our test suite couldn't catch it.

now make x64.release.check passes. Do I need to test something else ?

[targos]

CI: https://ci.nodejs.org/job/node-test-pull-request/825/

EDIT: run on ARM: https://ci.nodejs.org/job/node-test-commit-arm-fanned/629/

[bnoordhuis]

LGTM. The missing bit was the case Runtime::kInlineSubString: in linkage.cc, wasn't it?

now make x64.release.check passes. Do I need to test something else ?

If you're patient, you may want to check make x64.debug.check as well.

[targos]

The missing bit was the case Runtime::kInlineSubString: in linkage.cc, wasn't it?

Yep

[targos]

>>> Running tests for x64.debug
No connection to distribution server; running tests locally.
[80:01|% 100|+ 23568|- 0]: Done 

[targos]

Thanks for the help Ben.
Landed on master with ab25589

[rvagg]

After a quick check with v4.x it seems that V8 4.5 did not have this but, tagging as dont-land-on-v4.x but I'm open to correction because perhaps I'm running #3934 wrong.

[targos]

@rvagg I did another PR for V8 4.5 (#3938). So you're right to tag this as dont-land-on-v4.x