deps: Fix V8 build in 32bit SmartOS with GCC 4.9

[joaocgreis]

Currently in 32bit SmartOS, node compiles and runs well with GCC 4.7 but segfaults immediately when compiled with GCC 4.9.

This fix is a cherry pick of 3 commits from V8:

1. https://chromium.googlesource.com/external/v8.git/+/4e670fd05e439145b3f46dd2c54d332044a72954
   (only the variable rename in NonAsciiStart)
2. https://chromium.googlesource.com/external/v8.git/+/90dc5c9e66ef1778cb8425c43c2f37b6799260c3
3. https://chromium.googlesource.com/external/v8.git/+/7cb82a76b46ef18f8e05d40d4fe13b3410d446e4

V8 issue: https://code.google.com/p/chromium/issues/detail?id=412967

These 3 commits are separate in this PR for easier review, but will be squashed if accepted.

Fixes #25281

[joaocgreis]

@joyent/node-collaborators PTAL, thanks!

[misterdjules]

Adding as a P-1 for the same reasons as highlighted previously in the issue it fixes.

[misterdjules]

@joaocgreis I would suggest changing the commit messages to make them clearer.

More specifically, I would follow the guidelines about cherry-picking changes from V8 and include the original commit message in each commit, like in the following commit: 2fc5eeb. Please note that the PR-URL and Reviewed-By fields are added by the Jenkins job, so you don't have to add them to the commit message.

I like how you mentioned the other commits' sha though in each of the three commits, so I would keep that but I would make it more concise.

Other than that the changes look good, but before approving these changes I would like to know from @alhazred if these changes fix his problems (discussion is ongoing in github.com//issues/25281).

Anyway, great work @joaocgreis 👍 and thank you!

[misterdjules]

@joaocgreis #25281 (comment) confirms that your changes fix the problem for the original reporter, so once the commit messages are fixed, LGTM.

Also, I would not squash these commits, as it's easier to identify them when we have to investigate issues if they're as close to the original changes as possible.

If you need help landing these changes, please ping me on #libuv :)

[misterdjules]

@joaocgreis When this lands, you'll also need to update the list of V8 floating patches for v0.12.

[joaocgreis]

@misterdjules The first commit is only a variable rename needed only for the third commit to merge cleanly. It's huge, 2,221 additions and 2,548 deletions, and does not merge cleanly. Should I:

1. Pick it partially as I did above
2. Rename the variable with a commit of my own
3. Not pick it and solve the conflicts when picking commit 3

Which one do you think is the best option, considering this needs to be added as a floating patch?

[misterdjules]

@joaocgreis I would choose 3) because solving the conflicts doesn't change the commit enough such as it becomes too different from the original one, and it avoids us from floating one more patch.

[joaocgreis]

Updated. Let me know if the commit messages are what you had in mind, @misterdjules . Ok to land?

[misterdjules]

@joaocgreis I would change the title of each commit to really describe what they do, so the first one would be something like V8: remove V8_HOST_CAN_READ_UNALIGNED, and the second would be V8: do not use wide reads in CopyCharsUnsigned. Other than that LGTM, thank you very much! 👍

[joaocgreis]

Landed in 48b0ca2 and 13ea50e

Also added the commits to the list of floating patches.

[misterdjules]

@joaocgreis 👍 Thank you!