[8.x] n-api: backport changes from master

[gabrielschulhof]

The following commits (or relevant portions thereof) are included:

* c123467 timers: allow Immediates to be unrefed (the portion fixing
  the uv_test_loop test)
* a555397 n-api: add methods to open/close callback scope
* 47f664e n-api: wrap control flow macro in do/while
* 1286923 n-api: implement wrapping using private properties
* 316b5ef n-api: throw RangeError napi_create_typedarray()
* 8938c4c n-api: expose n-api version in process.versions
* 91c1ccd n-api: throw RangeError in napi_create_dataview() with
  invalid range
* 94e2951 n-api: fix memory leak in napi_async_destroy()
* dc389bf n-api: use nullptr instead of NULL in node_api.cc
* a4a9a3d src: add napi_handle_scope_mismatch to msg list

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

[gabrielschulhof]

Fixes nodejs/abi-stable-node#297.

[gibfahn]

CI: https://ci.nodejs.org/job/node-test-pull-request/13775/

[gabrielschulhof]

This also fails on an unrelated test: parallel/test-benchmark-misc.

[MylesBorins]

@gabrielschulhof we have generally backported every commit individually... that makes it easier to audit in the future and stop tooling from breaking

@nodejs/lts thoughts on landing a single commit comprising of many patches?

[gibfahn]

@nodejs/lts thoughts on landing a single commit comprising of many patches?

I'd rather backport each commit individually, although I'm +1 on it being in one PR (i.e. this one). @gabrielschulhof if you could split this into the individual commits that would be good. You don't need to change any commit metadata, so it should just be a case of cherry-picking and fixing merge conflicts.

[gabrielschulhof]

Can do.

…

On Thu, Mar 22, 2018 at 11:45 AM, Gibson Fahnestock < ***@***.***> wrote: @nodejs/lts <https://github.com/orgs/nodejs/teams/lts> thoughts on landing a single commit comprising of many patches? I'd rather backport each commit individually, although I'm +1 on it being in one PR (i.e. this one). @gabrielschulhof <https://github.com/gabrielschulhof> if you could split this into the individual commits that would be good. You don't need to change any commit metadata, so it should just be a case of cherry-picking and fixing merge conflicts. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#19265 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA7k0UYxc8Ju-XtmJ8yYplmjJ0YArQK-ks5tg8c0gaJpZM4Sksat> .

[gabrielschulhof]

@gibfahn there are some commits where I've intentionally not applied parts of it even though they would've applied cleanly, because those parts were not related to N-API.

[gabrielschulhof]

@gibfahn @MylesBorins I have now replaced the single commit with the sometimes-partially cherry-picked individual commits.

[gabrielschulhof]

CI: https://ci.nodejs.org/job/node-test-pull-request/13825/

[gabrielschulhof]

The same unrelated failure has reoccurred.

[gibfahn]

@gibfahn there are some commits where I've intentionally not applied parts of it even though they would've applied cleanly, because those parts were not related to N-API.

So if you think those changes should be applied to 8.x then you should leave them in, we try to backport whole commits where possible, otherwise understanding what's been backported becomes even harder.

If they don't need to go back to 8.x at all, then leaving them out is fine.

Sorry to keep complaining about this PR, I really appreciate you doing this work!

[gabrielschulhof]

@gibfahn NP. We have to do things right.

I'll do the cherry-picking again, being more inclusive. Any decisions I make I will document in a subsequent comment here.

[gabrielschulhof]

@MylesBorins @gibfahn the list of commits from master:

⁠	⁠	⁠	⁠	⁠	⁠
6572f63	d4b3fcb	c6c957d	be47094	c1695d8	f24d0ec
6a5a9ad	d20f1f0	5fb6f7f	cd7d7b1	a03c90b	33e63fe
48b5c11	35c7238	d83f830	449d1c8	c698017	a29089d
9cb96ac	6e1c25c	463d1a4	caee112	755e07c	a1bab82
2bead4b	1729af2	fe442f6	d3569b6	5ecd2e1	6101bd2
d379db3	a555397	c2b9048	629a447	47f664e	1286923
0c16b18	6e312c5	6c1906a	c123467	f054855	c90185c
316b5ef	f75bc2c	3e06276	8938c4c	91c1ccd	8a86d9c
93acfe5	094d92b	94e2951	246aeac	ef49f55

These are the places where I made a judgement call:

• c123467 timers: allow Immediates to be unrefed
  I did not backport the immediate.ref() and immediate.unref(), but only the modification to test/addons-napi/test_uv_loop/test_uv_loop.cc
• d3569b6 doc: remove **Note:** tags
  This touched N-API, so I grabbed the whole thing, dropping chunks which do not apply, and manually applying the fix in places where there are such tags that were not on master
• caee112 test: remove assert.doesNotThrow()
  I removed the files that were "deleted by us", checked out "our" version of those that conflicted, and removed assert.doesNotThrow() by hand
• 463d1a4 test,benchmark,doc: enable dot-notation rule
  I removed the files that were "deleted by us", checked out "our" version of those that conflicted, and moved to dot-notation by hand. I ran the new linter to make sure everything was covered.
• 6e1c25c errors: update all internal errors
  I removed the unused errors but I did not backport the updated definition of E() which accepts an additional parameter.
• 9cb96ac doc: remove extraneous "for example" text
  Did my best to merge this, since its documentation-only.
• a29089d doc: add new documentation lint rule
  I applied as much of the commit as possible, checking out "our" version of conflicting files, and manually wrapping lines at 80 characters to conform to the linting rule.
• 35c7238 doc: replace to Node.js
  I applied as much of the commit as possible, checking out "our" version of conflicting files, and using git grep -niH 'node' -- doc/api doc/guides tools/icu/README.md | grep -v 'Node[.]js' | grep -i --color=always node | less -R to find places that should read "Node.js" instead of "node". I found one place where it's better to replace "node" with "`node`" rather than "Node.js", because the text refers to the name of the binary.

[gabrielschulhof]

I'll close and open another because this needs to go on top of v8.x-staging.

[addaleax]

@gabrielschulhof Not sure if that’s what you mean by your last comment, but using the same button that you’d use to change a PR title, you can also change its base branch :)

[MylesBorins]

landed in 3225601...07a6770

edit: automated adding the backport-pr-url labels with the following command

gitfilter-branch --msg-filter 'perl -pe "s/PR-URL/Backport-PR-URL: https://github.com//pull/19265\nPR-URL/g"' upstream/v8.x-staging..v8.x-staging

[MylesBorins]

It seems like e15f577 was only a partial backport of a test from a semver-minor commit... this will make it impossible for us to audit that semver-minor commit in the future if we wanted to land it.

Were there any other commits / prs that were partially backported?

[gabrielschulhof]

@MylesBorins 2838f9b doesn't modify test/parallel/test-console-assign-undefined.js whereas the original does.

I used

git log --oneline 3225601...07a6770 | \
while read id message; do \
  echo $id; \
  diff -u \
    <(git show --stat=99 $id | grep -vE '^commit|files changed' | grep '|') \
    <(git log --oneline master | grep "$message" | awk '{print $1;}' | xargs \
        git show --stat=99 | grep -vE '^commit|files changed' | grep '|'); \
done

to check whether the files touched by the backported commits differ from the files touched by the commits on master.

[MylesBorins]

@gabrielf did that script identify e15f577 as well? If so this satisfied me!

Perhaps we should consider putting this into node-core-utils to help us with reviewing backports /cc @nodejs/automation

[gabrielf]

@MylesBorins you are asking the wrong Gabriel...

[gabrielschulhof]

@gabrielf sorry!

@MylesBorins it did. This was the output:

07a6770614
8b3ef4660a
92f699e021
367113f5d7
f0ba2c6ceb
24b8bb6708
3a6b7e610d
9949d55ae9
f29d8e0e8d
7c6fa183cb
584fadc605
4c1181dc02
faf94b1c49
df63adf7aa
b26410e86f
1abb168838
cb3f90a1a9
6129ff4b99
87d0fd8212
58688d97dc
2838f9b150
--- /dev/fd/63	2018-05-03 15:57:59.614213570 -0400
+++ /dev/fd/62	2018-05-03 15:57:59.614213570 -0400
@@ -1 +1,2 @@
- test/addons-napi/test_typedarray/test.js | 4 ++--
+ test/addons-napi/test_typedarray/test.js       | 4 ++--
+ test/parallel/test-console-assign-undefined.js | 3 +--
5c0983e5a2
4d43607474
--- /dev/fd/63	2018-05-03 15:58:00.621202048 -0400
+++ /dev/fd/62	2018-05-03 15:58:00.621202048 -0400
@@ -1 +1,2 @@
  doc/api/n-api.md | 2 +-
+ doc/api/n-api.md | 2 +-
9244e1d234
927fc0b19f
9729278007
7ed1dfef28
969a520990
d89f5937eb
bebcdfe382
af655f586c
5b1b74c5a5
e15f57745d
--- /dev/fd/63	2018-05-03 15:58:05.856142148 -0400
+++ /dev/fd/62	2018-05-03 15:58:05.856142148 -0400
@@ -1 +1,10 @@
- test/addons-napi/test_uv_loop/test_uv_loop.cc | 9 +++++++++
+ doc/api/timers.md                                   |  32 +++++++++
+ lib/timers.js                                       | 143 +++++++++++++++++++++++----------------
+ src/env-inl.h                                       |  40 +++++++++--
+ src/env.cc                                          |  36 +++++-----
+ src/env.h                                           |  17 ++++-
+ src/node_perf.cc                                    |  18 ++---
+ src/timer_wrap.cc                                   |  12 ++--
+ test/addons-napi/test_uv_loop/test_uv_loop.cc       |   9 +++
+ test/parallel/test-timers-immediate-unref-simple.js |   7 ++
+ test/parallel/test-timers-immediate-unref.js        |  37 ++++++++++
84e0a03727
8cfa87832d
ca10fda064
51513fdf9e
02ae3295d5
853b4d593c
48be8a4793
79ecc2c586
ad8c079af7
a744535f99
2e100c82be
077e1870ae

I didn't mention 4d43607 because there are two commits with the same message on master.