vm: add Script.createCodeCache() #20300

devsnek · 2018-04-25T17:24:22Z

Closes #20052

/cc @nodejs/vm

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

addaleax · 2018-04-25T17:55:21Z

lib/vm.js

+    if (cachedData && !cachedDataWarned) {
+      process.emitWarning(cachedDataMessage, 'DeprecationWarning', 'DEPXXXX');
+      cachedDataWarned = true;
+    }


Note that this makes this PR semver-major – it might be easier to split the deprecation out?

Also – What’s the motivation for not doing a docs-only deprecation first? Do we actually want a run-time deprecation at any point? It seems like, for cache information, just making them no-ops would be less breakage for users…

i can remove the runtime dep

~~The additional of a new throw also makes it semver-major.~~

Wait, nevermind, this is a check on a new property. Scratch that :-)

hashseed · 2018-04-26T14:00:38Z

doc/api/deprecations.md

+
+Type: Documentation-only
+
+The options `produceCachedData` and `cachedData` have been deprecated. Please use


I would also support splitting the deprecation into a separate PR, so that we can ship this earlier without semver changes.

I changed it to docs only, not sure why I made it runtime in the first place

Even as a docs-only deprecation, it should be separated out into a separate commit.

Also ... just a nit... s/DEPXXXX/DEP00XX ... the release tooling looks specifically for the DEP00XX pattern to ensure that the codes are properly assigned before release.

src/node_contextify.cc

+          env,
+          reinterpret_cast<const char*>(cached_data->data),
+          cached_data->length);
+      args.GetReturnValue().Set(buf.ToLocalChecked());


hashseed · 2018-04-26T14:30:04Z

src/env.h

  V(change_string, "change")                                                  \
  V(channel_string, "channel")                                                \
  V(chunks_sent_since_last_write_string, "chunksSentSinceLastWrite")          \
+  V(code_cache_used_string, "codeCacheUsed")                                  \


This is an API change. Not sure whether that belongs into a separate PR? Not sure if this would block this PR from going into Node 10 otherwise.

TimothyGu · 2018-04-25T18:27:51Z

doc/api/vm.md

+  * `codeCache` {Buffer} Provides an optional `Buffer` with V8's code cache
+    data for the supplied source. When supplied, the `codeCacheRejected` value
+    will be set to either `true` or `false` depending on acceptance of the data
+    by V8.
  * `cachedData` {Buffer} Provides an optional `Buffer` with V8's code cache


Is the code cache generated by createCodeCache compatible with cachedData

Yes. The latter is produced with createCodeCache.

In that case I’m reluctant to deprecate an older name of the same thing. Actually would rather change createCodeCache to createCachedData or something similar.

joyeecheung · 2018-04-27T08:29:59Z

@hashseed Can the code cache generated by a different version of v8 be loaded by another? (My guess is no?)

hashseed · 2018-04-27T08:33:59Z

We check a version hash before deserializing. If the version hash mismatches, V8 rejects the cache data.

joyeecheung · 2018-04-27T08:39:28Z

@hashseed Thanks. It would be good to document about this (otherwise the user only has a boolean codeCacheRejected without further information or possible cause)

devsnek · 2018-04-27T22:15:33Z

I'm going to work on some patches this weekend for v8 to provide more data on what happened with the caches besides just if it was rejected, which should help make this API a bit better

hashseed · 2018-04-28T06:31:41Z

Generally agree. Though I think landing the minimal change of adding this new API without adding the reject reason or deprecating the old API would make sense too.

TimothyGu

Code looks good.

doc/api/vm.md

+-->
+
+Creates a code cache that can be used with the Script constructor's
+`cachedData` option. Returns a Buffer. If the code cache cannot be created then


doc/api/vm.md

+  - version: REPLACEME
+    pr-url: https://github.com/nodejs/node/pull/20300
+    description: The `produceCachedData` is deprecated in favour of
+                 `script.createCachedData()`


TimothyGu · 2018-04-28T21:28:51Z

doc/api/vm.md

+
+Creates a code cache that can be used with the Script constructor's
+`cachedData` option. Returns a Buffer. If the code cache cannot be created then
+`ERR_CODE_CACHE_CREATION_FAILED` is thrown. This method may be called at any


@hashseed What can possibly lead to failure of creating cached data? It would be helpful here for the documentation.

TimothyGu · 2018-04-28T21:29:49Z

doc/api/vm.md

+Creates a code cache that can be used with the Script constructor's
+`cachedData` option. Returns a Buffer. If the code cache cannot be created then
+`ERR_CODE_CACHE_CREATION_FAILED` is thrown. This method may be called at any
+time and any number of times.


Maybe also link to the V8 blog post?

TimothyGu · 2018-04-28T21:31:18Z

test/parallel/test-vm-cached-data.js


-      if (script.cachedDataProduced)
-        data = script.cachedData.toString('base64');


The old code should still be tested.

TimothyGu · 2018-04-28T21:31:51Z

doc/api/deprecations.md

+<a id="DEP00XX"></a>
+### DEP00XX: vm.Script cached data
+
+Type: Documentation-only


Any reason this isn't runtime?

@TimothyGu Some previous comments on that in #20300 (comment) … for a caching functionality, there is no reason to introduce a runtime deprecation, right?

i never know when to make something documentation or runtime... do we have some sort of guide on deciding which to do? i always kinda saw documentation as things that are too ingrained in npm packages such that like any project would be spammed if it used a runtime dep

@devsnek I don’t know if you’ve read it, but https://github.com/nodejs/node/blob/master/COLLABORATOR_GUIDE.md#deprecations might help you?

Generally, we should avoid moving directly from a documented feature towards a runtime deprecation. But yes, obviously popularity of a feature has played a role in deciding how fast we move with those steps…

@addaleax i've seen that but it doesn't really explain when to use them, just how they fit into semver. is the generally expected process that something will do docs -> runtime -> eol?

is the generally expected process that something will do docs -> runtime -> eol?

Yes, unless we have reason to deviate from it. But in this case I do think we have reason to do so. :)

hashseed · 2018-04-29T05:06:26Z

These here are possible reject reasons.

TimothyGu · 2018-04-29T20:31:07Z

@hashseed I meant possible reasons for failure of creating a code cache, not consuming one :)

hashseed · 2018-04-30T01:39:42Z

Oh. Creating may fail if the script is actually an asm.js module or if the debugger is active. The latter can probably be fixed as soon as the debug context has been removed completely.

jdalton · 2018-05-28T17:42:57Z

Any updates here? What's the status?

devsnek · 2018-05-28T17:46:05Z

@jdalton hashseed is going to backport a bunch of changes to SharedFunctionInfo soonish, this is just waiting for that.

hashseed · 2018-05-29T07:17:33Z

Umm.. wait. Which changes am I supposed to backport? I am not aware of anything actionable on my end. Did I drop the ball?

devsnek · 2018-05-29T11:26:37Z

@hashseed a couple weeks ago I emailed you about backporting the changes for this and modules sfi and my understanding from that was you would be backporting the changes. if that's wrong sorry about the misunderstanding 😅

hashseed · 2018-05-29T12:03:20Z

I haven't found such an email unfortunately. Do you mean by any chance this change? I'm pessimistic that it will get green light for back porting since this is not a bug fix. We could just float this in Node?

jdalton · 2018-06-01T19:19:15Z

@devsnek Are you cool floating the change into Node?

jdalton · 2018-06-19T20:07:19Z

Okay!

jdalton · 2018-06-21T14:32:26Z

CI https://ci.nodejs.org/job/node-test-pull-request/15551/

TimothyGu · 2018-06-21T16:08:21Z

test/parallel/test-vm-cached-data.js

+
+      assert(cachedData instanceof Buffer);
+
+      data = cachedData.toString('base64');


This overwrites data created earlier. Is it possible to test the buffer returned from both produceCachedData and createCachedData?

jdalton · 2018-06-22T05:10:39Z

CI https://ci.nodejs.org/job/node-test-pull-request/15562/

addaleax · 2018-06-22T06:18:15Z

CI again just to make sure: https://ci.nodejs.org/job/node-test-commit/19379/

jdalton · 2018-06-22T17:42:42Z

While experimenting with createCachedData() locally I noticed the produced buffers are roughly the same size as the old cachedData results even when calling createCachedData() after running the scripts (~800 modules). Does this sound likely or could this indicate something is slightly off?

Update:

NM, on double checking my test I found I goofed it.
The new API does produce more info in the buffers 😋

devsnek · 2018-06-26T21:38:33Z

landed in 4f67c6f

PR-URL: #20300 Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: James M Snell <[email protected]> Reviewed-By: John-David Dalton <[email protected]> Reviewed-By: Tiancheng "Timothy" Gu <[email protected]> Reviewed-By: Yang Guo <[email protected]>

Notable changes: * build: * Node.js should now be about 60% faster to startup than the previous version, thanks to the use V8's code cache feature for core modules. [#21405](#21405) * dns: * An experimental promisified version of the dns module is now available. Give it a try with `require('dns').promises`. [#21264](#21264) * fs: * `fs.lchown` has been undeprecated now that libuv supports it. [#21498](#21498) * lib: * `Atomics.wake` is being renamed to `Atomics.notify` in the ECMAScript specification ([reference](tc39/ecma262#1220)). Since Node.js now has experimental support for worker threads, we are being proactive and added a `notify` alias, while emitting a warning if `wake` is used. [#21413](#21413) [#21518](#21518) * n-api: * Add API for asynchronous functions. [#17887](#17887) * util: * `util.inspect` is now able to return a result instead of throwing when the maximum call stack size is exceeded during inspection. [#20725](#20725) * vm: * Add `script.createCachedData()`. This API replaces the `produceCachedData` option of the `Script` constructor that is now deprecated. [#20300](#20300) * worker: * Support for relative paths has been added to the `Worker` constructor. Paths are interpreted relative to the current working directory. [#21407](#21407) PR-URL: #21629

Notable changes: * dns: * An experimental promisified version of the dns module is now available. Give it a try with `require('dns').promises`. [#21264](#21264) * fs: * `fs.lchown` has been undeprecated now that libuv supports it. [#21498](#21498) * lib: * `Atomics.wake` is being renamed to `Atomics.notify` in the ECMAScript specification ([reference](tc39/ecma262#1220)). Since Node.js now has experimental support for worker threads, we are being proactive and added a `notify` alias, while emitting a warning if `wake` is used. [#21413](#21413) [#21518](#21518) * n-api: * Add API for asynchronous functions. [#17887](#17887) * util: * `util.inspect` is now able to return a result instead of throwing when the maximum call stack size is exceeded during inspection. [#20725](#20725) * vm: * Add `script.createCachedData()`. This API replaces the `produceCachedData` option of the `Script` constructor that is now deprecated. [#20300](#20300) * worker: * Support for relative paths has been added to the `Worker` constructor. Paths are interpreted relative to the current working directory. [#21407](#21407) PR-URL: #21629

devsnek requested a review from hashseed April 25, 2018 17:24

nodejs-github-bot added the lib / src Issues and PRs related to general changes in the lib or src directory. label Apr 25, 2018

devsnek added vm Issues and PRs related to the vm subsystem. and removed lib / src Issues and PRs related to general changes in the lib or src directory. labels Apr 25, 2018

addaleax reviewed Apr 25, 2018

View reviewed changes

addaleax added the semver-major PRs that contain breaking changes and should be released in the next major version. label Apr 25, 2018

devsnek added semver-minor PRs that contain new features and should be released in the next minor version. and removed semver-major PRs that contain breaking changes and should be released in the next major version. labels Apr 25, 2018

devsnek force-pushed the feature/vm-snapshot-on-the-go-anytime-anywhere branch from 0a65658 to d4aa733 Compare April 25, 2018 18:06

hashseed reviewed Apr 26, 2018

View reviewed changes

jdalton mentioned this pull request Apr 26, 2018

vm module should expose V8::ScriptCompiler::CreateCodeCache #20052

Closed

hashseed reviewed Apr 26, 2018

View reviewed changes

TimothyGu reviewed Apr 26, 2018

View reviewed changes

devsnek force-pushed the feature/vm-snapshot-on-the-go-anytime-anywhere branch 2 times, most recently from f066d12 to b45e969 Compare April 28, 2018 14:53

TimothyGu reviewed Apr 28, 2018

View reviewed changes

devsnek force-pushed the feature/vm-snapshot-on-the-go-anytime-anywhere branch from ab4284b to c31b520 Compare June 19, 2018 17:59

jasnell approved these changes Jun 21, 2018

View reviewed changes

jdalton approved these changes Jun 21, 2018

View reviewed changes

TimothyGu reviewed Jun 21, 2018

View reviewed changes

This comment has been minimized.

Sign in to view

vm: add Script.createCodeCache()

9db4f30

devsnek force-pushed the feature/vm-snapshot-on-the-go-anytime-anywhere branch from c31b520 to 9db4f30 Compare June 21, 2018 19:16

TimothyGu approved these changes Jun 22, 2018

View reviewed changes

addaleax approved these changes Jun 22, 2018

View reviewed changes

addaleax added the c++ Issues and PRs that require attention from people who are familiar with C++. label Jun 22, 2018

addaleax added the author ready PRs that have at least one approval, no pending requests for changes, and a CI started. label Jun 22, 2018

hashseed approved these changes Jun 22, 2018

View reviewed changes

hashseed mentioned this pull request Jun 22, 2018

wip ryzokuken/node#1

Closed

devsnek closed this Jun 26, 2018

devsnek deleted the feature/vm-snapshot-on-the-go-anytime-anywhere branch June 26, 2018 21:38

targos mentioned this pull request Jul 3, 2018

v10.6.0 proposal #21629

Merged

devsnek removed the author ready PRs that have at least one approval, no pending requests for changes, and a CI started. label Jul 18, 2018

rauchg mentioned this pull request Nov 25, 2018

V8 Compile Cache vercel/ncc#51

Closed

2 tasks


		Type: Documentation-only

		The options `produceCachedData` and `cachedData` have been deprecated. Please use


		if (script.cachedDataProduced)
		data = script.cachedData.toString('base64');


		assert(cachedData instanceof Buffer);

		data = cachedData.toString('base64');

vm: add Script.createCodeCache() #20300

vm: add Script.createCodeCache() #20300

Conversation

devsnek commented Apr 25, 2018 • edited Loading

Checklist

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jasnell Apr 27, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

This comment was marked as resolved.

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

joyeecheung commented Apr 27, 2018

hashseed commented Apr 27, 2018

joyeecheung commented Apr 27, 2018

devsnek commented Apr 27, 2018

hashseed commented Apr 28, 2018

TimothyGu left a comment

Choose a reason for hiding this comment

This comment was marked as resolved.

This comment was marked as resolved.

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hashseed commented Apr 29, 2018 • edited Loading

TimothyGu commented Apr 29, 2018

hashseed commented Apr 30, 2018

jdalton commented May 28, 2018

devsnek commented May 28, 2018

hashseed commented May 29, 2018

devsnek commented May 29, 2018 • edited Loading

hashseed commented May 29, 2018

jdalton commented Jun 1, 2018

jdalton commented Jun 19, 2018

jdalton commented Jun 21, 2018

Choose a reason for hiding this comment

This comment has been minimized.

jdalton commented Jun 22, 2018

addaleax commented Jun 22, 2018

jdalton commented Jun 22, 2018 • edited Loading

devsnek commented Jun 26, 2018

devsnek commented Apr 25, 2018 •

edited

Loading

jasnell Apr 27, 2018 •

edited

Loading

hashseed commented Apr 29, 2018 •

edited

Loading

devsnek commented May 29, 2018 •

edited

Loading

jdalton commented Jun 22, 2018 •

edited

Loading