v8.10.0 proposal by gibfahn · Pull Request #18336 · nodejs/node
Commit d217b28 ("async_hooks: add trace events to async_hooks") used `NODE_MODULE_CONTEXT_AWARE_BUILTIN()` instead. After commit 8680bb9 ("src: explicitly register built-in modules") it no longer works for static library builds so remove it. Backport-PR-URL: #18179 PR-URL: #17071 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Timothy Gu <timothygu99@gmail.com>
This removes the `process._needImmediateCallback` property
and its semantics of having a 1/0 switch that tells C++ whether
immediates are currently scheduled.
Instead, a counter keeping track of all immediates is created,
that can be increased on `setImmediate()` or decreased when an
immediate is run or cleared.
This is faster, because rather than reading/writing a C++ getter,
this operation can be performed as a direct memory read/write via
a typed array. The only C++ call that is left to make is
activating the native handles upon creation of the first
`Immediate` after the queue is empty.
One other (good!) side-effect is that `immediate._destroyed` now
reliably tells whether an `immediate` is still scheduled to run or not.
Also, as a nice extra, this should make it easier to implement
an internal variant of `setImmediate` for C++ that piggybacks
off the same mechanism, which should be useful at least for
async hooks and HTTP/2.
Benchmark results:
$ ./node benchmark/compare.js --new ./node --old ./node-master-1b093cb93df0 --runs 10 --filter immediate timers | Rscript benchmark/compare.R
[00:08:53|% 100| 4/4 files | 20/20 runs | 1/1 configs]: Done
improvement confidence p.value
timers/immediate.js type="breadth" thousands=2000 25.61 % ** 1.432301e-03
timers/immediate.js type="breadth1" thousands=2000 7.66 % 1.320233e-01
timers/immediate.js type="breadth4" thousands=2000 4.61 % 5.669053e-01
timers/immediate.js type="clear" thousands=2000 311.40 % *** 3.896291e-07
timers/immediate.js type="depth" thousands=2000 17.54 % ** 9.755389e-03
timers/immediate.js type="depth1" thousands=2000 17.09 % *** 7.176229e-04
timers/set-immediate-breadth-args.js millions=5 10.63 % * 4.250034e-02
timers/set-immediate-breadth.js millions=10 20.62 % *** 9.150439e-07
timers/set-immediate-depth-args.js millions=10 17.97 % *** 6.819135e-10
Backport-PR-URL: #18179
PR-URL: #17064
Reviewed-By: Refael Ackermann <refack@gmail.com>
Reviewed-By: Minwoo Jung <minwoo@nodesource.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
async_hooks emits trace_events. This adds the executionAsyncId to the init events. In theory this could be inferred from the before and after events but this is much simpler and doesn't require knowledge of all events. Backport-PR-URL: #18179 PR-URL: #17196 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Franziska Hinkelmann <franziska.hinkelmann@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
When context is missing the executionAsyncId will be zero. For the default triggerAsyncId the zero value was used to default to the executionAsyncId. While this was not technically wrong because the functions are different themself, it poorly separated the two concepts. Backport-PR-URL: #18179 PR-URL: #17273 Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com>
Previously the getter would mutate the kDefaultTriggerAsncId value. This refactor changes the setter to bind the current kDefaultTriggerAsncId to a scope, such that the getter doesn't have to mutate its own value. Backport-PR-URL: #18179 PR-URL: #17273 Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com>
Backport-PR-URL: #18179 PR-URL: #17757 Reviewed-By: Anatoli Papirovski <apapirovski@mac.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Daniel Bevenius <daniel.bevenius@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de> Reviewed-By: Jon Moss <me@jonathanmoss.me> Reviewed-By: Yuta Hiroto <hello@about-hiroppy.com> Reviewed-By: Timothy Gu <timothygu99@gmail.com> Reviewed-By: vdeturckheim <vlad2t@hotmail.com>
SetupHooks is only available via `process.binding('async_wrap')`, so
there's no reason it shouldn't be called with the appropriate arguments,
since it is an internal-only function. The only place this function is
used is `lib/internal/async_hooks.js`.
Backport-PR-URL: #18179
PR-URL: #17832
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Anatoli Papirovski <apapirovski@mac.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Gireesh Punathil <gpunathi@in.ibm.com>
Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
- Communicate the current async stack length through a
typed array field rather than a native binding method
- Add a new fixed-size `async_ids_fast_stack` typed array
that contains the async ID stack up to a fixed limit.
This increases performance noticeably, since most of the time
the async ID stack will not be more than a handful of
levels deep.
- Make the JS `pushAsyncIds()` and `popAsyncIds()` functions
do the same thing as the native ones if the fast path
is applicable.
Benchmarks:
$ ./node benchmark/compare.js --new ./node --old ./node-master --runs 10 --filter next-tick process | Rscript benchmark/compare.R
[00:03:25|% 100| 6/6 files | 20/20 runs | 1/1 configs]: Done
improvement confidence p.value
process/next-tick-breadth-args.js millions=4 19.72 % *** 3.013913e-06
process/next-tick-breadth.js millions=4 27.33 % *** 5.847983e-11
process/next-tick-depth-args.js millions=12 40.08 % *** 1.237127e-13
process/next-tick-depth.js millions=12 77.27 % *** 1.413290e-11
process/next-tick-exec-args.js millions=5 13.58 % *** 1.245180e-07
process/next-tick-exec.js millions=5 16.80 % *** 2.961386e-07
Backport-PR-URL: #18179
PR-URL: #17780
Reviewed-By: James M Snell <jasnell@gmail.com>
The existing version of defaultTriggerAsyncIdScope creates an Array for the callback's arguments which is highly inefficient. Instead, use rest syntax and allow V8 to do that work for us. This yields roughly 2x performance for this particular function. Backport-PR-URL: #18179 PR-URL: #18004 Reviewed-By: Andreas Madsen <amwebdk@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
This allows more easy tracking of where HTTP requests come from. Before this change the HTTPParser would have the HTTPServer as the triggerAsyncId. The HTTPParser will still have the executionAsyncId set to the HTTP Server so that information is still directly available. Indirectly, the TCP socket itself also has its triggerAsyncId set to the TCP Server. Backport-PR-URL: #18179 PR-URL: #18003 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Anatoli Papirovski <apapirovski@mac.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Jon Moss <me@jonathanmoss.me> Reviewed-By: Daijiro Wachi <daijiro.wachi@gmail.com>
If IPv6 is not supported on a machine, the IPv6 handle will first be created, this will then fail and default to an IPv4 handle. This causes the graph to change, as there now is an extra handle. Backport-PR-URL: #18179 PR-URL: #18143 Fixes: #18003 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Rich Trott <rtrott@gmail.com> Reviewed-By: Anatoli Papirovski <apapirovski@mac.com> Reviewed-By: James M Snell <jasnell@gmail.com>
- Remove pointless pointers - Make `WriteHost` take a const argument so that it’s functionality is clear from the signature - Make `FindLongestZeroSequence` templated to accommodate the constness in `WriteHost` and because using `uint16_t` is an articifial, unnecessary restriction - Remove string copying when no copies are needed - Make `PercentDecode` just return its return value - Make `ParseHost` (string-only version) take its constant argument as a constant reference PR-URL: #17470 Reviewed-By: Timothy Gu <timothygu99@gmail.com>
This helps because `static` doesn’t work for C++ classes, but refactoring `url_host` into a proper C++ class seems the most reasonable soluation for the memory leak fixed by the next commit. PR-URL: #17470 Reviewed-By: Timothy Gu <timothygu99@gmail.com>
- Gives `URLHost` a proper destructor that clears memory depending on the type of the host (This fixes a memory leak) - Hide the host type enums and class layout as implementation details - Make the `Parse` methods members of `URLHost` - Turn `WriteHost` into a `ToString()` method on the `URLHost` class - Verify that at the beginning of a parse attempt, the type is set to “failed” - Remove a lot of `goto`s from the source code 🐢🚀 PR-URL: #17470 Fixes: #17448 Reviewed-By: Timothy Gu <timothygu99@gmail.com>
`last_piece` pointed to the end of the 8×16 bit array, so `piece_pointer == last_piece` already means that the pointer is not writable any longer. Previously, this still worked most of the time but could result in an out-of-bounds-write. Also, rename `last_piece` to `buffer_end` to avoid this pitfall. PR-URL: #17470 Reviewed-By: Timothy Gu <timothygu99@gmail.com>
The mustCall is actually only necessary in case it is used as callback. Otherwise it works as a must call on its own. PR-URL: #17616 Reviewed-By: Anatoli Papirovski <apapirovski@mac.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Jon Moss <me@jonathanmoss.me>
PR-URL: #17610 Refs: #17054 (comment) Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Evan Lucas <evanlucas@me.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com> Reviewed-By: Jon Moss <me@jonathanmoss.me> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Fix a nullptr dereference on abrupt termination when async call stack recording is enabled. Bug discovered while trying to write a regression test for the bug fix in commit df79b7d ("src: fix missing handlescope bug in inspector".) PR-URL: #17577 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: Timothy Gu <timothygu99@gmail.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Refael Ackermann <refack@gmail.com> Reviewed-By: Eugene Ostroukhov <eostroukhov@google.com>
Currently the documentation states that promisify() will result in undefined behavior if bad arguments are passed. This is not necessarily the case, since the behavior is well defined, but just not useful. PR-URL: #17593 Fixes: #17569 Reviewed-By: Anatoli Papirovski <apapirovski@mac.com> Reviewed-By: Rich Trott <rtrott@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de> Reviewed-By: Benjamin Gruenbaum <benjamingr@gmail.com> Reviewed-By: Jon Moss <me@jonathanmoss.me> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
According to the linked document: "The length of the randomly-generated portion of the salt shall be at least 128 bits." [NIST SP 800-132] PR-URL: #17524 Reviewed-By: Tobias Nießen <tniessen@tnie.de> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Daniel Bevenius <daniel.bevenius@gmail.com> Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl> Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
test-child-process-pass-fd needs to launch many processes simultaneously. On Fedora 24, this can result in EAGAIN "Resource temporarily unavailable" errors. When this occurs, simply try to launch a worker again. PR-URL: #17598 Fixes: #17589 Reviewed-By: Santiago Gimeno <santiago.gimeno@gmail.com> Reviewed-By: Gibson Fahnestock <gibfahn@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gireesh Punathil <gpunathi@in.ibm.com>
