Installing an uncaughtException listener has a side effect that process
is not aborted. This is quite bad for monitoring/logging tools which
tend to be interested in errors but don't want to cause side effects
like swallow an exception or change the output on console.
There are some workarounds in the wild like monkey patching emit or
rethrow in the exception if monitoring tool detects that it is the only
listener but this is error prone and risky.
This PR allows to install a listener to monitor uncaughtException
without the side effect to consider the exception has handled.
PR-URL: https://github.com/nodejs/node/pull/31257
Refs: https://github.com/nodejs/node/pull/30932
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
This patch splits the handling of `isMainThread` and
`ownsProcessState` from conditionals in
`lib/internal/bootstrap/node.js` into different scripts under
`lib/internal/bootstrap/switches/`, and call them accordingly
from C++ after `node.js` is run.
This:
- Creates a common denominator of the main thread and the worker
thread bootstrap that can be snapshotted and shared by
both.
- Makes it possible to override the configurations on-the-fly.
PR-URL: https://github.com/nodejs/node/pull/30862
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Chengzhong Wu <legendecas@gmail.com>
It's likely that anyone using `process.report.getReport()` will be
processing the return value thereafter (e.g., filtering fields or
redacting secrets). This change eliminates boilerplate by calling
`JSON.parse()` on the return value.
Also modified the `validateContent()` and `validate()` test helpers in
`test/common/report.js` to be somewhat more obvious and helpful. Of
note, a report failing validation will now be easier (though still not
_easy_) to read when prepended to the stack trace.
- Refs: https://github.com/nodejs/diagnostics/issues/315
PR-URL: https://github.com/nodejs/node/pull/28630
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Richard Lau <riclau@uk.ibm.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Jiawen Geng <technicalcute@gmail.com>
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Reviewed-By: Rich Trott <rtrott@gmail.com>
Previously the enhancement were done right after emitting
`'uncaughtException'`, which meant by the time we knew the
exception was fatal in C++, the error.stack had already been
patched.
This patch moves those routines to be called later during the
fatal exception handling, and split them into two stages:
before and after the inspector is notified by the invocation of
`V8Inspector::exceptionThrown`. We now expand the stack to include
additional informations about unhandled 'error' events before
the inspector is notified, but delay the highlighting of the
frames until after the inspector is notified, so that the
ANSI escape sequences won't show up in the inspector console.
PR-URL: https://github.com/nodejs/node/pull/28308
Fixes: https://github.com/nodejs/node/issues/28287
Reviewed-By: Benjamin Gruenbaum <benjamingr@gmail.com>
The C++ land `node::FatalException()` is not in fact fatal anymore.
It gives the user a chance to handle the uncaught exception
globally by listening to the `uncaughtException` event. This patch
renames it to `TriggerUncaughtException` in C++ to avoid the confusion.
In addition rename the JS land handler to `onGlobalUncaughtException`
to reflect its purpose - we have to keep the alias
`process._fatalException` and use that for now since it has been
monkey-patchable in the user land.
This patch also
- Adds more comments to the global uncaught exception handling routine
- Puts a few other C++ error handling functions into the `errors`
namespace
- Moves error-handling-related bindings to the `errors` binding.
Refs: 2b252acea4
PR-URL: https://github.com/nodejs/node/pull/28257
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Benjamin Gruenbaum <benjamingr@gmail.com>
Reviewed-By: Franziska Hinkelmann <franziska.hinkelmann@gmail.com>
Emit the deprecation warning in the `kDefaultUnhandledRejections`
case to reduce the number of branches on unhandled rejection mode -
there is now only one switch case on it.
Also rename `emitWarning()` to `emitUnhandledRejectionWarning()`
to avoid ambiguity with `process.emitWarning()`
PR-URL: https://github.com/nodejs/node/pull/28258
Reviewed-By: Benjamin Gruenbaum <benjamingr@gmail.com>
Reviewed-By: Franziska Hinkelmann <franziska.hinkelmann@gmail.com>
- Use constants instead of a dictionary and add comments
about the behavior of each mode.
- Use switch cases to handle the unhandled rejection modes.
- Rename the run time value of the CLI option from `state`
to `unhandledRejectionsMode`.
- Return in the call site of `emitWarning` when
`--unhandled-rejections=none` instead of inside
the function.
PR-URL: https://github.com/nodejs/node/pull/28228
Reviewed-By: Rich Trott <rtrott@gmail.com>
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Fix a few edge cases and non-obvious issues with nextTick:
1. Emit destroy hook in a try-finally rather than triggering
it before the callback runs.
2. Re-word comment for processPromiseRejections and make sure
it returns true in the rejectionHandled case too.
3. Small readability improvements.
PR-URL: https://github.com/nodejs/node/pull/28047
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Reviewed-By: Joyee Cheung <joyeec9h3@gmail.com>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
This makes sure that errors that shut down the application are
inspected with `util.inspect()`. That makes sure that all extra
properties on the error will be visible and also that the stack trace
is highlighted (Node.js internal frames will be grey and node modules
are underlined).
PR-URL: https://github.com/nodejs/node/pull/27243
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Michaël Zasso <targos@protonmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
Previously, we call the JS land `runNextTicks` implementation
immediately from JS land after evaluating the main module or the
input, so these synchronous JS call frames would show up in the stack
trace of the async errors, which can be confusing. This patch moves
those calls into C++ so that more of these internal scheduler
implementation details can be hidden and the users can see a cleaner
a cleaner async JS stack trace.
PR-URL: https://github.com/nodejs/node/pull/27392
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Reviewed-By: Gus Caplan <me@gus.host>
Reviewed-By: Minwoo Jung <minwoo@nodesource.com>
Reviewed-By: Benjamin Gruenbaum <benjamingr@gmail.com>
This patch adds an internal function that prints to stdout or
stderr by directly writing to the known file descriptor, and
uses it internally in common cases to avoid the overhead
of the console implementation.
PR-URL: https://github.com/nodejs/node/pull/27320
Reviewed-By: James M Snell <jasnell@gmail.com>
Since the bootstrap does not actually use ESM at all, there
is no need to create this map so early. This patch moves
the initialization of the map to pre-execution,
so that the only binding loaded in loaders is native_module.
In addition, switch to SafeWeakMap.
PR-URL: https://github.com/nodejs/node/pull/27323
Reviewed-By: Gus Caplan <me@gus.host>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
This patch reduce usage of `process._breakFirstLine` and
`process._eval` in the internals and use
`getOptionValue('--inspect-brk')` and `getOptionValue('--eval')`
instead wherever possible.
PR-URL: https://github.com/nodejs/node/pull/27278
Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Reviewed-By: Anna Henningsen <anna@addaleax.net>
