diagnostics_channel: capture console messages #56292
Conversation
Qard
added
console
Qard
force-pushed
the
console-diagnostics-channels
branch
from
6ba11e5 to
07fbcd4
Compare
github-actions
bot
removed
the
request-ci
Qard
force-pushed
the
console-diagnostics-channels
branch
from
07fbcd4 to
8094626
Compare
Qard
force-pushed
the
console-diagnostics-channels
branch
from
8094626 to
b3830bf
Compare
|
I lack context on this, so I might be missing some point. Can this use the Inspector protocol instead of adding hooks directly into console object. In-process inspector API can be configured to watch for the messages? This way there will be less coupling. |
|
The Inspector API is substantially more expensive. This is meant for being able to rewrite potentially all log messages in production, so it needs to be fast. |
|
Thanks for the reply. |
|
Maybe worth to extend test to verify adding/removing/replacing elements, Main problem I see with mutating diagnostics channels is the undefined sequence if there are more subscribers. |
|
|
||
| * `args` {any\[]} | ||
|
|
||
| Emitted when `console.log()` is called. Receives and array of the arguments |
There was a problem hiding this comment.
Maybe describe that this is the arguments array and therefore mutation is possible.
Drain the loop and platform tasks before creating a snapshot. This is necessary to ensure that the no roots are held by the the platform tasks, which may reference objects associated with a context. For example, a WeakRef may schedule an per-isolate platform task as a GC root, and referencing an object in a context, causing an assertion in the snapshot creator. PR-URL: #56403 Refs: #56292 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Joyee Cheung <[email protected]>
|
Would you mind rebasing on top of the main branch? The build failure should be fixed once rebased. |
I've added diagnostics_channel support to capture and modify inputs to the main console methods. This enables some useful things like APMs injecting tracing data into log messages to correlate log messages to the requests from which they originated. You can also replace the args list entirely, so you could apply your own custom formatting to produce JSON logs instead or things like that.
cc @nodejs/diagnostics @nodejs/console