You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Jul 27, 2023. It is now read-only.
I've wired up rusttrace(https://github.com/flowerinthenight/rusttrace) crate in xi-win ui.
It's a small mess to setup(can make a powershell script to automate it), but can be handy for debugging, in absence of println!, also ETW events can be used by tools like perfview for profiling.
Was wondering if you see any value in a pull request for this, if not so I could avoid the trouble of signing google license agreements.
Here's an example trace as displayed in the command line by mftrace.
Hey Codri - were you able to receive logging information from the core as well? I'm mostly interested in the errors, as currently they are ignored, but this might be a useful solution. (See #60)
Also, the CLA is no longer necessary. :)
No, I didn't wire it up for core. Only for requests/responses going from/into xi-win.
I was thinking about the whole state of xi logging, and tracing, as well as how to integrate with etw.
Here are a few notes/toughts:
There already seems to be an effort to add logging to xi-editor. see Real logging in xi-core (+ core RPC logging) xi-editor#408 but, this is not quite what etw can provide, since etw can be used for performance tracing and can be enabled/disabled on runtime, with it being zero cost when not enabled.
The library mentioned above(rusttrace) does only string tracing for etw, which is not ideal, etw has the concept of structured logging, which is more flexible and performant. For example, for the xi-rpc method one can define an enum, and not collect redundant strings, which will impact perf test more than collecting an integer would.
LTTng and ETW are very similar conceptually, maybe it's worth making a structured logging/tracing library that uses ETW on Windows and LTTng on Linux. Not sure it makes sense respecting the log interface, as that is mainly for string logging.
ETW has a series of modes to do logging, in one you specify a XML template(if I remember correctly) defining the schema of your logger, which is messy. There is another mode where you use system calls to register event schema on startup, and the client can ask the OS to get it for a certain process. The latter I think is the way to go.
So, long story short, ETW and LTTng are interesting pieces of technology, can be a great fit for high performance crossplatform tracing. But, I'm yet to make the thin wrapper library that can log into both.
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Hi,
I've wired up rusttrace(https://github.com/flowerinthenight/rusttrace) crate in xi-win ui.
It's a small mess to setup(can make a powershell script to automate it), but can be handy for debugging, in absence of println!, also ETW events can be used by tools like perfview for profiling.
Was wondering if you see any value in a pull request for this, if not so I could avoid the trouble of signing google license agreements.
Here's an example trace as displayed in the command line by mftrace.
The text was updated successfully, but these errors were encountered: