-
Notifications
You must be signed in to change notification settings - Fork 122
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix printing issue in perfstub thirdparty code #3913
Conversation
The CI errors are unrelated to the changes from what I can tell so the PR can be reviewed
|
LGTM |
I'm kind of late to this, but changes should not be made directly to third party libraries like it says in |
Whoops, @caitlinross is absolutely correct. Oddly, I thought that perfstubs was actually part of ADIOS, despite its location in thirdparty. But this needs to be rolled back, fixed in Kevin's repo and pulled back in from that. (Or, if we don't want to depend upon a third party for this, we need to adopt perfstubs and move it into toolkit.) |
Yeah it used to be under |
So my vote is not to undo that work that was done a couple of years ago, and leave it as thirdparty and do what @eisenhauer said to rollback, make the change upstream and pull it in correctly. |
yeah, that's my bad - I should have done it in the main repo. I think we should remove it entirely, or make it optional. There's a possibility that PerfStubs could get initialized before MPI_Init, so it wouldn't know the MPI rank. It's also possible to do something like this to get the MPI rank without making an MPI call: https://github.com/UO-OACISS/tau2/blob/df1b6975cf9fd47456793b64277af7af4117cd18/src/Profile/RtsLayer.cpp#L149-L206 ...but it would have to be maintained/updated to support all job schedulers. Maybe the best thing to do is remove the printf entirely. |
Yeah I'd probably just remove the printf. |
Ups I didn't know this about thirdparty code :) |
I have removed the printf in the master branch of perfstubs, you should be good to run the magic script ( |
I don't have time to figure out why the update is not working until tomorrow evening so if someone else could help, that would be great. This is what I get:
|
It appears that you have an existing worktree in the path |
During the SC23 ADIOS2 tutorial @khuck noticed that only rank 0 should print the initialization message in perfstub and not all ranks. If this PR is considered an overkill (passing the rank all the way to the initializer) the second option is to just comment out the print.