-
Notifications
You must be signed in to change notification settings - Fork 126
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
Design and implement user experience around linker analysis warnings #1165
Comments
For this we could make sure that all analysis warnings declare a subcategory and then check in
For this we'd need to cache all warnings and print them when the linker exits. I don't think this adds much overhead to the overall process, should we do it? |
We've discussed this to some degree already - I think the conclusion was to disable a set of warning codes on the MSBuild side. (I think it's the
The off-by-default mentioned above via So the end result is that linker always produces the warnings, but sometimes (most of the time) they're filtered out by MSBuild. This needs to be tried though - I'm not sure it works well end-to-end. |
I am a little wary of doing this in MSBuild. It seems unusual to introduce warnings that aren't ready to be on by default, but still emit them, then filter them out by default, and have an option on top of that to un-filter them. It would be more straightforward not to emit experimental warnings by default, and pass a flag to the linker that turns them on. Then there would be no MSBuild magic, and Just wanted to express this concern - in this particular case adding it at the MSBuild layer might be the most expedient thing to do, just to avoid introducing a new command-line and task input. It just means there won't be a simple way to disable them when running illink from the command line or using the task directly. My hope is that we can avoid this pattern, and get rid of the MSBuild logic once we are confident that the warnings are "ready". |
I'm all for clearly separating analyzer warnings by introducing a new category or even using a special number range for quick overview check. If we go with category approach we could have something like I don't think we have to sort the warnings at the end as we would need to come up with sorting rules which will be imperfect anyway. Issuing multiple warnings is something that will need to be tweaked manually to reduce cascading or overlapping warnings (csc has that problem as well) and it's not a generally serious concern. |
I opened a PR with a proposal: #1303. |
What would we say the status of this issue is? |
I think we're getting close - things we should do:
|
I think this is basically done. Notable improvements from the last comment here:
|
This mostly concerned with the warnings produced by the data flow analysis and correctness checks, but in theory it applies to all linker warnings.
Things to consider
The text was updated successfully, but these errors were encountered: