-
Notifications
You must be signed in to change notification settings - Fork 4.7k
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
Target .NET 10 / net10.0 #106599
base: main
Are you sure you want to change the base?
Target .NET 10 / net10.0 #106599
Conversation
Tagging subscribers to this area: @dotnet/runtime-infrastructure |
to be able to build System.Text.Json on net9.0.
...libraries/System.Diagnostics.DiagnosticSource/src/System.Diagnostics.DiagnosticSource.csproj
Show resolved
Hide resolved
@ViktorHofer I already had this PR open for this purpose: #106421 |
Shouldn't we change this line?
|
Ah, it was added four days ago b14e2f5, and the branch I was grepping in was a week old. 🤞 |
|
Yup, it's picked up. I have reapplied the workaround (which I removed to see if it's needed after the clone is fixed). Please retry. |
Worked, performance is green. We should apply it to perf repo and then revert the changes to cloned url. Edit: my bad, not fully green: Edit2: please, change |
We're stuck on
shouldn't we change the arg to be |
Is it so that before #101143, major version change wasn't getting blocked on perf leg? It could be we might have missed considering this detail during that refactoring work. cc @DrewScoggins, @caaavik-msft One way to move forward could be to disable perf legs and enable it once the major version transition has been completed. |
Typically, we continue to target 9 but install the net10 sdk during transition periods like this. I'd rather not disable the perf tests, it has lead to large gaps in testing in the past that included regressions that were extremely difficult to isolate. |
This should still be possible. Looking at the wasm-runtime-perf build logs the only thing that was stopping this from working was the version parsing issue. I think if we just revert the changes to channel_map.py from dotnet/performance#4470 then it should work. If it doesn't, I'd be curious to know what the error is. |
dotnet/performance#4470 (comment) <- this was the error before making changes in channel_map. Feel free to push to that PR and just rerun this leg https://dev.azure.com/dnceng-public/public/_build/results?buildId=813887&view=logs&j=59ba1e23-1c3f-5e83-bfa2-2b4c458b95b8&t=606691c3-9bf3-5776-e07f-204fdb23192e (it's already on attempt 7). |
After looking into it some more, I think the reason it isn't working is because of the new SDK validation feature (dotnet/BenchmarkDotNet#2523, dotnet/BenchmarkDotNet#2538) that is part of BenchmarkDotNet which will validate that the SDK is installed. The SDK validation feature was introduced 6 months ago and so was not an issue when we went from .NET 8 to 9. |
Does this mean the perf runs will fail for all the runtimes but we're only seeing wasm fail because of our validation run? |
Just a reminder that runtime flow into the sdk is stuck at dotnet/sdk#43061 (or dotnet/sdk#43070 ) and this is the next step in unblocking that. |
I ran our full performance test suite, and although it is still running I think it only affects wasm and nativeaot because for them we specify a custom runtime argument (link to source) which specifies which SDK to run against whereas for other runs like coreclr there is no SDK passed in, it just uses the corerun path. We have a few options here:
I'm going to see if option 3 works first, but if it doesn't then I'll probably proceed with option 2 which might take until Monday to get merged into BDN so we can start using it. One other thing is that it has been quite a while since we last updated our BDN version, I am concerned that if we pull the latest version of BDN that it might cause other things to break and it might take longer than we expected if other bug fixes need to be made. If that happens we might just make a custom build of BDN based off the current version we are using but only with the fix for the SDK validator. |
This will scale better. Let SDK handle the validations and delete those dozen of back and forth mappings. If BDN’s excuse is that public moniker enum used in the attribute, it’s probably worth taking a breaking change; delete the enum and replace with parameterized one |
Agreed, whatever the solution let's assume the same thing will happen again next September and plan for that this time. |
Made the BDN PR here dotnet/BenchmarkDotNet#2641 |
Contributes to #105343