-
Notifications
You must be signed in to change notification settings - Fork 10.3k
Update SDK to 9.0.100-alpha.1.23421.9 #49911
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
Conversation
Hey @dotnet/aspnet-build, looks like this PR is something you want to take a look at. |
Both this and the efcore SDK update PR are failing to build with this NETSDK1210 error. All of our Assuming that these NETSDK1210 errors are legitimate, I could update all our |
@sbomer are you aware of any relevant recent changes? |
We don't have It's confusing to have an error message about a property that isn't present in source code. SDK should complain about |
The Windows build servers are also failing with a seemingly unrelated error when trying to build some ANCM vcxproj's:
I'm not sure why ANCM would need an emscripten workload or why it's missing. |
@sbomer - I believe this is caused by dotnet/sdk#34077. Can you take a look? |
It is probably due to |
I updated Directory.Build.Targets to unset IsTrimmable when not targeting DefaultNetCoreTargetFramework. I tried just conditioning EnableAOTAnalyzer on the target framework or IsAotCompatible without overriding IsTrimmable, but then I still got NETSDK1195 errors. I'm worried I might disable something important if I remove the workaround at https://github.com/dotnet/aspnetcore/blob/1d6e871b4a171a1d3cd34460d745b8722642b5d5/Directory.Build.targets#L4C1-L5, so I'm leaving that for a follow up. I made a similar change in EF, but for IsAotCompatible instead of IsTrimmable. |
We're still getting the following error I mentioned yesterday:
I noticed there was a workload-related failure in the last SDK update (#49761) for a different wasm linking workload, but that seems to have had a different cause. Does anyone know what's going on here? @radical @lewing @joeloff @marcpopMSFT |
Do you know which workloads are present? Wasm has a dependency on EMSDK and that should get installed as part of installing wasm-tools. |
I just ran restore.cmd before and after this SDK update and I noticed a difference to the location where emscripten's WorkloadManifest.json is installed. Before it was Noticed the new Is this intentional? |
It is intentional. The new path is related to workload sets (version and grouping) that's being introduced. @dsplaisted as an FYI |
Are you building using VS/fullframework and what version of VS is used in the builds? We had to make a fix in 17.7-preview5 and it just released 17.7.0 today which may affect this. Can you update to a preview image of VS? |
This definitely has something to do with this SDK update and not just the environment considering the main branch and other PRs are still building fine. The latest failing Build.native.binlog for Windows_Test_Logs shows This is the first SDK update that the aspnetcore repo has consumed where the Is MSBuild version 17.6.3 is too old to get the latest changes to SdkDirectoryWorkloadManifestProvider that cause it to look in versioned subfolders for the workload manifests? How do we update it? |
I believe you need 17.7 and the only way to get that is to move to a different OS image. |
Since this is failing in the azure pipeline before we get the chance to see how this affects our helix tests, should I update default-build.yml here? aspnetcore/.azure/pipelines/jobs/default-build.yml Lines 124 to 131 in 0b35f03
It looks like we're currently using Will Helix.Common.props need a similar update? |
We use Windows.Amd64.VS2022.Pre.Open for our test runs which pulls in preview versions of VS. For our builds, we don't use full framework so use the same images that you use. Not sure if there is a version with VS as an alternative. |
We queue jobs to more than just Windows.Amd64.VS2022.Pre.Open: aspnetcore/eng/targets/Helix.Common.props Lines 33 to 34 in 12b48e6
I'm not sure if the additional Windows Helix queues will be an issue, but first we need to figure out how to get arcade working on the dnceng 1ES agent pools. Right now, it looks like every job using a Windows AzDO agent is failing with this workload resolution error. |
We only need to update aspnetcore/.azure/pipelines/jobs/default-build.yml Lines 124 to 131 in 0b35f03
|
One sec, let me figure out what to update to |
Looks like |
I think this is because the breaking change is in |
No updates yet. Highlighting a point @dougbu made:
Has that been considered? Is there something in 17.7 that is expected to allow this to work that's not present in 17.8? |
Checking to see if the error was the same in the 1es images vs the Scout images |
1es updated from 17.6 to 17.7 yesterday, so this might just work now - I think we need the combination of VS 17.7 (or newer), plus the windows toolsets/native components on the 1es images |
/azp run |
Azure Pipelines successfully started running 3 pipeline(s). |
Wait until after #50386 is in to resolve the conflict, but it looks like this should work now |
/azp run |
Azure Pipelines successfully started running 3 pipeline(s). |
@wtgodbe to confirm: you're not expecting a need for VS 17.7 in the build pools DNCEng provides i.e., the 1es images should now work here❔ |
That's right, sorry for the confusion - I mixed up the initial workload error w/ the missing components error. Now that 1es has updated to 17.7 we are good to go |
Down to a couple test failures:
|
@adityamandaleeka could you assign someone to take a look at the WebSocketMiddleware test failures above? Hopefully it's a trivial fix |
I took a look, and the tests passed locally. I added some logging to get information about where the ArgumentNullExceptions originated. |
This is the error message:
Moving the buffer to where it is used fixed this. I don't know the cause, but something about the previous method shape caused the error. Probably JIT related. Edit: Created an issue: dotnet/roslyn#69777 |
No description provided.