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
There is currently a Native AOT test that gets run as part of the VMR scenario tests. But this test only validates that the publish worked. It doesn't verify a critical aspect of the source build requirements which are to avoid any dependency on a NuGet feed.
In fact the Native AOT publishing scenario test doesn't fulfill that requirement today. It ends up loading from a NuGet feed:
Installed runtime.linux-x64.Microsoft.DotNet.ILCompiler 10.0.0-alpha.1.25071.14 from https://pkgs.dev.azure.com/dnceng/public/_packaging/dotnet10/nuget/v3/index.json to /__w/1/s/src/scenario-tests/.packages/runtime.linux-x64.microsoft.dotnet.ilcompiler/10.0.0-alpha.1.25071.14 with content hash pzLeGawFrvq3a3MyljNCHurehIhhVMEFS+K/xTgHxAM602oqVU/8qJIFtgxMIexNMN7aGB+QSXSraDBJjjueIg==.
Note that such validation behavior should be defined in Microsoft.DotNet.SourceBuild.Tests since it is source build-specific; it should not be in the scenario tests repo.
There is currently a Native AOT test that gets run as part of the VMR scenario tests. But this test only validates that the publish worked. It doesn't verify a critical aspect of the source build requirements which are to #1215.
I had assumed it wouldn't be able find an ilcompiler for linux-x64. I wasn't aware one was being provided through extra NuGet sources.
Note that such validation behavior should be defined in Microsoft.DotNet.SourceBuild.Tests since it is source build-specific; it should not be in the scenario tests repo.
I think it would be more clear to have the functional test for PublishAOT in a single test suite and have that test suite include the difference is between portable vs non-portable publish.
Placing the source-build specific validation in SourceBuild.Tests means duplicate code, overlapping coverage, and less understanding of how this works portable vs non portable.
Yeah, I agree with that. It should be straightforward to provide a property to scenario tests that indicates this state and have the tests do the validation accordingly.
There is currently a Native AOT test that gets run as part of the VMR scenario tests. But this test only validates that the publish worked. It doesn't verify a critical aspect of the source build requirements which are to avoid any dependency on a NuGet feed.
In fact the Native AOT publishing scenario test doesn't fulfill that requirement today. It ends up loading from a NuGet feed:
This would be fixed by dotnet/scenario-tests#184. Having validation would have caught this. This kind of validation is already done for the self-contained publishing test at https://github.com/dotnet/sdk/blob/b324c2e074f7691705b85b6ed69be23d350164b8/src/SourceBuild/content/test/Microsoft.DotNet.SourceBuild.Tests/WebScenarioTests.cs#L45.
Note that such validation behavior should be defined in Microsoft.DotNet.SourceBuild.Tests since it is source build-specific; it should not be in the scenario tests repo.
cc @tmds
The text was updated successfully, but these errors were encountered: