Speed up native AOT testing #98939
Speed up native AOT testing #98939
Conversation
We don't need to run analyzers on native AOT legs because they run on the same code elsewhere. Locally drops `clr.aot+libs` build from 5 mins 30 secs to 4 mins 45 secs.
|
Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas Issue DetailsWe don't need to run analyzers on native AOT legs because they run on the same code elsewhere. Locally drops Cc @dotnet/ilc-contrib
|
ghost
assigned 
|
cc @dotnet/area-infrastructure-libraries |
Do you mean whether building If it's 1, I would prefer us to be consistent with the rest of the product. Doing this on specific CI legs is just a perf optimization (we might have more legs where this could save time without impacting anything). Disabling analyzers locally for the whole product is the same discussion we had about disabling warnaserror locally. Some people had strong feelings about it. I disable warnaserror locally with an environment variable. I might end up doing the same for analyzers because their cost to my inner loop doesn't justify what they do; they usually just get in my way. |
|
I think it makes sense to make most, if not all, analyzers run only in a separate pipeline built specifically for that purpose. I'm fine with this specific change as well. |
I explored this in the past (for the warnaserror - so that we don't fail all pipelines if there's a warning) and my main concern was that because we have so many |
We don't need to run analyzers on native AOT legs because they run on the same code elsewhere. Locally drops
clr.aot+libsbuild from 5 mins 30 secs to 4 mins 45 secs.Cc @dotnet/ilc-contrib