================
@@ -68,7 +68,7 @@ function compute-projects-to-test() {
       done
     ;;
     clang)
-      for p in clang-tools-extra compiler-rt flang lldb cross-project-tests; do
+      for p in clang-tools-extra compiler-rt lldb cross-project-tests; do
----------------
AaronBallman wrote:

> The principle here is conservative: if a change to clang might affect another 
> project, then it should be tested.

This approach is very expensive in practice. It costs us significant amount of 
overhead in precommit CI (14+ hour-long waits, as with 
https://buildkite.com/llvm-project/github-pull-requests/builds/64748) and the 
compute resources are also monetarily expensive (as I understand things).

> Basically take the finest grain dependency in your head, then collapse the 
> dependency graph per top-level projects.
"clangDriver" and "clangBasic" aren't projects, right now we can only track 
dependencies at this granularity (patches welcome of course, but that seems 
hard to keep maintained).

Ah, that's good to know, thanks! If we don't have the granularity to support 
precommit CI in a workable way, then I think the Flang tests should only be run 
only when Flang changes or in post-commit (so test coverage remains the same, 
but failures are found slightly later than if we did full precommit CI 
testing). After all, any change in LLVM can also impact clang, which can also 
impact Flang. Given the expense and pain involved, I don't think the 
conservative approach works any longer.

> Because you're trying to "not test" something that should be tested, only for 
> the sake of a problem on windows.

I think this trivializes the current situation -- this is not "for the sake of 
a problem on windows", this is for the sake of the entire community that has 
been raising concerns with how unusable precommit CI has been for the past ~six 
months. This significantly impacts folks doing code reviews because they feel 
they need to wait for days to for precommit CI to come back green -- Windows 
just happens to be the largest pain point at the moment.

https://github.com/llvm/llvm-project/pull/92740
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to