Currently, we cannot run the flang tests on Windows since it uses some shell scripts (like ./test/Semantics/test_errors.sh). The Windows buildbot (https://lab.llvm.org/buildbot/#/builders/172) set up by Michael Kruse currently builds flang but does not run the tests due to this reason. Porting these shell scripts to Python was suggested as a possible solution. We have an intern who is interested in taking up this task. I expect no change in functionality and the Python script will be a straightforward replacement of the shell script.
Is porting the testing scripts to Python OK with everyone?
Yes, using python would be great. Python is required anyway and even
on non-Windows systems bash cannot be assumed to be just present (e.g.
MacOS does not install it anymore by default and only installs a very
old version of Bash due to GPLv3 issues).
By comparison, Clang has a `-verify` option in the binary which would
be a possibility for Flang as well without needing to regex the
+1 to this proposal!
Clang's `-verify` is implemented in terms of Clang's sophisticated diagnostics API. It would be fantastic to introduce something similar in Flang, but it would require re-designing how diagnostics are generated and dumped. Something worth considering.