Andrey Semashev wrote:
There was a single failure, clang-win in release,
...failed testing.capture-output > ..\..\bin.v2\libs\filesystem\test\operations_unit_test.test\clang-win-10.0.0\release\threading-multi\operations_unit_test.run... but it's also present on the master branch, so it's not a regression.
Could you provide more details for the error?
The output from the test is:
Does it also fail when you run tests sequentially? I.e. just one compiler and just one variant, with -j1?
This looks like a Clang x86 codegen bug. The compiler returns the `file_status` from `get_status` on the stack and then overwrites it with `push` instructions for some reason. Not the first x86 clang-cl problem. I'll upgrade from 10.0 to 11.0 to see if it's fixed.