Carl Dehlin

Hi @julescmay . Can you please provide a minimal reproducible example (MRE) to debug this issue? It's impossible to know what is going on otherwise. I am pretty sure you...

Hi, this ideas have been circulating for quite some while. I'll try to organize them under the new label category/value-parameterization. I believe this is the oldest issue on this topic:...

@jeaye I think this is a good feature. If you can implement that decorator and make a PR I'm sure it will be appreciated by others using your workflow as...

This is similar to this feature request: https://github.com/doctest/doctest/issues/322 Both request touch upon the subject of statistical testing. There needs to be a discussion if this is in scope of this...

I wonder how this differs from running the text exe itself multiple times with a test case filter? One liner workaround, does that work for you?

Thanks for the elaborate motivation for this feature. I guess it could be implemented with a decorator, or maybe as a value parameterized test. To motivate this further, could you...

Hi @douglasmacharet , have you checked out the reporters API? https://github.com/doctest/doctest/blob/dev/doc/markdown/reporters.md

@onqtam What is the policy on merging back to master? I would like these changes to be incorporated into the documentation, which people read on the master branch.

Improved implementation by utilizing exact negation for IEEE floats. Now `ulpadd` supports addition of negative ulps (i.e. you get `ulpsub` by passing a negative argument), and `ulpeq` can be implemented...

NVCC is not tested ATM. CUDA is quite different from standard C++, so this is a very minor bug.