evelez7 wrote: > Actually, thinking about this for another second. If our plan is to remove > the old backend, do we care about test parity? Is it valuable? We already > think the old output is bad, and have basically stopped adding new features > to it.
What I do find useful is having a verifiable source saying this is what the new backend emits compared to the old one. These tests are the only verification of what is actually implemented in the backend which is a pretty big problem because the actual output can have way more stuff. I learned that while implementing a Mustache MD backend - the old MD backend has way more functionality and output than what's tested. In this case, the precommit shows that we don't have certain important things in the templates, but you're right that the history might not look as good once the `MUSTACHE` prefixes turn into `HTML`. That might be double true since I wont be using the current test's global function markup, even though I put the `-NOT` prefix as if it was a precommit. I'm already not copying the output 1:1 now anyways. You can see that for example on line 128 where I omitted the `<p>` from the above HTML just because I didn't think it was useful right now. There's better ways to show `pwd` style info. I'm not looking to match the output, just give comparison. https://github.com/llvm/llvm-project/pull/169107 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
