Code coverage support #154
Conversation
This appears to be a bit tricky since this implementation processes all files as single awk source and does some validations (like making sure functions are defined). This is a problem if we want to annotate on per-file basis, since we need to parse files separately.
This appears to be a bit tricky since this implementation processes all files as single awk source and does some validations (like making sure functions are defined). This is a problem if we want to annotate on per-file basis, since we need to parse files separately.
This appears to be a dead end... This reverts commit 3a3a6c2. Revert "Work on annotating multiple files (multiple -f provided)." This reverts commit fb42783. Revert "Work on annotating multiple files (multiple -f provided)." This reverts commit 5dbafa3. Revert "Work on annotating multiple files (multiple -f provided)." This reverts commit bb09b88. Revert "Work on annotating multiple files (multiple -f provided)." This reverts commit 3e0c70b.
|
Actually I managed to fix tests for windows. Still there is problem with macOS. |
|
Thanks! I'll do the final review later today. It looks like the macOS tests are failing, not because of this PR, but because homebrew |
Conflicts: - goawk.go
|
FYI, I resolve the merge conflicts after merging #155. |
|
Yeah, I think we should add a few basic tests for the |
|
Actually, I just opened an issue to track that: #156 |
|
This is great, thank you. Just a couple of nit comments above, and then I'll merge this in. Looking forward to playing with AWK coverage! :-) |
…clude the leading hyphen: `"-covermode can only be one of: set, count"`
Hopefully, done. |
One more PR for #144 that addresses couple of minor things, missed in #154 In this PR: 1. Remove unintuitive behavior (outputting annotated code when only `-covermode` flag set) - just use `-d` flag 2. Update cover.png screenshot: now we show if/for/while conditions in green (covered) 3. Fix for `./goawk -covermode=set 'BEGIN { print 123 }'` to not hang waiting for stdin. This started to happen because processing stops if `prog.Actions == nil` and we by mistake were changing it to empty slice during coverage annotation.
This PR implements #144 and tries to address all CR notes in #144 (comment).
This PR is based on top of previous related PRs #148, #153 in order to eliminate the necessity for different hacky solutions.
I think code-wise this is ready. The only thing I still want to add - a documentation in own file
coverage.md. While I'm working on this I would be really glad if you @benhoyt check the current implementation in this PR. Hopefully now it should be in much better shape than with prevous hacky approach.