This is an archive of the discontinued LLVM Phabricator instance.

[llvm-cov] Unify region marker placement between text/html modes
ClosedPublic

Authored by vsk on Jul 28 2017, 2:24 PM.

Details

Summary

Make sure that the text and html emitters always emit the same set of
region markers, and avoid emitting redundant markers for line segments
which don't end on the line they start on.

Here is the html output from the deferred-region.cpp test:

This is related to D35925, and depends on D36014

Diff Detail

Repository
rL LLVM

Event Timeline

vsk created this revision.Jul 28 2017, 2:24 PM
vsk added a subscriber: efriedma.Aug 3 2017, 7:20 PM

+ Eli (heads-up, this is the last patch that's needed to make deferred regions look/behave nicely. If you're using ToT llvm-cov and start noticing strange region highlighting, where semi-colons at the ends of a line get weird highlights, this will fix it)

vsk added a comment.Aug 4 2017, 5:37 PM

Unfortunately, no, is a separate problem in llvm-cov exposed by cfe/r310010. Clang is producing reasonable mappings, but llvm-cov stops being able to find line segments after a certain point, and keeps printing out the line execution count from the last line segment it encountered. I reverted the clang change for now so I can look into what's happening.

This revision was automatically updated to reflect the committed changes.