https://github.com/JuliaCI/Coverage.jl/issues/187#issuecomment-470198083 #1163
Conversation
Codecov Report
@@ Coverage Diff @@
## master #1163 +/- ##
==========================================
+ Coverage 99.11% 99.87% +0.76%
==========================================
Files 92 92
Lines 4181 4149 -32
==========================================
Hits 4144 4144
+ Misses 37 5 -32 |
|
This won't work if you only do it for Travis but not AppVeyor, as the results of the two are merged. |
|
Aha, but it seems that while you have an |
|
Aha, but you did not actually ensure that it uses the latest Coverage.jl master branch -- so it is still using Coverge.jl 0.8.0, which does not know about |
|
In fact, not even the master branch is enough, you'll need the code from that PR, resp. from my |
|
@fingolfin I just reran travis w/ codecov - does this look correct to you? |
|
No... hmm, the coverage displayed by codecov for some reason meshes the results of two Travis build with two jobs each, see This is because once the branch is built as a PR, and once as branch of your repo. Now, looking at the build logs, three of the four jobs used Coverage.jl 0.9.0, but one used 0.7.0, and that is tainting the results. Perhaps a cache for installed packages was used and not properly reset on some of the builders?!? Anyway, try restarting |
|
@fingolfin restarted.... |
|
Both Julia 0.7 jobs are using Coverage.jl 0.7.0 (which is three versions behind); hmm, I just realize this is way behind, as 0.8.0 and 0.8.1 had been released some time ago. Perhaps try disabling Julia 0.7 for the sake of testing this? |
|
yup! see #1167. It fixed things. |
|
I'm pretty sure we don't want this since it won't count functions that aren't executed at all. @fingolfin - do I have this right? |
|
Yes, agreed |
No description provided.