Causing a lot of false-negatives, here are a few examples:
I was seeing breakages when there was a new file added. Fujii suggested core.autocrlf should be false so I went ahead and did that to the bots earlier today.
I'll keep monitoring that to see if that helps any.
The other thing I'm seeing is that libpng's version isn't being reported properly in some builds. It seems to think its 1.6.29 despite the header clearly being 1.6.34. I'm not sure why this is happening there as I haven't seen it on our buildbot.
Did you try to delete CMakeCache.txt?
> -- Found ICU data libraries:
> CMake Error at C:/tools/cmake/share/cmake-3.9/Modules/FindPackageHandleStandardArgs.cmake:137 (message):
> Could NOT find PNG: Found unsuitable version "1.6.29", but required is at
> least "1.6.34" (found
(In reply to Fujii Hironori from comment #3)
> Did you try to delete CMakeCache.txt?
> > -- Found ICU data libraries:
> > CMake Error at C:/tools/cmake/share/cmake-3.9/Modules/FindPackageHandleStandardArgs.cmake:137 (message):
> > Could NOT find PNG: Found unsuitable version "1.6.29", but required is at
> > least "1.6.34" (found
> > C:/WebKit-EWS/WebKit/WebKitLibraries/win/lib64/libpng16.lib)
The bots should be doing a clean build on each 10 iterations. I also looked at the header files and verified it was a correct version. I'm really not sure what the root cause of it is...
Looking further there is a libpng 1.6.29 within perl's directories.
I'm not sure why EWS would be picking this header up but then finding the library in lib64.
My only thought here is maybe its something to do with https://bugs.webkit.org/show_bug.cgi?id=180137 landing which started detecting perl modules. I'm not sure if that might cause some paths to get added to CMake which would then cause this particular issue.
Its detecting multiple libraries within the perl c directory. I think the fix is to see if within perl there is a c/include directory then we need to use CMAKE_IGNORE_PATH to make sure that nothing appears from there.
Ok there are even more problems...
webkit-patch appears to call update-webkit which will invoke update-webkit-auxiliary-libs. This will overwrite what was in WinCairoRequirements.
Ok EWS has had a couple successes after the dependent patches landed. I'm spinning up the other bots for a total of 4 running things.
If things look like they've stabilized then I'll close this.
Its been behaving for a few weeks now so hopefully all is resolved.