Conversation
|
Hmm any obvious reason why the CI hasn't triggered here? |
|
It failed parsing YAML. I got a little ahead of myself on the syntax after finding an example online :) |
|
Finally. All follow up commits are just fixing mistakes in the initial one, we don't need to preserve their messages if testing succeeds. I am assigning 5 parallel tasks to the CMake job and 5 to MSVC. We can try to combine them into one big job, it should yield faster builds, but at expense of readability. |
|
For now, I am going to give 6 jobs to CMake (covering all needed) and the remaining 4 to Visual Studio, this should work as VS builds are much faster. I am also going to rebase all changes together. |
Add MSVC CI jobs, testing on latest Windows VM image using build and test wrappers. Assign 6 jobs to 'nix CI (all it needs) and the remaining 4 - to MCVS, as it is much faster. For #6547
|
Looking at the speed comparisons - those windows builds are crazy fast. Stand out differences:
|
Use the pattern in Windows CI to create two separate steps for build and test. For #6547
173fede to
ed068f0
Compare
|
Maybe we should start with hiding the output - it produces enormous amount of text. If rl.exe cap number of concurrent tests it might also be the reason why it is faster. |
Testing offline the excess output vanishes if runtests.py is called without a --logfile parameter - I don't want to open a PR because I'll delay your updates but two changes to try updating test/CMakeLists.txt with:
Edit: tried this and it didn't work. BUT a slight tweak to runtests.py did do it - see #6559 |
|
@Fly-Style I am going to merge, but will definitely address any feedback in a follow-up PR. |
Add MSVC CI jobs, testing on latest Windows VM image using provided build and test wrappers.
For #6547