`R CMD check --as-cran` causes "No output has been received..." in master but works in topic branch. How do I add `travis_wait` to it?


I am trying to run this build, which succeeds on a topic branch, but not on the master branch for some reason. I tried to find ways to speed up the build, but I haven’t been able to find any way to speed up the R CMD check --as-cran part, which seems to be the main culprit.

The reason why I tried to customize the build in the first place was to be able to add travis_wait somewhere. Originally, the build looked like this build , just the additional apt packages and using stock options, but I don’t know where to insert travis_wait in that scenario.

The other issue is that a build with the same config succeeds in one branch, but not the other, and I don’t see where/how they might differ (maybe some branch specific caching I am missing?). So for example this build and this build have the same config, but 147 (which is on master) fails. I do understand that the deploy behavior would be different, but I don’t see from the build log that the deploy phase is the part where it breaks down.

I would be grateful for any suggestions. Thank you.

In the latest build as of now, the error is rather a missing C header which should be installed via an appropriate apt package.

Thank you for the response. That is even weirder because I did have those C header files in the config.. Is it possible that the apt_packages key is not run in order (so those packages are not dowloaded?) or I have some other error in the yaml file?

Oh, I see. That’s because you override the install: step where the apt_packages key is currently handled.

I already suggested a fix for this unintended effect in BiocCheck installation cannot install scripts; thus "R CMD BiocCheck" fails .


Looking at your commands, you don’t really need the before_install: and install: keys in your configuration because the stock logic would already do all that.

In other words, a fix for you will be to remove them.

You need to overwrite the R CMD check invocation which means you need to override the script: section rather than install:.

Look for other differences (diffing the raw logs with WinMerge proved helpful for me). Package cache is the primary suspect: the tests may be hanging with some older package versions.

1 Like

Oh I hadn’t tried that, I thought once there was customization I had to include both script and install. Thank you, I will report back.

It worked!! Thank you very much!