Greetings,
The PolyML 5.7 build has been failing seemingly randomly on the Fedora builders. On one run it fails on the ppc64le build; on the next, ppc64le succeeds but the 32-bit ARM build fails. The part of the build where the failure happens varies from run to run, as well. The builder always reports that polyml exited with error code 1.
I looked in the code. There are several exit(1) calls in there. That leads to my first request. Would you, please, give the different reasons why polyml might exit different exit codes? That would make tracking down problems like this easier.
That's essentially what I did. I patched the polyml code to differentiate, and found that these build failures stem from the crowbar code in libpolyml/processes.cpp. I'm sure that 20 seconds is more than adequate for a typical desktop environment, but it is apparently often not enough on a busy build box with multiple compile jobs running simultaneously. My second request is: would you please increase that timeout? Since the builds almost always fail on only a single architecture, and a different one each time at that, I think that 20 seconds is *almost* enough. Perhaps 30 seconds would be enough to avoid this kind of build failure.
Thank you and regards,