Patrice Dumas wrote: > I do not have a good understanding, actually, whether the CI > system we use is transparent in that it could be reproduced somewhere > else or not.
Generally, in 90% or 95% of the cases, I can reproduce CI failures locally on VMs that run on my own hardware or on compilefarm machines. In this case I couldn't, probably because the /bin/sh bug on this OS is specific to an OS version or is sensitive to memory addresses that arise at runtime. > Also there is the SaaSS issue Yes. In the worst case, GitHub changes their terms and conditions, and we cannot use it any more. Then we'll have to move somewhere else, which means that our investment (in terms of CI configuration tweaks) is lost. But the benefits that I got from this CI, for the Gnulib project alone, in 5 months, already exceed the investment that I made. Gavin Smith wrote: > If people want to use Github to run tests of Texinfo, that is up to them. > They are not discouraged from doing so and reports from such tests are > welcome. Yep, that's what I'm doing. Without an automated CI, I could not afford the time to test every texinfo prerelease; I would be able to do so only once every 2 or 3 years. Bruno
