Bastian Germann wrote: > The dpkg-architecture call in the multiarch+cross patch can have the > side effect of printing a warning.
Thanks for the report. This was completely intentional; dpkg prints both warnings and errors to stderr and in case of an error, we would like to see dpkg's error message before the process aborts. > I have identified this issue via > https://github.com/markokr/rarfile/issues/104 This is the same issue as unar's zip autopkgtest which I changed to allow-stderr as a workaround, as I couldn't reproduce. My very wild guess is that the CI runner is a not so pure Debian system or a system with some odd/non-standard setup. Note that the GNUstep Base Library can output a lot of stuff (warnings, exceptions, deprecation notices, etc) so if your program is relying on it being silent, that is a wrong assumption. > Please make sure that all of the execs in the patch do not randomly > influence gnustep programs. I don't think it is a good idea to redirect stderr to /dev/null because in case of dpkg-architecture failure it would be very hard to analyze what exactly happened.
