On 28.12.2017 18:01, Sébastien Villemot wrote: > Note that octave-interval FTBFS on armel: > > https://buildd.debian.org/status/fetch.php?pkg=octave-interval&arch=armel&ver=3.1.0-1&stamp=1514325840&raw=0 > > There are two options: the preferred one is of course to fix the build; the > alternative is to request the removal of the binary on armel. The issue with armel is, that floating-point rounding-mode switches are not working. This is not a problem for the package, since it does not depend on rounding mode switches to work. The broken function is used for *fast* matrix multiplication only and the package detects at runtime when it is not working. You can find the following warning in several test cases on armel (also for older versions of the package): warning: mtimes: rounding modes not supported, falling back to slow but accurate matrix multiplication From the test logs I can tell that the package is working on armel, but matrix multiplication might be slower than expected. Thus, it should not be necessary to request removal on that architecture. However, I'm not sure how this should be handled. I guess that a global floating-point rounding mode is not available on that platform. Should the test simply be ignored / skipped for the particular platform? If yes, how? With an arch-specific patch? Oliver
Attachment:
signature.asc
Description: OpenPGP digital signature