control: severity -1 serious Hi,
On 2025-02-15 11:06, Aurelien Jarno wrote: > Source: bart-cuda > Version: 0.8.00-2 > Severity: important > Tags: ftbfs patch > Justification: fails to build from source > X-Debbugs-Cc: debian-gl...@lists.debian.org > User: debian-gl...@lists.debian.org > Usertags: glibc2.41 > > Dear maintainer, > > The bart-cuda autopkgtest fails when run against glibc 2.41, which is > currently in experiemntal. It fails with [1]: > > | 3772s touch tests/test-traj-custom > | 3772s set -e ; mkdir > /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp/tests/tmp/$$/ ; cd > /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp/tests/tmp/$$/ > ;\ > | 3772s /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp//traj -R0. -r > -y360 -D t0.ra ;\ > | 3772s /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp//phantom -k -t > t0.ra k0.ra ;\ > | 3772s /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp//traj -R30. -r > -y360 -D t30.ra ;\ > | 3772s /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp//phantom -k -t > t30.ra k30.ra ;\ > | 3772s /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp//estshift 4 > k0.ra k30.ra | grep "30.00000" ;\ > | 3772s rm *.ra ; cd .. ; rmdir > /tmp/autopkgtest-lxc.oedkyjwg/downtmp/autopkgtest_tmp/tests/tmp/$$/ > | 3774s make: *** [tests/traj.mk:53: tests/test-traj-rot] Error 1 > > Investigation shows that this is due to improved accuracy of the atan2f > function in glibc 2.41. In details bart-cuda computes > atan2f(0x1.166202p-4, 0x1.e22c5ap-4) and it expects it to be 30.00000 > when converted from radian to degrees, ie it expects the atan2f function > to return 0x1.0c1524p-1. > > With glibc 2.41, the atan2f function has been improved [2] and now > returns instead 0x1.0c1522p-1 which is closer to the real value than > 0x1.0c1524p-1. This corresponds to 29.999998 when converted to degrees. > The patch below fixes the issue by accepting both values. > glibc 2.41 is now in unstable, bumping the severity accordingly. Regards Aurelien -- Aurelien Jarno GPG: 4096R/1DDD8C9B aurel...@aurel32.net http://aurel32.net
