Hi! I recently uploaded an update to the globus-gass-copy package to address a FTBFS caused by the update of openssl to version 1.1.1. The new openssl version is stricter than the old one and requires a minimum of 2048 bits RSA keys. The reason for the FTBFS in globus-gass- copy was that the test suite was using 1024 bit keys when running tests. So the tests started failing with the new openssl due to the key being to short. The fix was a simple one-line patch changing the size of the key generated for the test. However, the tests run a lot of TLS operations and increasing the key size means that the tests use more cpu cycles and take longer time. For most of the architectures this is not problematic, the tests runs fast enough, but for two of the architectures increasing the key size meant that the tests ran out of time and were killed: mips: E: Build killed with signal TERM after 150 minutes of inactivity hppa: E: Build killed with signal TERM after 150 minutes of inactivity On the other architectures the builds completed fine. Total build time for successful architectures from https://buildd.debian.org/status/logs.php?pkg=globus-gass-copy&ver=9.28-2&suite=sid alpha 1h 7m arm64 18m armel 2h 31m armhf 1h 18m hurd-i386 18m i386 26m ia64 22m kfreebsd-amd64 6m kfreebsd-i386 15m m68k 10m mips64el 45m mipsel 2h 14m powerpc 22m powerpcspe 6m ppc64 8m ppc64el 8m riscv64 2h 16m s390x 10m sh4 9m sparc64 3h 41m x32 8m As can be seen the build completes fast on most architectures but runs a long time on some of the slower ones. I tried to run the build on the porter boxes for the two failing architectures. They completed fine after running for a long time. On mips the longest running test ran for 3h 14m (which is above the 2h 30m no-activity limit). The complete build including the tests ran for 3h 44m. Om hppa the longest running test ran for 7h 30m (which is well above the 2h 30m limit). The complete build including the tests ran for 7h 58m. I don't know if the hppa build servers are as slow as the porter box. The no-activity limit should have some scaling so that builds are allowed to take longer time on slow builders. Mattias
Attachment:
smime.p7s
Description: S/MIME cryptographic signature