ruby-gsl dep8 fails on ppc64el
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
gsl (Ubuntu) |
Fix Released
|
High
|
Sergio Durigan Junior |
Bug Description
With the latest update of gsl from Debian, ruby-gsl's dep8 test start failing on ppc64el:
421s =======
421s Failure: test_random32_
421s random32-bsd, 10000 steps (852261210 observed vs 1663114331 expected).
421s <false> is not true.
421s /tmp/autopkgtes
421s =======
421s : (0.007176)
421s test_random32_
421s test_random32_
421s =======
421s Failure: test_random32_
421s random32-libc5, 10000 steps (367802360 observed vs 1967452027 expected).
421s <false> is not true.
421s /tmp/autopkgtes
421s =======
421s : (0.002615)
421s test_random64_
421s =======
421s Failure: test_random64_
421s random64-bsd, 10000 steps (210970120 observed vs 864469165 expected).
421s <false> is not true.
421s /tmp/autopkgtes
421s =======
421s : (0.002551)
421s test_random64_
421s test_random64_
421s =======
421s Failure: test_random64_
421s random64-libc5, 10000 steps (221021662 observed vs 2106639801 expected).
421s <false> is not true.
421s /tmp/autopkgtes
421s =======
421s : (0.002627)
This looks like a precision problem caused by excessive optimization in the architecture.
I built a version of gsl using -O2 instead of -O3 on ppc64el and confirmed that the problem went away.