=?ISO-8859-1?Q?Frank Strau=DF?= writes:
Frank> Can anyone else on this list confirm any of our - good or Frank> failing - observations on the last test of the "make check" Frank> test suite in 0.4.2?
I just compiled on "SunOS thesun 5.8 Generic_108528-15 sun4u sparc SUNW,Sun-Fire-280R" and I also see that a several tests fail. It turns out that the order of some of the identifiers is different on these platforms. For example, smidump -f metrics on Debian Gnu/Linux produces something like
MODULE TYPE EXT-USAGE Integer32 31.6% SNMPv2-SMI Counter32 31.6% SNMPv2-TC AutonomousType 15.8% IF-MIB InterfaceIndex 10.5% SNMPv2-SMI Counter64 5.3% SNMPv2-TC TruthValue 5.3%
while the Solaris box produces
MODULE TYPE EXT-USAGE Integer32 31.6% SNMPv2-SMI Counter32 31.6% SNMPv2-TC AutonomousType 15.8% IF-MIB InterfaceIndex 10.5% SNMPv2-TC TruthValue 5.3% SNMPv2-SMI Counter64 5.3%
Note that the last two lines are swapped and that this list is sorted by the percentages. The code just calls qsort() and I guess this is why see these differences since qsort implementations differ in how they order identical values...
/js