Ticket #1813 (closed bug: done)
t/op/infnan.t: failures on Darwin/PPC
Reported by: | jkeenan | Owned by: | plobsing |
---|---|---|---|
Priority: | normal | Milestone: | |
Component: | core | Version: | 2.8.0 |
Severity: | medium | Keywords: | |
Cc: | Language: | ||
Patch status: | Platform: | darwin |
Description
Somewhere between r49383 and r49432, we began to experience failures in t/op/infnan.t.
$ prove -v t/op/inf_nan.t t/op/inf_nan.t .. 1..37 not ok 1 - basic arithmetic: = # Have: inf # Want: Inf ok 2 - ... -= ok 3 - ... *= -1 ok 4 - ... *= 0 ok 5 - ... += 5 ok 6 - ... -= 42 ok 7 - ... inc ok 8 - ... dec ok 9 - ... abs NaN not ok 10 - ... abs Inf # Have: inf # Want: Inf not ok 11 - ... abs -Inf # Have: inf # Want: Inf not ok 12 - sqrt: assignment # Have: inf # Want: Inf ok 13 - ... sqrt -Inf ok 14 - ... sqrt NaN ok 15 - ... sqrt -1 not ok 16 - negative: neg Inf # Have: -inf # Want: -Inf not ok 17 - ... neg -Inf # Have: inf # Want: Inf ok 18 - ... neg NaN ok 19 - mixing NaN and Inf: NaN * Inf ok 20 - ... NaN / Inf ok 21 - ... NaN - Inf ok 22 - ... NaN + Inf ok 23 - rounding n: floor NaN ok 24 - ... ceil NaN not ok 25 - ... floor Inf # Have: inf # Want: Inf not ok 26 - ... ceil Inf # Have: inf # Want: Inf not ok 27 - ... floor -Inf # Have: -inf # Want: -Inf not ok 28 - ... ceil -Inf # Have: -inf # Want: -Inf ok 29 #skip rounding nan/inf gives something like -2147483648 ok 30 #skip rounding nan/inf gives something like -2147483648 ok 31 #skip rounding nan/inf gives something like -2147483648 ok 32 #skip rounding nan/inf gives something like -2147483648 ok 33 #skip 1+i + NaN should be NaN ok 34 #skip fdiv/mod do not play nicely with PMCs and NaN ok 35 #skip fdiv/mod do not play nicely with PMCs and NaN ok 36 #skip fdiv/mod do not play nicely with PMCs and NaN ok 37 #skip fdiv/mod do not play nicely with PMCs and NaN Failed 10/37 subtests (less 9 skipped subtests: 18 okay) Test Summary Report ------------------- t/op/inf_nan.t (Wstat: 0 Tests: 37 Failed: 10) Failed tests: 1, 10-12, 16-17, 25-28 Files=1, Tests=37, 0 wallclock secs ( 0.14 usr 0.04 sys + 0.06 cusr 0.04 csys = 0.28 CPU) Result: FAIL
Since a new configuration step, auto::infnan, was added at r49425, I suspect that was the revision where these test failures emerged.
Note: AFAICT there was no discussion on list about adding this new configuration step, nor was there a Trac ticket presenting a rationale for this new configuration step. The new step was committed directly to trunk without prior development in a branch. I thought we had gotten away from this style of work.
This is not a good development practice, as it prevents Parrot developers other than the committer from:
(a) evaluating the rationale for the new step -- is it needed at all? is this the best approach?
(b) testing the code on our normal set of platforms.
Thank you very much.
kid51