| Summary: | =dev-libs/ppl-0.12.1-r1 on HPPA fails polyhedron1 and polyhedron2 tests | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Jeroen Roovers (RETIRED) <jer> |
| Component: | [OLD] Core system | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
| Status: | RESOLVED OBSOLETE | ||
| Severity: | normal | CC: | hppa |
| Priority: | Normal | Keywords: | PMASKED, TESTFAILURE |
| Version: | unspecified | ||
| Hardware: | HPPA | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
dev-libs:ppl-0.12.1-r1:20130508-135340.log.gz
test output |
||
|
Description
Jeroen Roovers (RETIRED)
2013-05-10 16:02:39 UTC
Are there any test logs? @toolchain, fyi on amd64 the tests are fine. (In reply to comment #2) > @toolchain, fyi on amd64 the tests are fine. The point is, can you reproduce it on PARISC? Still waitin on those logs. (In reply to comment #4) > Still waitin on those logs. What logs? Does the ppl test suite produce logs on your system? Hmm I guess not. Can you try setting PPL_NOISY_TESTS (or PPL_VERY_NOISY_TESTS) in the env? karsten /var/tmp/portage/dev-libs/ppl-0.12.1-r1/work/ppl-0.12.1/tests/Concrete_Expression # export PPL_NOISY_TESTS=yesplz
karsten /var/tmp/portage/dev-libs/ppl-0.12.1-r1/work/ppl-0.12.1/tests/Concrete_Expression # ./polyhedron1
=== test01 ===
*** ph.affine_form_image(B, 3) ***
C = 0,
A = -2,
-B >= -3,
B >= 3.
*** know_result ***
C = 0,
A = -2,
B = 3.
=== test02 ===
*** ph.affine_form_image(A, A + B + 1) ***
B >= 0,
A - 2*B >= 1,
-A + B >= -4.
*** known_result ***
A - 2*B >= 1,
B >= 0,
-A + B >= -4.
=== test03 ===
*** ph.affine_form_image(A, (A + 1) / 2) ***
2*A >= 1,
B >= 0.
*** known_result ***
2*A >= 1,
B >= 0.
=== test04 ===
*** ph.affine_form_image(A, (B + 2) / (-3)) ***
0 = 1.
*** known_result ***
B = 0,
3*A = -2.
=== test05 ===
*** ph.affine_form_image(B, (A - B + 2) / (-3)) ***
0 = 1.
*** known_result ***
A >= 2,
A + 3*B >= -1,
-A >= -3,
A - 3*B >= 2.
=== test06 ===
*** ph.affine_form_image(A, (-A - 1) / (-1)) ***
A >= 1,
B >= 0,
-A >= -2,
-B >= -1.
*** known_result ***
A >= 1,
B >= 0,
-A >= -2,
-B >= -1.
=== test07 ===
*** p1.affine_form_image(B, A) ***
A - B >= 0,
-A > 0,
-A + B >= 0.
*** known_result1 ***
A - B >= 0,
-A > 0,
-A + B >= 0.
*** p2.affine_form_image(B, A + 2) ***
A >= 0,
A - B >= -2,
-A + B >= 2.
*** known_result2 ***
A >= 0,
A - B >= -2,
-A + B >= 2.
tests failed: test04 test05
karsten /var/tmp/portage/dev-libs/ppl-0.12.1-r1/work/ppl-0.12.1/tests/Concrete_Expression # ./polyhedron2 === test01 === incompatible dimensions. === test02 === *** A <= [57, 57] *** -A >= -57. *** known_result *** -A >= -57. === test03 === *** ph *** -A >= -2, B >= 1, A >= 0, -C >= 0, C >= 0. *** known_result *** A >= 0, -A >= -2, B >= 1, C = 0. === test04 === *** ph *** 0 = 1. *** known_result1 *** 3*A >= 1, -3*A >= -2, B >= 0, -3*B >= -1. *** FP_Interval_Abstract_Store *** A = [] B = [] === test05 === *** ph *** -A >= -1, -B >= -1, A >= -1, B >= -1, -A - B >= 2. *** known_result1 *** -2*A + 4*B >= -3, A >= -1, -A >= -1, B >= -1, -B >= -1. *** FP_Interval_Abstract_Store *** A = -1 B = -1 tests failed: test04 test05 Created attachment 348518 [details]
test output
Of course copy/paste somehow bungled the line endings. :-\
This seems to actually affect sys-devel/gcc[graphite] in that it miscompiles certain loops. USE=graphite is now use.masked for sys-devel/gcc on HPPA. Thanks. Sorry I lost track of this one. Seems upstream isn't interested in older versions. No problem. I'm dying to see if ppl-1.0 fixes the stack direction problems, but I doubt it will. the 0.12.1 track is dead and no one cares and nothing uses it, so we can just let it go quietly into the night |