[bind] Intel/Linux failures
Peter, Could you please comment on "bind_cv_test" and "bind_stateful_test" tests' failures on all Intel compilers on Linux -- http://tinyurl.com/5p3or? Should we mark these up as expected? -- Aleksey Gurtovoy MetaCommunications Engineering
Aleksey Gurtovoy wrote:
Peter,
Could you please comment on "bind_cv_test" and "bind_stateful_test" tests' failures on all Intel compilers on Linux -- http://tinyurl.com/5p3or?
Seemed like a compiler bug to me (wrong overload being called) last time I looked at the logs. If someone'd like to investigate further, I'd be grateful.
Should we mark these up as expected?
"Expected" turns red failures into yellow failures, right? If so, and the failure is indeed a compiler bug, then yes, we should mark them as expected.
Peter Dimov writes:
Aleksey Gurtovoy wrote:
Peter,
Could you please comment on "bind_cv_test" and "bind_stateful_test" tests' failures on all Intel compilers on Linux -- http://tinyurl.com/5p3or?
Seemed like a compiler bug to me (wrong overload being called) last time I looked at the logs. If someone'd like to investigate further, I'd be grateful.
Anyone?
Should we mark these up as expected?
"Expected" turns red failures into yellow failures, right?
Green, actually, but with a link to a corresponding note/explanation.
If so, and the failure is indeed a compiler bug,
That's basically what I wanted to know -- is it?
then yes, we should mark them as expected.
OK, unless somebody proves otherwise, I'm going to mark the failures as expected with a comment saying that it's a compiler bug. -- Aleksey Gurtovoy MetaCommunications Engineering
Aleksey Gurtovoy wrote:
Peter Dimov writes:
"Expected" turns red failures into yellow failures, right?
Green, actually, but with a link to a corresponding note/explanation.
Can we make/leave them yellow? :-) The tests fail, after all. How can a failure be green? Maybe we need a way to mark a test "yellow failure, does not affect overall greenness".
If so, and the failure is indeed a compiler bug,
That's basically what I wanted to know -- is it?
I don't know, but the two tests do pass on 8.0 based on the report. Only 7.1 fails. The bind_cv_test failure sure looks like a compiler bug (or an artifact of some bug compatibility mode) because it seems that the compiler invokes the non-const operator() on a const function object.
Peter Dimov writes:
Aleksey Gurtovoy wrote:
Peter Dimov writes:
"Expected" turns red failures into yellow failures, right?
Green, actually, but with a link to a corresponding note/explanation.
Can we make/leave them yellow? :-)
They are going to be yellow on the user-level library report.
The tests fail, after all. How can a failure be green?
Well, it still says "fail" :).
Maybe we need a way to mark a test "yellow failure, does not affect overall greenness".
That's basically the current meaning of the "expected failures" markup; the expected failures rendered as green on the developer report because basically, once the failures are marked, they become of the same interest to you as the passing tests -- that is, of no interest.
If so, and the failure is indeed a compiler bug,
That's basically what I wanted to know -- is it?
I don't know, but the two tests do pass on 8.0 based on the report. Only 7.1 fails. The bind_cv_test failure sure looks like a compiler bug (or an artifact of some bug compatibility mode) because it seems that the compiler invokes the non-const operator() on a const function object.
OK, thanks! -- Aleksey Gurtovoy MetaCommunications Engineering
participants (2)
-
Aleksey Gurtovoy -
Peter Dimov