_1basic.t and constants.t fails on Arch Linux about math-mpfr HOT 2 CLOSED

I've not seen this problem before, despite the large number of successful tests run on Linux by the cpan-testers - other distros apparently ensure that mpfr_get_version() and MPFR_VERSION_STRING match.
I'm actually thinking that someone forgot to edit the MPFR_VERSION_STRING entry in mpfr.h.
Would you be able to check with them as to whether that's the case ?

I'm a bit loathe to remove the suffix from the test for two reasons:

1. I think mpfr_get_version() and MPFR_VERSION_STRING should match exactly;
2. we would then fail to detect a discrepancy between 3.1.3-p5 and (say) 3.1.3-p6.

Perhaps the second point doesn't really apply ... perhaps 3.1.3-p5 headers are guaranteed to be compatible with the 3.1.3-p6 library ... I don't know.

If they insist on distributing mpfr packages where header and library can disagree on the version string, then I guess I'll have to do as you suggest - but only for those machines that might have installed the ArchLinux distros.

On Windows, Ubuntu and Debian, I build Math::MPFR against svn trunk, where both Rmpfr_get_version() and MPFR_VERSION_STRING() report "3.2.0-dev".

Cheers,
Rob

from math-mpfr.

It was my mistake. I apologize for it. Somehow, I had an old mpfr.h file in /usr/local/include/ which took precedence over /usr/include/mpfr.h. By removing it, Math::MPFR passes all the tests.

Thank you for your patience.

from math-mpfr.