Skip Menu |

This queue is for tickets about the Math-FractionManip CPAN distribution.

Report information
The Basics
Id: 132402
Status: resolved
Worked: 1.6 hours (95 min)
Priority: 0/
Queue: Math-FractionManip

People
Owner: gene [...] cpan.org
Requestors: SREZIC [...] cpan.org
Cc:
AdminCc:

Bug Information
Severity: (no value)
Broken in: 0.5403
Fixed in: 0.5503



Subject: t/01-methods.t fails (with -Duselongdouble)
If perl was compiled with increased floating point accuracy (i.e. with the -Duselongdouble configuration option), then the test suite fails: ... # Failed test 'exp' # at t/01-methods.t line 79. # got: 'NaN/NaN' # expected: '229739670999407/373719281884655' # Failed test 'exp' # at t/01-methods.t line 80. # got: '0.614738607654485177' # expected: '0.614738607654485' # Failed test 'exp' # at t/01-methods.t line 82. # got: 'NaN/NaN' # expected: '70710678118655/86602540378444' # Failed test 'sqrt' # at t/01-methods.t line 83. # got: 'NaN/NaN' # expected: '70710678118655/86602540378444' # Looks like you failed 4 tests of 66. t/01-methods.t ........... Dubious, test returned 4 (wstat 1024, 0x400) Failed 4/66 subtests ...
Weird. On my "perl 5, version 31, subversion 5 (v5.31.5) built for x86_64-linux-ld" with -Duselongdouble Ubuntu box this test passes. Thinking about how to wrangle this one. :\
Aha. I think I just need to force using Math::BigFloat. Trying that.
Ok. I went and installed a -Duselongdouble perl v5.30.2, and the test now fails as in the report. On Thu Apr 23 12:01:47 2020, GENE wrote: Show quoted text
> Weird. On my "perl 5, version 31, subversion 5 (v5.31.5) built for > x86_64-linux-ld" with -Duselongdouble Ubuntu box this test passes.
Ok. Hopefully the new version 0.5500 will fix things up. I removed code that was being fired but appeared broken. Also I re-worked the ** and sqrt tests to not check for an exact fraction.
On 2020-04-23 20:57:08, GENE wrote: Show quoted text
> Ok. Hopefully the new version 0.5500 will fix things up. I removed > code that was being fired but appeared broken. Also I re-worked the > ** and sqrt tests to not check for an exact fraction.
0.5500 looks better. But there's still one smoker system where the test suite fails (freebsd 10 with perl 5.20.1 + -Duselongdouble): ... # Failed test 'exp numerator' # at t/01-methods.t line 94. # got: '141421356237309505' # expected: '141421356237309515' # Failed test 'exp denominator' # at t/01-methods.t line 95. # got: '173205080756887729' # expected: '173205080756887719' # Looks like you failed 2 tests of 71. t/01-methods.t ........... Dubious, test returned 2 (wstat 512, 0x200) Failed 2/71 subtests ... A full report will be available within some hours on CPAN Testers.
I had added 2 tests that were looking for a specific float... Naughty! I have contsrained them with sprintf(). Version 0.5502 should fix things up...
Ok. I have fixed things once again with 0.5503. This time by just /removing/ the two failing tests, in which I was trying to compare ** and sqrt operation values. All should be well with this resurrected module now.
On 2020-04-25 16:20:43, GENE wrote: Show quoted text
> Ok. I have fixed things once again with 0.5503. This time by just > /removing/ the two failing tests, in which I was trying to compare ** > and sqrt operation values. All should be well with this resurrected > module now.
Confirmed. I don't see any failures anymore.