develooper Front page | perl.perl5.porters | Postings from December 2017

Re: [perl #74864] test suite failure on powerpc with -Duselongdouble

Thread Previous
From:
sisyphus1
Date:
December 11, 2017 06:40
Subject:
Re: [perl #74864] test suite failure on powerpc with -Duselongdouble
Message ID:
7BA855EC18274334961801C417177210@OwnerPC311012


-----Original Message----- 
From: Zefram
Sent: Monday, December 11, 2017 12:10 PM
To: perl5-porters@perl.org
Subject: Re: [perl #74864] test suite failure on powerpc 
with -Duselongdouble

> Can someone offer me an account on one of these PowerPC Linux systems with 
> the suspect long double?  Having never experienced a double-double system 
> before, I'd like to experiment to see exactly where the numeric semantics 
> break down, and thus how to efficiently detect the problem.

I have such a system - though I'm unable to provide you with access to it.

For some additional info on the M::BI failures, see:
https://rt.perl.org/Public/Bug/Display.html?id=120758

I've been consistently seeing those failures, right through to the present 
(5.27.6).
They arise solely because the value 1e+129 is miscalculated.
Replace it with a value that gets assigned correctly, and those tests pass.

Until recently, those have been the only tests that were failing. (I think 
Jarkko provided some fixes for the double-double builds a year or two ago.)
Currently, however, there is one (and only one)  *other* test file that is 
failing - namely dist/Time-HiRes/t/utime.t.

With perl-5.27.6, I get:

#######################################
$ perl t/utime.t

# Testing for subsecond file timestamps (mtime) in 
/home/sisyphus-sis/comp/perl-5.27.6/dist/Time-HiRes
# Subsecond file timestamps in 
/home/sisyphus-sis/comp/perl-5.27.6/dist/Time-HiRes: OK
1..18
# I am the main process 3288, starting the watchdog process...
# The watchdog process 3289 launched, continuing testing...
# I am the watchdog process 3289, sleeping for 360 seconds...
# $^O = linux, atime = 1.111111111, mtime = 2.222222222
# utime $fh
ok 1 - One file changed
not ok 2 - atime set correctly
#   Failed test 'atime set correctly'
#   at t/utime.t line 76.
#          got: '1.11111111'
#     expected: '1.111111111'
not ok 3 - mtime set correctly
#   Failed test 'mtime set correctly'
#   at t/utime.t line 77.
#          got: '2.222222221'
#     expected: '2.222222222'
#utime $filename
ok 4 - One file changed
not ok 5 - atime set correctly
#   Failed test 'atime set correctly'
#   at t/utime.t line 85.
#          got: '1.11111111'
#     expected: '1.111111111'
not ok 6 - mtime set correctly
#   Failed test 'mtime set correctly'
#   at t/utime.t line 86.
#          got: '2.222222221'
#     expected: '2.222222222'
utime $filename and $fh
ok 7 - Two files changed
not ok 8 - File 1 atime set correctly
#   Failed test 'File 1 atime set correctly'
#   at t/utime.t line 96.
#          got: '1.11111111'
#     expected: '1.111111111'
not ok 9 - File 1 mtime set correctly
#   Failed test 'File 1 mtime set correctly'
#   at t/utime.t line 97.
#          got: '2.222222221'
#     expected: '2.222222222'
not ok 10 - File 2 atime set correctly
#   Failed test 'File 2 atime set correctly'
#   at t/utime.t line 101.
#          got: '1.11111111'
#     expected: '1.111111111'
not ok 11 - File 2 mtime set correctly
#   Failed test 'File 2 mtime set correctly'
#   at t/utime.t line 102.
#          got: '2.222222221'
#     expected: '2.222222222'
# utime undef sets time to now
ok 12 - Two files changed
ok 13 - File 1 atime set correctly
ok 14 - File 1 mtime set correctly
ok 15 - File 2 atime set correctly
ok 16 - File 2 mtime set correctly
# negative atime dies
ok 17 - negative time error
# negative mtime dies;
ok 18 - negative time error
# I am the main process 3288, terminating the watchdog process 3289 before 
it terminates me in 359 seconds (testing took 1 seconds).
# kill KILL 3289 = 1
# All done.
# Looks like you failed 8 tests of 18.
#######################################

I have neither investigated nor filed a report for those utime.t failures. 
(I'm happy to file a separate report for that ... I've just been waiting for 
the tuits.)

Cheers,
Rob

Thread Previous


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About