test suite failure on powerpc with -Duselongdouble

[p5pRT]

Migrated from rt.perl.org#74864 (status was 'open')

Searchable as RT74864$

[p5pRT]

From @ntyni

This is a bug report for perl from Niko Tyni <ntyni@​debian.org>,
generated with the help of perlbug 1.39 running under perl 5.12.0.

---

I'm seeing 5.12.0 test failures on the Debian powerpc architecture
with -Duselongdouble -Duse64bitint (but not with just -Duse64bitint.)

Test Summary Report

---

op/inc.t (Wstat​: 65280 Tests​: 34 Failed​: 0)
  Non-zero exit status​: 255
  Parse errors​: Bad plan. You planned 54 tests but ran 34.
op/pack.t (Wstat​: 0 Tests​: 14697 Failed​: 28)
  Failed tests​: 521, 694, 749, 804, 1024, 1079, 1134, 1354
  1409, 1464, 1684, 1739, 1794, 2014, 2069
  2124, 2344, 2399, 2454, 2894, 2949, 3004
  3059, 3444, 3499, 13234, 13289, 13344
op/stash.t (Wstat​: 0 Tests​: 31 Failed​: 0)
  TODO passed​: 26
../cpan/Math-BigInt/t/bare_mbi.t (Wstat​: 0 Tests​: 3273 Failed​: 1)
  Failed test​: 3087
../cpan/Math-BigInt/t/bigintpm.t (Wstat​: 0 Tests​: 3279 Failed​: 1)
  Failed test​: 3093
../cpan/Math-BigInt/t/sub_mbi.t (Wstat​: 0 Tests​: 3278 Failed​: 1)
  Failed test​: 3087
Files=1803, Tests=357905, 12620 wallclock secs (1354.30 usr 35.30 sys + 10177.06 cusr 276.30 csys = 11842.96 CPU)
Result​: FAIL

At least op/pack.t fails on 5.10.1 too, so this is not a 5.12 regression.

I assume the 128-bit long doubles on this platform are actually a sum of
two 64-bit doubles, which "does not conform to the IEEE floating-point
standard" [1] and "has serious arithmetic anomalies" [2].

If this something that can not be supported, perhaps Configure should
detect such a platform and disable the longdouble setting or error out?

(This is also http​://bugs.debian.org/578295 )

[1] http​://en.wikipedia.org/wiki/Long_double
[2] http​://gcc.gnu.org/ml/gcc-help/2006-10/msg00138.html

---

---

Flags​:
  category=core
  severity=low

---

Site configuration information for perl 5.12.0​:

Configured by Debian Project at Sun May 2 12​:18​:53 UTC 2010.

Summary of my perl5 (revision 5 version 12 subversion 0) configuration​:
 
  Platform​:
  osname=linux, osvers=2.6.26-2-powerpc, archname=powerpc-linux-gnu-thread-multi-64int-ld
  uname='linux pescetti 2.6.26-2-powerpc #1 tue mar 9 23​:07​:18 utc 2010 ppc gnulinux '
  config_args='-Dusethreads -Duselargefiles -Dccflags=-DDEBIAN -Dcccdlflags=-fPIC -Darchname=powerpc-linux-gnu -Dprefix=/usr -Dprivlib=/usr/share/perl/5.12 -Darchlib=/usr/lib/perl/5.12 -Dvendorprefix=/usr -Dvendorlib=/usr/share/perl5 -Dvendorarch=/usr/lib/perl5 -Dsiteprefix=/usr/local -Dsitelib=/usr/local/share/perl/5.12.0 -Dsitearch=/usr/local/lib/perl/5.12.0 -Dman1dir=/usr/share/man/man1 -Dman3dir=/usr/share/man/man3 -Dsiteman1dir=/usr/local/man/man1 -Dsiteman3dir=/usr/local/man/man3 -Dusemorebits -Dman1ext=1 -Dman3ext=3perl -Dpager=/usr/bin/sensible-pager -Uafs -Ud_csh -Ud_ualarm -Uusesfio -Uusenm -DDEBUGGING=both -Doptimize=-O2 -O0 -Uuseshrplib -des'
  hint=recommended, useposix=true, d_sigaction=define
  useithreads=define, usemultiplicity=define
  useperlio=define, d_sfio=undef, uselargefiles=define, usesocks=undef
  use64bitint=define, use64bitall=undef, uselongdouble=define
  usemymalloc=n, bincompat5005=undef
  Compiler​:
  cc='cc', ccflags ='-D_REENTRANT -D_GNU_SOURCE -DDEBIAN -DDEBUGGING -fno-strict-aliasing -pipe -fstack-protector -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64',
  optimize='-O2 -O0 -g',
  cppflags='-D_REENTRANT -D_GNU_SOURCE -DDEBIAN -DDEBUGGING -fno-strict-aliasing -pipe -fstack-protector -I/usr/local/include'
  ccversion='', gccversion='4.4.3', gccosandvers=''
  intsize=4, longsize=4, ptrsize=4, doublesize=8, byteorder=87654321
  d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=16
  ivtype='long long', ivsize=8, nvtype='long double', nvsize=16, Off_t='off_t', lseeksize=8
  alignbytes=16, prototype=define
  Linker and Libraries​:
  ld='cc', ldflags =' -fstack-protector -L/usr/local/lib'
  libpth=/usr/local/lib /lib /usr/lib /lib64 /usr/lib64
  libs=-lgdbm -lgdbm_compat -ldb -ldl -lm -lpthread -lc -lcrypt
  perllibs=-ldl -lm -lpthread -lc -lcrypt
  libc=/lib/libc-2.10.2.so, so=so, useshrplib=false, libperl=libperl.a
  gnulibc_version='2.10.2'
  Dynamic Linking​:
  dlsrc=dl_dlopen.xs, dlext=so, d_dlsymun=undef, ccdlflags='-Wl,-E'
  cccdlflags='-fPIC', lddlflags='-shared -O2 -O0 -g -L/usr/local/lib -fstack-protector'

Locally applied patches​:
  DEBPKG​:debian/arm_thread_stress_timeout - http​://bugs.debian.org/501970 Raise the timeout of ext/threads/shared/t/stress.t to accommodate slower build hosts
  DEBPKG​:debian/cpan_config_path - Set location of CPAN​::Config to /etc/perl as /usr may not be writable.
  DEBPKG​:debian/cpan_definstalldirs - Provide a sensible INSTALLDIRS default for modules installed from CPAN.
  DEBPKG​:debian/db_file_ver - http​://bugs.debian.org/340047 Remove overly restrictive DB_File version check.
  DEBPKG​:debian/doc_info - Replace generic man(1) instructions with Debian-specific information.
  DEBPKG​:debian/enc2xs_inc - http​://bugs.debian.org/290336 Tweak enc2xs to follow symlinks and ignore missing @​INC directories.
  DEBPKG​:debian/errno_ver - http​://bugs.debian.org/343351 Remove Errno version check due to upgrade problems with long-running processes.
  DEBPKG​:debian/extutils_hacks - Various debian-specific ExtUtils changes
  DEBPKG​:debian/fakeroot - Postpone LD_LIBRARY_PATH evaluation to the binary targets.
  DEBPKG​:debian/instmodsh_doc - Debian policy doesn't install .packlist files for core or vendor.
  DEBPKG​:debian/ld_run_path - Remove standard libs from LD_RUN_PATH as per Debian policy.
  DEBPKG​:debian/libnet_config_path - Set location of libnet.cfg to /etc/perl/Net as /usr may not be writable.
  DEBPKG​:debian/m68k_thread_stress - http​://bugs.debian.org/495826 Disable some threads tests on m68k for now due to missing TLS.
  DEBPKG​:debian/mod_paths - Tweak @​INC ordering for Debian
  DEBPKG​:debian/module_build_man_extensions - http​://bugs.debian.org/479460 Adjust Module​::Build manual page extensions for the Debian Perl policy
  DEBPKG​:debian/prune_libs - http​://bugs.debian.org/128355 Prune the list of libraries wanted to what we actually need.
  DEBPKG​:debian/use_gdbm - Explicitly link against -lgdbm_compat in ODBM_File/NDBM_File.
  DEBPKG​:fixes/net_smtp_docs - http​://bugs.debian.org/100195 [rt.cpan.org #36038] Document the Net​::SMTP 'Port' option
  DEBPKG​:fixes/processPL - http​://bugs.debian.org/357264 [rt.cpan.org #17224] Always use PERLRUNINST when building perl modules.
  DEBPKG​:debian/perlivp - http​://bugs.debian.org/510895 Make perlivp skip include directories in /usr/local
  DEBPKG​:debian/disable-zlib-bundling - Disable zlib bundling in Compress​::Raw​::Zlib
  DEBPKG​:debian/cpanplus_definstalldirs - http​://bugs.debian.org/533707 Configure CPANPLUS to use the site directories by default.
  DEBPKG​:debian/cpanplus_config_path - Save local versions of CPANPLUS​::Config​::System into /etc/perl.
  DEBPKG​:fixes/autodie-flock - http​://bugs.debian.org/543731 Allow for flock returning EAGAIN instead of EWOULDBLOCK on linux/parisc
  DEBPKG​:debian/devel-ppport-ia64-optim - http​://bugs.debian.org/548943 Work around an ICE on ia64
  DEBPKG​:fixes/cpanplus-without-home - http​://bugs.debian.org/577011 [rt.cpan.org #52988] Fix CPANPLUS test failures when HOME doesn't exist
  DEBPKG​:patchlevel - http​://bugs.debian.org/567489 List packaged patches for 5.12.0-1 in patchlevel.h

---

@​INC for perl 5.12.0​:
  lib
  /etc/perl
  /usr/local/lib/perl/5.12.0
  /usr/local/share/perl/5.12.0
  /usr/lib/perl5
  /usr/share/perl5
  /usr/lib/perl/5.12
  /usr/share/perl/5.12
  /usr/local/lib/site_perl
  .

---

Environment for perl 5.12.0​:
  HOME=/home/ntyni
  LANG=en_US.UTF-8
  LANGUAGE (unset)
  LC_CTYPE=fi_FI.UTF-8
  LD_LIBRARY_PATH (unset)
  LOGDIR (unset)
  PATH=/usr/local/bin​:/usr/bin​:/bin​:/usr/bin/X11​:/usr/games
  PERL_BADLANG (unset)
  SHELL=/bin/zsh

[p5pRT]

From @nwc10

On Mon, May 03, 2010 at 01​:30​:34AM -0700, Niko Tyni wrote​:

I'm seeing 5.12.0 test failures on the Debian powerpc architecture
with -Duselongdouble -Duse64bitint (but not with just -Duse64bitint.)

I assume the 128-bit long doubles on this platform are actually a sum of
two 64-bit doubles, which "does not conform to the IEEE floating-point
standard" [1] and "has serious arithmetic anomalies" [2].

That seems a reasonable assumption, given that sparc64 Debian with 16 byte
long doubles does work.

If this something that can not be supported, perhaps Configure should
detect such a platform and disable the longdouble setting or error out?

That seems reasonable. Not sure how to do it.

(This is also http​://bugs.debian.org/578295 )

[1] http​://en.wikipedia.org/wiki/Long_double
[2] http​://gcc.gnu.org/ml/gcc-help/2006-10/msg00138.html

Nicholas Clark

[p5pRT]

The RT System itself - Status changed from 'new' to 'open'

[p5pRT]

From zefram@fysh.org

Niko Tyni wrote​:

I assume the 128-bit long doubles on this platform are actually a sum of
two 64-bit doubles, which "does not conform to the IEEE floating-point
standard" [1] and "has serious arithmetic anomalies" [2].

If this something that can not be supported, perhaps Configure should
detect such a platform and disable the longdouble setting or error out?

A double-double format *can* be well-behaved, if its two halves are
constrained to have a fixed exponent difference and all the edge cases
are handled properly. I concur that we should detect the badly-behaved
ones and reject their use.

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.

-zefram

[p5pRT]

From @sisyphus

-----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