[Bug libc/16623] New: cos change of behavior in 2.19 causes scipy testfailures

[Bug libc/16623] New: cos change of behavior in 2.19 causes scipy testfailures

[jtaylor.debian at googlemail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=16623

            Bug ID: 16623
           Summary: cos change of behavior in 2.19 causes scipy
                    testfailures
           Product: glibc
           Version: 2.19
            Status: NEW
          Severity: normal
          Priority: P2
         Component: libc
          Assignee: unassigned at sourceware dot org
          Reporter: jtaylor.debian at googlemail dot com
                CC: drepper.fsp at gmail dot com

#include <math.h>
#include <stdio.h>
int main()
{
  double x= 4831921577.4601803;
  printf("%.15g\n", sin(x));
  printf("%.15g\n", cos(x));
}


in glibc 2.18:

gcc test.c -lm && ./a.out 
-0.918193578775631
-0.396131987972694

while 2.19 has the sign of the cosinus has changed:
gcc test.c -lm && ./a.out 
-0.918193578775631
0.396131987972694

I can't judge if that is within the allowed accuracy of libc but it is causing
wrong results in funpack used for special functions in python-scipy.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

Joseph Myers <jsm28 at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |siddhesh at redhat dot com
          Component|libc                        |math

--- Comment #1 from Joseph Myers <jsm28 at gcc dot gnu.org> ---
Siddhesh, you've been doing the most with these functions lately....

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[bugdal at aerifal dot cx]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

Rich Felker <bugdal at aerifal dot cx> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bugdal at aerifal dot cx

--- Comment #2 from Rich Felker <bugdal at aerifal dot cx> ---
For the record, the old behavior with the negative sign is correct. Tested
against other libm implementations that handle large trig arguments and Wolfram
Alpha.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[siddhesh at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

Siddhesh Poyarekar <siddhesh at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at sourceware dot org   |siddhesh at redhat dot com

--- Comment #3 from Siddhesh Poyarekar <siddhesh at redhat dot com> ---
Must be due to some of the code consolidation I did recently.  I'll take a
look.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[cvs-commit at gcc dot gnu.org]

http://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #4 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, master has been updated
       via  1cadc85813d736f7682fa2eeadae639ab6b66c65 (commit)
      from  f8c17e79fab13a3d1de976c1c3564df7f8c2a175 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=1cadc85813d736f7682fa2eeadae639ab6b66c65

commit 1cadc85813d736f7682fa2eeadae639ab6b66c65
Author: Siddhesh Poyarekar <siddhesh@redhat.com>
Date:   Thu Feb 27 21:12:09 2014 +0530

    Fix sign of input to bsloww1 (BZ #16623)

    In 84ba214c, I removed some redundant sign computations and in the
    process, I incorrectly got rid of a temporary variable, thus passing
    the absolute value of the input to bsloww1.  This caused #16623.

    This fix undoes the incorrect change.

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog                         |   10 +++
 NEWS                              |    2 +-
 math/auto-libm-test-in            |    2 +
 math/auto-libm-test-out           |  138 +++++++++++++++++++++++++++++++++++++
 sysdeps/ieee754/dbl-64/s_sin.c    |   16 +++--
 sysdeps/x86_64/fpu/libm-test-ulps |   44 ++++++++++++
 6 files changed, 205 insertions(+), 7 deletions(-)

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[cvs-commit at gcc dot gnu.org]

http://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #5 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, release/2.19/master has been updated
       via  ffe768a90912f9bce43b70a82576b3dc99e3121c (commit)
      from  ac0b347743833452b8ff56cb94eab5da95136b70 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=ffe768a90912f9bce43b70a82576b3dc99e3121c

commit ffe768a90912f9bce43b70a82576b3dc99e3121c
Author: Siddhesh Poyarekar <siddhesh@redhat.com>
Date:   Thu Feb 27 21:29:16 2014 +0530

    Fix sign of input to bsloww1 (BZ #16623)

    In 84ba214c, I removed some redundant sign computations and in the
    process, I incorrectly got rid of a temporary variable, thus passing
    the absolute value of the input to bsloww1.  This caused #16623.

    This fix undoes the incorrect change.

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog                      |    7 +++++++
 NEWS                           |    2 +-
 sysdeps/ieee754/dbl-64/s_sin.c |   16 ++++++++++------
 3 files changed, 18 insertions(+), 7 deletions(-)

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[siddhesh at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

Siddhesh Poyarekar <siddhesh at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #6 from Siddhesh Poyarekar <siddhesh at redhat dot com> ---
The fix is now in master and the 2.19 branch.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[jtaylor.debian at googlemail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #7 from Julian Taylor <jtaylor.debian at googlemail dot com> ---
confirmed master and 2.19 branch fix the failing application (python scipy),
thanks

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[schwab@linux-m68k.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |alduc1 at free dot fr

--- Comment #8 from Andreas Schwab <schwab@linux-m68k.org> ---
*** Bug 16705 has been marked as a duplicate of this bug. ***

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[alduc1 at free dot fr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #9 from alduc1 <alduc1 at free dot fr> ---
(In reply to Andreas Schwab from comment #8)
> *** Bug 16705 has been marked as a duplicate of this bug. ***

Sorry, it is now OK for sin(4831921577.4601803) but it is KO with my test case
:

sin(2*M_PI*0.27*7390843)         =  -0.63742399  OK
sin(2*M_PI*2.27*7390843)         =   0.63742398  KO

cos(2*M_PI*0.27*7390879)         =  -0.48175367  OK
cos(2*M_PI*2.27*7390879)         =   0.48175367  KO

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[siddhesh at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #10 from Siddhesh Poyarekar <siddhesh at redhat dot com> ---
I don't see your incorrect results on master.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[alduc1 at free dot fr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #11 from alduc1 <alduc1 at free dot fr> ---
(In reply to Siddhesh Poyarekar from comment #10)
> I don't see your incorrect results on master.

I tested with sources from 
http://www.eglibc.org/cgi-bin/viewvc.cgi/branches/eglibc-2_19/.

Is not it the right way to do ?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[alduc1 at free dot fr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #12 from alduc1 <alduc1 at free dot fr> ---
I don't know if eglibc and glibc are the same thing but the fix in s_sin.c of
glibc is not applied to s_sin.c in eglibc 2.19.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

Re: [Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[Ondřej Bílka]

On Fri, Mar 14, 2014 at 02:51:21PM +0000, alduc1 at free dot fr wrote:
> https://sourceware.org/bugzilla/show_bug.cgi?id=16623
> 
> --- Comment #12 from alduc1 <alduc1 at free dot fr> ---
> I don't know if eglibc and glibc are the same thing but the fix in s_sin.c of
> glibc is not applied to s_sin.c in eglibc 2.19.
>
A correct glibc repository is
 
git clone git://sourceware.org/git/glibc.git

a eglibc gets merged into glibc so 2.19 its last release.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[neleai at seznam dot cz]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #13 from Ondrej Bilka <neleai at seznam dot cz> ---
On Fri, Mar 14, 2014 at 02:51:21PM +0000, alduc1 at free dot fr wrote:
> https://sourceware.org/bugzilla/show_bug.cgi?id=16623
> 
> --- Comment #12 from alduc1 <alduc1 at free dot fr> ---
> I don't know if eglibc and glibc are the same thing but the fix in s_sin.c of
> glibc is not applied to s_sin.c in eglibc 2.19.
>
A correct glibc repository is

git clone git://sourceware.org/git/glibc.git

a eglibc gets merged into glibc so 2.19 its last release.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[schwab@linux-m68k.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |warshall at 99main dot com

--- Comment #14 from Andreas Schwab <schwab@linux-m68k.org> ---
*** Bug 16708 has been marked as a duplicate of this bug. ***

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[joseph at codesourcery dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #15 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
I only expect to merge from glibc release branches to EGLIBC release 
branches every few months now that EGLIBC is being phased out.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16623] cos change of behavior in 2.19 causes scipy testfailures

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16623

--- Comment #16 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, gentoo/2.19 has been updated
       via  4545280f9c1ab06efa9348ab46bac436f5885244 (commit)
       via  e1449bcd91f738ea4b0b6d75bb3e1f21827cf047 (commit)
       via  49ca74fb26ce91b00de6df7bcae7ee2923f5f047 (commit)
      from  b76e856d84e14d5ece6988fee589aff26d6ac2f2 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=4545280f9c1ab06efa9348ab46bac436f5885244

commit 4545280f9c1ab06efa9348ab46bac436f5885244
Author: Siddhesh Poyarekar <siddhesh@redhat.com>
Date:   Thu Feb 27 21:29:16 2014 +0530

    Fix sign of input to bsloww1 (BZ #16623)

    In 84ba214c, I removed some redundant sign computations and in the
    process, I incorrectly got rid of a temporary variable, thus passing
    the absolute value of the input to bsloww1.  This caused #16623.

    This fix undoes the incorrect change.

    (cherry picked from commit ffe768a90912f9bce43b70a82576b3dc99e3121c)

    https://bugs.gentoo.org/509494
    https://sourceware.org/bugzilla/show_bug.cgi?id=16623

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=e1449bcd91f738ea4b0b6d75bb3e1f21827cf047

commit e1449bcd91f738ea4b0b6d75bb3e1f21827cf047
Author: Stefan Liebler <stli@linux.vnet.ibm.com>
Date:   Thu Jun 12 14:15:25 2014 +0200

    posix_spawn_faction_addopen: Add missing string.h include directive

    This is needed to avoid a PLT call on s390.

    (cherry picked from commit 35a5e3e338ae17f3d42c60a708763c5d498fb840)

    https://bugs.gentoo.org/513090
    https://sourceware.org/bugzilla/show_bug.cgi?id=17048

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=49ca74fb26ce91b00de6df7bcae7ee2923f5f047

commit 49ca74fb26ce91b00de6df7bcae7ee2923f5f047
Author: Florian Weimer <fweimer@redhat.com>
Date:   Wed Jun 11 23:12:52 2014 +0200

    posix_spawn_file_actions_addopen needs to copy the path argument (BZ 17048)

    POSIX requires that we make a copy, so we allocate a new string
    and free it in posix_spawn_file_actions_destroy.

    Reported by David Reid, Alex Gaynor, and Glyph Lefkowitz.  This bug
    may have security implications.

    (cherry picked from commit 89e435f3559c53084498e9baad22172b64429362)

    https://bugs.gentoo.org/513090
    https://sourceware.org/bugzilla/show_bug.cgi?id=17048

-----------------------------------------------------------------------

Summary of changes:
 posix/spawn_faction_addopen.c  |   14 +++++++++++---
 posix/spawn_faction_destroy.c  |   22 ++++++++++++++++++++--
 posix/spawn_int.h              |    2 +-
 posix/tst-spawn.c              |   10 +++++++++-
 sysdeps/ieee754/dbl-64/s_sin.c |   16 ++++++++++------
 5 files changed, 51 insertions(+), 13 deletions(-)

-- 
You are receiving this mail because:
You are on the CC list for the bug.