forwarded message from Anders H. Abrahamsen

forwarded message from Anders H. Abrahamsen

[Brian Gough]

[-- Attachment #1: message body text --]
[-- Type: text/plain, Size: 12 bytes --]

Any ideas?


[-- Attachment #2: forwarded message --]
[-- Type: message/rfc822, Size: 1932 bytes --]

From: "Anders H. Abrahamsen" <andersa@ii.uib.no>
To: Brian Gough <bjg@network-theory.co.uk>
Subject: Re: problems with test?
Date: Fri, 7 Dec 2001 15:20:30 +0100 (MET)
Message-ID: <Pine.SOL.4.40.0112071516380.12109-100000@apal.ii.uib.no>

I have tried both gcc 2.96 and gcc 3.1-0.10. For three days ago I ran
Rehat 7.1, now I've just upgraded to 7.2.

andersa

> Anders H A writes:
>  > When I run the tests (make check), it locks up after this test:
>  >
>  > ..
>  > PASS: gsl_sf_ellint_E_e(M_PI/3.0, 0.01, mode, &r)m
>  >
>  > (All tests before PASSes.)
>  >
>
> Hi,
> What platform, compiler, etc are you using?
>
> If possible can you run the test under the debugger and see where
> it is going into an infinite loop.
>
> regards
> Brian
>

-- 
Anders Hagelin Abrahamsen
andersa@ii.uib.no
andersa@c2i.net

[forwarded message from Anders H. Abrahamsen]

[Dirk Eddelbuettel]

  "Brian" == Brian Gough <bjg@network-theory.co.uk> writes:
  Brian>  Any ideas?  

  Anders> I have tried both gcc 2.96 and gcc 3.1-0.10. For three days ago I
  Anders> ran Rehat 7.1, now I've just upgraded to 7.2.
[...]
  Anders> Anders H A writes: > When I run the tests (make check), it locks up
  Anders> after this test: > > ..  > PASS: gsl_sf_ellint_E_e(M_PI/3.0, 0.01,
  Anders> mode, &r)m > > (All tests before PASSes.)  >

You can assess success (or lack thereof) on various platform and compilers at
Debian's "Build Daemon Success/Failure" database. Goto

       http://buildd.debian.org/build.php?arch=&pkg=gsl

to pull the logs for the gsl package. As the package build process calls
'make check', these logs are fairly long and detailed.

Among the platforms that compiled gsl-1.0 successfully (i.e. i386, ia64 [aka
"itanium"], alpha, arm, mips, mipsel, sparc), the ia64 port uses a gcc 2.96:

edd@caballero:~$ gcc -v
Reading specs from /usr/lib/gcc-lib/ia64-linux/2.96/specs
gcc version 2.96 20000731 (Debian GNU/Linux IA64 experimental)

edd@caballero:~$ uname -a
Linux caballero 2.4.9-itanium-smp #1 SMP Wed Nov 21 13:44:54 MST 2001 ia64 unknown

Now, as everybody knows, what Red Hat calls "gcc 2.96" is probably not what
the rest of the world calls "gcc 2.96" so you might want to check for an
upgraded gcc toolchain there.

I can't think of a platform (other than hppa) that enforces
gcc-3.0. Unfortunately, hppa currently has toolchain issues so there is no
attempt beyond the failure of 0.9.4 back in October. I do have other packages
with issues on that platform.

Hope this helps, 

Dirk  (Debian maintainer for GSL)

-- 
Good judgment comes from experience; experience comes from bad judgment. 
							    -- F. Brooks

Re: [forwarded message from Anders H. Abrahamsen]

[Anders H A]

This is the platforms for which there exists a log file, I use as previously 
stated Redhat 7.2, but I will be using IRIX64 (Origin 2000) to run the 
programs later on:

1.0-1 (powerpc) (latest build at Nov 3 17:27: successful)
1.0-1 (alpha) (latest build at Nov 3 18:02: successful)
1.0-1 (ia64) (latest build at Nov 3 18:27: successful)
1.0-1 (sparc) (latest build at Nov 3 18:57: successful)
1.0-1 (mipsel) (latest build at Nov 3 19:45: successful)
1.0-1 (m68k) (latest build at Nov 23 02:03: failed)
1.0-1 (mips) (latest build at Nov 3 22:37: successful)
1.0-1 (arm) (latest build at Nov 4 02:28: successful)

No i386 log file as far as I can see.
-----------------------
My problem with running the checks with Redhat's gcc 2.96 is this:

>...
>FAIL: herm(10), orthogonal(6,9), abs/desc (1.45946345652517906e+142 observed 
>vs 0 expected)
>PASS: herm(10), orthogonal(7,8), abs/desc (nan observed vs 0 expected)
>PASS: herm(10), orthogonal(7,9), abs/desc (1.38532e-32 observed vs 0 
>expected)
>PASS: herm(10), orthogonal(8,9), abs/desc (nan observed vs 0 expected)
>FAIL: test
>===================
>1 of 1 tests failed
>===================
>make[2]: *** [check-TESTS] Error 1
>make[2]: Leaving directory `/home/angel/gsl/gsl-1.0/eigen'
>make[1]: *** [check-am] Error 2
>make[1]: Leaving directory `/home/angel/gsl/gsl-1.0/eigen'
>make: *** [check-recursive] Error 1
>[angel@villand gsl-1.0]$

I get a couple of hundred FAILs on the eigenvalue tests, all tests prior to 
these PASSes. Cannot find these tests in the log files found at:

http://buildd.debian.org/build.php?arch=&pkg=gsl

Is there an easy way to eliminate the eigenvalue tests? Then I could check if 
the rest worked. Anyone knows why these tests don't PASS?

(This is a different problem than described in an earlier mail)

andersa

Re: [forwarded message from Anders H. Abrahamsen]

[Brian Gough]

Anders H A writes:
 > My problem with running the checks with Redhat's gcc 2.96 is this:
 > FAIL: herm(10), orthogonal(6,9), abs/desc (1.45946345652517906e+142
 > observed

Hi,

gcc 2.96 has known bugs, which cause that failure. You'll need to use
a different version of gcc, e.g. gcc-3.0.

regards
Brian Gough

[forwarded message from Anders H. Abrahamsen]

[Dirk Eddelbuettel]

  "Brian" == Brian Gough <bjg@network-theory.co.uk> writes:
  Brian>  Any ideas?  

  Anders> I have tried both gcc 2.96 and gcc 3.1-0.10. For three days ago I
  Anders> ran Rehat 7.1, now I've just upgraded to 7.2.
[...]
  Anders> Anders H A writes: > When I run the tests (make check), it locks up
  Anders> after this test: > > ..  > PASS: gsl_sf_ellint_E_e(M_PI/3.0, 0.01,
  Anders> mode, &r)m > > (All tests before PASSes.)  >

You can assess success (or lack thereof) on various platform and compilers at
Debian's "Build Daemon Success/Failure" database. Goto

       http://buildd.debian.org/build.php?arch=&pkg=gsl

to pull the logs for the gsl package. As the package build process calls
'make check', these logs are fairly long and detailed.

Among the platforms that compiled gsl-1.0 successfully (i.e. i386, ia64 [aka
"itanium"], alpha, arm, mips, mipsel, sparc), the ia64 port uses a gcc 2.96:

edd@caballero:~$ gcc -v
Reading specs from /usr/lib/gcc-lib/ia64-linux/2.96/specs
gcc version 2.96 20000731 (Debian GNU/Linux IA64 experimental)

edd@caballero:~$ uname -a
Linux caballero 2.4.9-itanium-smp #1 SMP Wed Nov 21 13:44:54 MST 2001 ia64 unknown

Now, as everybody knows, what Red Hat calls "gcc 2.96" is probably not what
the rest of the world calls "gcc 2.96" so you might want to check for an
upgraded gcc toolchain there.

I can't think of a platform (other than hppa) that enforces
gcc-3.0. Unfortunately, hppa currently has toolchain issues so there is no
attempt beyond the failure of 0.9.4 back in October. I do have other packages
with issues on that platform.

Hope this helps, 

Dirk  (Debian maintainer for GSL)

-- 
Good judgment comes from experience; experience comes from bad judgment. 
							    -- F. Brooks

Re: [forwarded message from Anders H. Abrahamsen]

[Anders H A]

This is the platforms for which there exists a log file, I use as previously 
stated Redhat 7.2, but I will be using IRIX64 (Origin 2000) to run the 
programs later on:

1.0-1 (powerpc) (latest build at Nov 3 17:27: successful)
1.0-1 (alpha) (latest build at Nov 3 18:02: successful)
1.0-1 (ia64) (latest build at Nov 3 18:27: successful)
1.0-1 (sparc) (latest build at Nov 3 18:57: successful)
1.0-1 (mipsel) (latest build at Nov 3 19:45: successful)
1.0-1 (m68k) (latest build at Nov 23 02:03: failed)
1.0-1 (mips) (latest build at Nov 3 22:37: successful)
1.0-1 (arm) (latest build at Nov 4 02:28: successful)

No i386 log file as far as I can see.
-----------------------
My problem with running the checks with Redhat's gcc 2.96 is this:

>...
>FAIL: herm(10), orthogonal(6,9), abs/desc (1.45946345652517906e+142 observed 
>vs 0 expected)
>PASS: herm(10), orthogonal(7,8), abs/desc (nan observed vs 0 expected)
>PASS: herm(10), orthogonal(7,9), abs/desc (1.38532e-32 observed vs 0 
>expected)
>PASS: herm(10), orthogonal(8,9), abs/desc (nan observed vs 0 expected)
>FAIL: test
>===================
>1 of 1 tests failed
>===================
>make[2]: *** [check-TESTS] Error 1
>make[2]: Leaving directory `/home/angel/gsl/gsl-1.0/eigen'
>make[1]: *** [check-am] Error 2
>make[1]: Leaving directory `/home/angel/gsl/gsl-1.0/eigen'
>make: *** [check-recursive] Error 1
>[angel@villand gsl-1.0]$

I get a couple of hundred FAILs on the eigenvalue tests, all tests prior to 
these PASSes. Cannot find these tests in the log files found at:

http://buildd.debian.org/build.php?arch=&pkg=gsl

Is there an easy way to eliminate the eigenvalue tests? Then I could check if 
the rest worked. Anyone knows why these tests don't PASS?

(This is a different problem than described in an earlier mail)

andersa

Re: [forwarded message from Anders H. Abrahamsen]

[Brian Gough]

Anders H A writes:
 > My problem with running the checks with Redhat's gcc 2.96 is this:
 > FAIL: herm(10), orthogonal(6,9), abs/desc (1.45946345652517906e+142
 > observed

Hi,

gcc 2.96 has known bugs, which cause that failure. You'll need to use
a different version of gcc, e.g. gcc-3.0.

regards
Brian Gough

forwarded message from Anders H. Abrahamsen

[Brian Gough]

Any ideas?

forwarded message from Anders H. Abrahamsen

[Brian Gough]

[-- Attachment #1: message body text --]
[-- Type: text/plain, Size: 12 bytes --]

Any ideas?


[-- Attachment #2: forwarded message --]
[-- Type: message/rfc822, Size: 1932 bytes --]

From: "Anders H. Abrahamsen" <andersa@ii.uib.no>
To: Brian Gough <bjg@network-theory.co.uk>
Subject: Re: problems with test?
Date: Fri, 7 Dec 2001 15:20:30 +0100 (MET)
Message-ID: <Pine.SOL.4.40.0112071516380.12109-100000@apal.ii.uib.no>

I have tried both gcc 2.96 and gcc 3.1-0.10. For three days ago I ran
Rehat 7.1, now I've just upgraded to 7.2.

andersa

> Anders H A writes:
>  > When I run the tests (make check), it locks up after this test:
>  >
>  > ..
>  > PASS: gsl_sf_ellint_E_e(M_PI/3.0, 0.01, mode, &r)m
>  >
>  > (All tests before PASSes.)
>  >
>
> Hi,
> What platform, compiler, etc are you using?
>
> If possible can you run the test under the debugger and see where
> it is going into an infinite loop.
>
> regards
> Brian
>

-- 
Anders Hagelin Abrahamsen
andersa@ii.uib.no
andersa@c2i.net

[forwarded message from Anders H. Abrahamsen]

[Dirk Eddelbuettel]

  "Brian" == Brian Gough <bjg@network-theory.co.uk> writes:
  Brian>  Any ideas?  

  Anders> I have tried both gcc 2.96 and gcc 3.1-0.10. For three days ago I
  Anders> ran Rehat 7.1, now I've just upgraded to 7.2.
[...]
  Anders> Anders H A writes: > When I run the tests (make check), it locks up
  Anders> after this test: > > ..  > PASS: gsl_sf_ellint_E_e(M_PI/3.0, 0.01,
  Anders> mode, &r)m > > (All tests before PASSes.)  >

You can assess success (or lack thereof) on various platform and compilers at
Debian's "Build Daemon Success/Failure" database. Goto

       http://buildd.debian.org/build.php?arch=&pkg=gsl

to pull the logs for the gsl package. As the package build process calls
'make check', these logs are fairly long and detailed.

Among the platforms that compiled gsl-1.0 successfully (i.e. i386, ia64 [aka
"itanium"], alpha, arm, mips, mipsel, sparc), the ia64 port uses a gcc 2.96:

edd@caballero:~$ gcc -v
Reading specs from /usr/lib/gcc-lib/ia64-linux/2.96/specs
gcc version 2.96 20000731 (Debian GNU/Linux IA64 experimental)

edd@caballero:~$ uname -a
Linux caballero 2.4.9-itanium-smp #1 SMP Wed Nov 21 13:44:54 MST 2001 ia64 unknown

Now, as everybody knows, what Red Hat calls "gcc 2.96" is probably not what
the rest of the world calls "gcc 2.96" so you might want to check for an
upgraded gcc toolchain there.

I can't think of a platform (other than hppa) that enforces
gcc-3.0. Unfortunately, hppa currently has toolchain issues so there is no
attempt beyond the failure of 0.9.4 back in October. I do have other packages
with issues on that platform.

Hope this helps, 

Dirk  (Debian maintainer for GSL)

-- 
Good judgment comes from experience; experience comes from bad judgment. 
							    -- F. Brooks