On 13 September 2005 at 09:41, Laurent Bonnaud wrote:
| Package: libgsl0
| Version: 1.6.92-1
| Severity: normal
|
|
| Hi,
|
| here is what I see in /var/log/prelink.log:
|
| /usr/sbin/prelink.bin: Warning: /usr/lib/libgsl.so.0 has undefined non-weak
symbols
|
| and indeed:
|
| $ ldd -r /usr/lib/libgsl.so.0.7.0
| linux-gate.so.1 => (0xffffe000)
| libm.so.6 => /lib/tls/i686/cmov/libm.so.6 (0x4c3cc000)
| libc.so.6 => /lib/tls/i686/cmov/libc.so.6 (0x4c293000)
| /lib/ld-linux.so.2 (0x56555000)
| undefined symbol: cblas_dsdot (/usr/lib/libgsl.so.0.7.0)
| undefined symbol: cblas_dswap (/usr/lib/libgsl.so.0.7.0)
[... many more snipped ... ]
I have no idea what this refers to, and why it would be a bug. Can you
enlighten me? What are non-weak symbols? What do you expect to be done here?
As you may know, GSL links to Blas via its libgslcblas library. Here is ldd
output
from one of the examples:
[EMAIL PROTECTED]:~> ldd /usr/bin/gsl-histogram
linux-gate.so.1 => (0xffffe000)
libgsl.so.0 => /usr/lib/libgsl.so.0 (0xb7e61000)
libgslcblas.so.0 => /usr/lib/libgslcblas.so.0 (0xb7e32000)
libm.so.6 => /lib/tls/libm.so.6 (0xb7e0c000)
libc.so.6 => /lib/tls/libc.so.6 (0xb7cd4000)
/lib/ld-linux.so.2 (0xb7feb000)
As I understand it, Gsl can use its own Blas, as well as Atlas and
friends. Hence two options for gsl-config:
[EMAIL PROTECTED]:~> gsl-config --libs
-L/usr/lib -lgsl -lgslcblas -lm
[EMAIL PROTECTED]:~> gsl-config --libs-without-cblas
-L/usr/lib -lgsl -lm
It seems to me that hard-wiring BLAS back into libgsl would be in conflict
with this degree of freedom / choice regarding BLAS.
Dirk
|
|
| -- System Information:
| Debian Release: testing/unstable
| APT prefers unstable
| APT policy: (100, 'unstable'), (99, 'experimental')
| Architecture: i386 (i686)
| Shell: /bin/sh linked to /bin/bash
| Kernel: Linux 2.6.12-1-686
| Locale: [EMAIL PROTECTED], [EMAIL PROTECTED] (charmap=ISO-8859-15)
|
| Versions of packages libgsl0 depends on:
| ii libc6 2.3.5-6 GNU C Library: Shared libraries
an
|
| libgsl0 recommends no packages.
|
| -- no debconf information
--
Statistics: The (futile) attempt to offer certainty about uncertainty.
-- Roger Koenker, 'Dictionary of Received Ideas of Statistics'
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
