[Bug 744339] Review Request: dieharder - Random number generator tester and timer

bugzilla at redhat.com bugzilla at redhat.com
Tue Nov 15 21:29:07 UTC 2011


Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=744339

--- Comment #13 from Jiri Hladky <hladky.jiri at gmail.com> 2011-11-15 16:29:06 EST ---
Hi

I have informed upstream about the rpmlint warnings

dieharder.i686: W: unused-direct-shlib-dependency
/usr/lib/libdieharder.so.3.31.1 linux-gate.so.1
dieharder.i686: W: unused-direct-shlib-dependency
/usr/lib/libdieharder.so.3.31.1 /usr/lib/libgslcblas.so.0

and I got following reply

======================================================================
linux-gate.so.1
First, read:

 http://www.trilithium.com/johan/2005/08/linux-gate/

This isn't a real library, and I have no control over it.  It is a
kernel artifact and completely irrelevant to bloat.  IIRC it is also
known as vdso and listed (by e.g. ldd) on some systems as linux-vsdo.so
instead of linux-gate.so.  rpmlint should "probably" just ignore this
when it pops up in a binary as I didn't turn it on and don't have any
idea how to turn it off, under system/kernel control (artifact to help
manage shared objects) not mine.

The gslblas library is "required" for full transparent GSL
functionality. To quote the GSL manual's recommended build line:

 $ gcc -L/usr/local/lib example.o -lgsl -lgslcblas -lm

I'd argue that this should remain, partly because while I don't use BLAS
at this particular instant to do linear algebra associated with some of
the tests, at some point I will (I have in some of the unreleased stuff
I'm playing with).  In any event, since the primary distribution form of
the tarball (which is in the src rpm) is designed to encourage
tinkering, and I really want the tinkering to be done using GSL
functions and linear algebra where possible, I think there is virtue in
keeping -lgslcblas as is.  That way people (including me!) won't be
surprised if they link a gsl function that requires it while playing
around and the make breaks....;-)

Hope that helps.  If I absolutely must, I can try turning BLAS off in
the configuration in a way that people can turn back on again easily,
but it will probably remain on in my default build.  Linear algebra is a
big part of statistical analysis and it's just a matter of time before
some matrix or another will need to be inverted in dieharder.

======================================================================

I think that both points are valid. rpmlint should not warn about
linux-gate.so.1

As for -lgslcblas I agree with the upstream that it should stay as it is. See

http://www.gnu.org/s/gsl/manual/html_node/Linking-programs-with-the-library.html

Please let me know your opinion on it.

Thanks
Jirka

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.



More information about the package-review mailing list