Bug 2142316

Scratch build failed. Details below:

BuilderException: Build failed:
Command '['rpmbuild', '-D', '_sourcedir .', '-D', '_topdir .', '-bs', '/var/tmp/thn-36na1m_9/lapack.spec']' returned non-zero exit status 1.

StdOut:
error: Bad source: ./v3.11.0.tar.gz: No such file or directory


Traceback:
  File "/usr/local/lib/python3.10/site-packages/hotness/use_cases/package_scratch_build_use_case.py", line 56, in build
    result = self.builder.build(request.package, request.opts)
  File "/usr/local/lib/python3.10/site-packages/hotness/builders/koji.py", line 188, in build
    raise BuilderException(

If you think this issue is caused by some bug in the-new-hotness, please report it on the-new-hotness issue tracker: https://github.com/fedora-infra/the-new-hotness/issues

Lapack-3.11 apears to have regression in NaN/Inf handling. See e.g.: 

https://savannah.gnu.org/bugs/?63384

Filed the issue upstream:

https://github.com/Reference-LAPACK/lapack/issues/763

Dmitri.

3.11.0 (plus the fix for 763) is in rawhide.

Created attachment 1969084 [details]
bug test  fortran program

May be not directly related to this bug -- trying to recompile src.rpm on Centos Strema 9 with gcc-12 I got the error:

    [100%] Linking C executable VerifyFortranC
    /usr/bin/cmake -E cmake_link_script CMakeFiles/VerifyFortranC.dir/link.txt --verbose=1
    /opt/rh/gcc-toolset-12/root/usr/bin/gcc -O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1  -m64 -march=x86-64-v2 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -frecursive --no-optimize-sibling-calls -DNDEBUG -Wl,-z,relro -Wl,--as-needed  -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 CMakeFiles/VerifyFortranC.dir/main.c.o CMakeFiles/VerifyFortranC.dir/VerifyC.c.o -o VerifyFortranC   -L/usr/lib/gcc/x86_64-redhat-linux/11  libVerifyFortran.a -lgfortran -lm -lquadmath -lm 
    /opt/rh/gcc-toolset-12/root/usr/libexec/gcc/x86_64-redhat-linux/12/ld: /tmp/cc8cLNcL.ltrans0.ltrans.o: in function `main':
    /usr/share/cmake/Modules/FortranCInterface/Verify/main.c:14: undefined reference to `VerifyFortran'
    collect2: error: ld returned 1 exit status
    gmake[3]: *** [CMakeFiles/VerifyFortranC.dir/build.make:115: VerifyFortranC] Error 1

With gcc-11 (default) it rebuilds fine.

Dmitri.
--

I just noticed when I attached the program it wiped my comments. 

With lapack-3.11.0-4.fc39 I see:

$ gfortran bug-63384.f -llapack -lblas
$ ./a.out 
 x = 
                       NaN                       NaN
                       NaN                       NaN
 anorm =                        NaN
 dgetrf info:            0
 ipiv = 
           1           2
 ** On entry to DGECON parameter number  5 had an illegal value
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG
I am not sure if this is a desired behavior. 

With OpenBlas (and with  lapack-3.10) I see:

$ LD_PRELOAD=/usr/lib64/libopenblas.so.0 ./a.out 
 x = 
                       NaN                       NaN
                       NaN                       NaN
 anorm =                        NaN
 dgetrf info:            0
 ipiv = 
           1           2
 dgecon info:            0
 rcond =                        NaN
 dgetri info:            0
 xinv = 
                       NaN                       NaN
                       NaN                       NaN

Dmitri.
--