Hi, I have a problem with my installation of gretl on ArchLinux. While trying to help on SO<http://stackoverflow.com/questions/15997478/gretl-test-alpha-and-be...;, I get a segfault error

uname -a Linux dickoa 3.8.6-1-ARCH #1 SMP PREEMPT Sat Apr 6 07:27:01 CEST 2013 x86_64 GNU/Linux I didn't make it also in make check I get stucked at *** Filip.dat *** ("Higher Level of Difficulty") cholbeta: test[6] = 3.77499e-09 make[1]: *** [check] Segmentation fault (core dumped)

On Sun, 14 Apr 2013, Riccardo (Jack) Lucchetti wrote: Uhm, that's strange. The script you sent runs fine here. What are your > configure options? Also, could you try running the script that provokes the > segfault under valgrind and see what happens? > Wild guess: something strange may be happening when doing QR decomposition. What version of lapack are you using? ------------------------------**------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/**servizi/hpp/lucchetti<http://www2.econ.un... ------------------------------**------------------------- _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users

I checked with Valgrind and you were right about lapack and It's related to my openblas (built which 64 bit interface) I have something like this when check for restriction on lfp (lfp=0) in my example ==26438== Process terminating with default action of signal 11 (SIGSEGV) ==26438== Bad permissions for mapped region at address 0x9EBD000 ==26438== at 0x55E1CC4: idamax_k (in /usr/lib/ libopenblas_sandybridgep-r0.2.6.so) ==26438== by 0x5B275A6: dgetf2_ (in /usr/lib/liblapack.so) ==26438== by 0x5B27D00: dgetrf_ (in /usr/lib/liblapack.so) ==26438== by 0x4F24468: gretl_invert_general_matrix (gretl_matrix.c:7111) ==26438== by 0x4F442ED: check_R_matrix (gretl_restrict.c:100) ==26438== by 0x4F4645B: test_restriction_set (gretl_restrict.c:2285) ==26438== by 0x4F47FBA: gretl_restriction_finalize_full (gretl_restrict.c:2469) ==26438== by 0x4F5BA68: gretl_cmd_exec (interact.c:4847) ==26438== by 0x40442C: exec_line (gretlcli.c:1113) ==26438== by 0x403295: main (gretlcli.c:611) When I do the same example as above is just complained that the matrix is singular. On Sun, Apr 14, 2013 at 12:57 PM, Ahmadou Dicko <dicko.ahmadou(a)gmail.com&gt;wrote: > I have the version 3.4.2.1 of lapack > I don't if it is related but I also use openblas as blas. > > > On Sun, Apr 14, 2013 at 12:50 PM, Riccardo (Jack) Lucchetti < > r.lucchetti(a)univpm.it&gt; wrote: > >> On Sun, 14 Apr 2013, Riccardo (Jack) Lucchetti wrote: >> >> Uhm, that's strange. The script you sent runs fine here. What are your >>> configure options? Also, could you try running the script that provokes the >>> segfault under valgrind and see what happens? >>> >> >> Wild guess: something strange may be happening when doing QR >> decomposition. What version of lapack are you using? >> >> >> ------------------------------**------------------------- >> Riccardo (Jack) Lucchetti >> Dipartimento di Scienze Economiche e Sociali (DiSES) >> >> Università Politecnica delle Marche >> (formerly known as Università di Ancona) >> >> r.lucchetti(a)univpm.it >> http://www2.econ.univpm.it/**servizi/hpp/lucchetti<http://www2.econ.un... >> ------------------------------**------------------------- >> >> _______________________________________________ >> Gretl-users mailing list >> Gretl-users(a)lists.wfu.edu >> http://lists.wfu.edu/mailman/listinfo/gretl-users >> > > > > -- > Ahmadou H. DICKO > statistician and applied economist > PhD student in Climate change economics > Faculty of economics and managment - Cheikh Anta Diop University > West African Science Service Center on Climate Change and Adaptated Land > Use (WASCAL) > Center for Development Research (ZEF) - University of Bonn > twitter : @dickoah > github : github/dickoa > tel : +221 33 827 55 16 > portable: +221 77 123 81 69 > -- Ahmadou H. DICKO statistician and applied economist PhD student in Climate change economics Faculty of economics and managment - Cheikh Anta Diop University West African Science Service Center on Climate Change and Adaptated Land Use (WASCAL) Center for Development Research (ZEF) - University of Bonn twitter : @dickoah github : github/dickoa tel : +221 33 827 55 16 portable: +221 77 123 81 69

On Sun, 14 Apr 2013, Ahmadou Dicko wrote: > Finally, I decided to remove openblas from /usr/lib and I installed it in a > custom directory not in the PATH and it works fine. It looks as if some possibly relevant openblas bugs have been fixed in recent releases. http://xianyi.github.io/OpenBLAS/Changelog.txt Seems a bit experimental, though. No fun having an unstable BLAS. Allin Cottrell _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users

On Sun, 14 Apr 2013, Ahmadou Dicko wrote: > I already use the experimental version of OpenBLAS and it's really fast. > > For now, I use R and Julia with openblas and use the normal blas for gretl. > I just hope that I'll be able to build gretl using openblas soon I'll try > from time to time. How are you configuring gretl (and building openblas)? If you do a default "make" of openblas, this builds a library that includes both lapack and blas. In that case you must not link against the standard liblapack: you should do something like the following when configuring gretl LAPACK_LIBS='-lopenblas' ./configure <options> I've tested that here (on 32-bit Linux) with openblas 0.2.6 and it works fine. Your valgrind log showed ==26438== Bad permissions for mapped region at address 0x9EBD000 ==26438== at 0x55E1CC4: idamax_k (in /usr/lib/ libopenblas_sandybridgep-r0.2.6.so) ==26438== by 0x5B275A6: dgetf2_ (in /usr/lib/liblapack.so) which suggests you're mixing incompatible libraries. libopenblas should be providing dgetf2_ waverley:~$ nm /usr/lib/libopenblas_sandybridgep-r0.2.6.so \ | grep dgetf 00072ab0 T dgetf2_ 00981cc0 T LAPACKE_dgetf2 00981d70 T LAPACKE_dgetf2_work Allin Cottrell _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users