Well, I have seen horrible stuff as package maintainer in a distro and software installation specialist on a cluster. The cases where it would be needed should be fixed upstream. However sometimes upstream is dead or uncaring :( In my opinion tentative of autodetection with an autotool script or a cmake module is doomed (as in it can do a good job but if you are in my job you will always make them run in a corner case they don’t). It is far easier to provide a configuration option.
There are other tales to tell but they drift away from the initial query. François > On 21/10/2016, at 21:28, Emmanuel Charpentier <emanuel.charpent...@gmail.com> > wrote: > > Indeed... (I didn't saw the "--with-lapack" option. Time to change my > spectacles?). > > My patch has no use for this case. Can you think of other use cases ? > > If not, this ticket should be resolved as "Invalid/won't fix"... > > -- > Emmanuel Charpentier > > > Le 21 oct. 2016 10:22, "Francois Bissey" <francois.bis...@canterbury.ac.nz> a > écrit : > Doesn’t have to use .pc - directly. But it has to be able to accept a user > configuration one way or another. Using .pc to provide said configuration > in a script is OK. > You might want to explore the way I did set up numpy in sage for example. > > I looked at JAGS it takes --with-blas and —with-lapack options. You could > solve your problem by doing > —with-lapack=-lopenblas > or to keep with Volker spirit > —with-lapack=`pkg-cong —libs lapack` > > Francois > > > On 21/10/2016, at 21:14, Emmanuel Charpentier > > <emanuel.charpent...@gmail.com> wrote: > > > > While it is the "right" solution, it seems a bit heavy handed to ask all > > authors of LAPACK-using programs susceptible to Sage's use to conform to > > the convention suggested by Volker. This might not even be possible, > > depending on system's infrastructure and source's configuration. And > > effectively forking all such LAPACK-using programs puts an heavy load on > > the users... > > > > Meanwhile, this trivial patch needs review. > > > > HTH, > > > > -- > > Emmanuel Charpentier > > > > Le vendredi 21 octobre 2016 00:17:11 UTC+2, François a écrit : > > I agree it would be best. Unfortunately it may not always be possible > > to do that without heavy hacking. > > Adding the liblapack link has trade-off, it would make some stuff work > > out of the box but it could also hide problems on the long term. > > > > Really upstream of any project using blas/lapack needs to have a way > > to check a user provided configuration before trying a automated detection > > routine. Looking at you numpy (does the reverse by default without hacking, > > look at user provided configuration only when its auto detection has > > failed). > > > > François > > > > > On 21/10/2016, at 11:08, Volker Braun <vbrau...@gmail.com> wrote: > > > > > > We do have a pc file so IMHO the best way is to rely on the output of > > > > > > $ pkg-config --cflags lapack > > > -I/mnt/disk/home/release/Sage/local/include > > > $ pkg-config --libs lapack > > > -L/mnt/disk/home/release/Sage/local/lib -lopenblas > > > > > > > > > > > > On Thursday, October 20, 2016 at 11:12:43 PM UTC+2, François wrote: > > > This is because automatic blas/lapack detection is a hopeless task. > > > You should pass your lapack libraries to the configuration script. > > > If you can’t, hack it. > > > And now that we have switched to openblas, -lopenblas provides lapack. > > > > > > François > > > > > > > On 21/10/2016, at 10:04, Emmanuel Charpentier <emanuel.c...@gmail.com> > > > > wrote: > > > > > > > > Something just changed in the Sage build process : as of 7.4, LAPACK is > > > > no longer available to programs running in the Sage shell. > > > > > > > > Case in point : I want to install JAGS (an MCMC sampler), which is > > > > needed by the rjags R package, of some use to Bayesian R users... This > > > > package compiled fine in the Sage shell up to Sage 7.4beta6. Now, the > > > > --configure step fails with : > > > > > > > > checking for cheev_ in -llapack... no > > > > checking for cheev_ in -llapack_rs6k... no > > > > configure: error: "You need to install the LAPACK library" > > > > > > > > Indeed : > > > > > > > > charpent@asus16-ec:~$ find /usr/local/sage-7/ -iname "*lapack*so*" > > > > /usr/local/sage-7/local/lib/R/modules/lapack.so > > > > /usr/local/sage-7/local/lib/python2.7/site-packages/cvxopt/lapack.so > > > > /usr/local/sage-7/local/lib/python2.7/site-packages/numpy/linalg/lapack_lite.so > > > > /usr/local/sage-7/local/lib/python2.7/site-packages/scipy/linalg/_flapack.so > > > > /usr/local/sage-7/local/lib/python2.7/site-packages/scipy/linalg/cython_lapack.so > > > > charpent@asus16-ec:~$ find /usr/local/sage-7/ -iname "*lapack*h*" > > > > /usr/local/sage-7/local/lib/R/include/R_ext/Lapack.h > > > > /usr/local/sage-7/local/include/lapacke_mangling.h > > > > /usr/local/sage-7/local/include/linbox/algorithms/numeric-solver-lapack.h > > > > /usr/local/sage-7/local/include/lapacke_config.h > > > > /usr/local/sage-7/local/include/lapacke.h > > > > > > > > > > > > Short of forcing the Sage build process to use the system's LAPACK > > > > (which can be done but, as far as I know, is not recommended), I don't > > > > see how to force the creation of this library and its header file. > > > > > > > > Note that could be useful to reduce overhead : R (standard package) > > > > creates its own version (probably for lack of finding it at build > > > > time...) ; it also exists in the python package cvxopt. > > > > > > > > 1) What do you think ? > > > > > > > > 2) Should a ticket be filed ? > > > > > > > > HTH, > > > > > > > > -- > > > > Emmanuel Charpentier > > > > > > > > -- > > > > You received this message because you are subscribed to the Google > > > > Groups "sage-devel" group. > > > > To unsubscribe from this group and stop receiving emails from it, send > > > > an email to sage-devel+...@googlegroups.com. > > > > To post to this group, send email to sage-...@googlegroups.com. > > > > Visit this group at https://groups.google.com/group/sage-devel. > > > > For more options, visit https://groups.google.com/d/optout. > > > > > > > > > -- > > > You received this message because you are subscribed to the Google Groups > > > "sage-devel" group. > > > To unsubscribe from this group and stop receiving emails from it, send an > > > email to sage-devel+...@googlegroups.com. > > > To post to this group, send email to sage-...@googlegroups.com. > > > Visit this group at https://groups.google.com/group/sage-devel. > > > For more options, visit https://groups.google.com/d/optout. > > > > > > -- > > You received this message because you are subscribed to the Google Groups > > "sage-devel" group. > > To unsubscribe from this group and stop receiving emails from it, send an > > email to sage-devel+unsubscr...@googlegroups.com. > > To post to this group, send email to sage-devel@googlegroups.com. > > Visit this group at https://groups.google.com/group/sage-devel. > > For more options, visit https://groups.google.com/d/optout. > > -- > You received this message because you are subscribed to a topic in the Google > Groups "sage-devel" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/sage-devel/3cPELP9DXKg/unsubscribe. > To unsubscribe from this group and all its topics, send an email to > sage-devel+unsubscr...@googlegroups.com. > To post to this group, send email to sage-devel@googlegroups.com. > Visit this group at https://groups.google.com/group/sage-devel. > For more options, visit https://groups.google.com/d/optout. > > -- > You received this message because you are subscribed to the Google Groups > "sage-devel" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to sage-devel+unsubscr...@googlegroups.com. > To post to this group, send email to sage-devel@googlegroups.com. > Visit this group at https://groups.google.com/group/sage-devel. > For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
