On Sat, Jul 16, 2005 at 10:33:26AM -0700, James R. Phillips wrote: >--- "a.rburgers" wrote: >>I am starting to think we have a g77 bug here or my understanding >>ofparameter passing is incomplete (though I have quite a bit of >>experience withcalling Fortran from C). If I change the wrapper >>function to have the dot argument as *first*argument instead of the >>last argument as in the current sdsdotsub it works. See theattached >>bug.tar.gz for a self contained example. So as a work-around you could >>replace thesdsdotsub wrapper with this one (cblas_f77.h should then >>also be updated) > >You are on the right track here. But in fact the problem is in the >code, not the compiler. The reference cblas author copied the >dsdotsub.f file to sdsdotsub.f as a starting point for editing, but >didn't complete the edits properly. He left the calling argument >signature unchanged, whereas it requires an extra scalar parameter. >It's surprising to find an actual coding error in code from netlib, but >there it is. > >I am requesting a new upload shortly. Have a great weekend.
Can we move away from using cygwin-apps as an all-lapack-all-the-time mailing list and maybe start using the cygwin mailing list for bug reports, please? cgf
