Wierd PETSc build behavior...
Hello,
I'm having the darndest time figuring out what's going wrong with the PETSc
build. It was fine when I built and uploaded PPC packages three weeks ago,
and alpha, sparc and arm seemed to build fine, but it breaks now in exactly
that way on unstable i386 and powerpc and testing alpha.
The problem is that it's setting MPI_LIBS to -L{MPI_HOME}/build/... like it's
substituting ${MPI_HOME} with {MPI_HOME} instead of the value of the
variable. This variable is set in bmake/$(PETSC_ARCH)/base.site where
$(PETSC_ARCH) is linux (i386/arm) or linux_alpha or linux_ppc
(powerpc/sparc/m68k/mips) or hurd.
The libraries in the package actually build just fine, but when you try to
link anything against them using the provided makefile includes (e.g. by
building the examples that come with it), the bogus library variable makes it
fail. And the thing is, that's the *only* variable in all of the makefile
includes that it gets wrong! (You can see the erroneous value of the libs
variable go by when building the petsc package, right before it actually
starts to build the libs, the line starting "Using libraries:", so you don't
have to wait more than a couple of seconds to see the error.)
Any ideas? I'm completely baffled...
-Adam P.
GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6
Welcome to the best software in the world today cafe!
Reply to: