#20075 closed defect (duplicate)
petsc does not build anymore
Reported by: | root42 | Owned by: | mamoll (Mark Moll) |
---|---|---|---|
Priority: | Normal | Milestone: | |
Component: | ports | Version: | 1.7.1 |
Keywords: | Cc: | mww@… | |
Port: | petsc openmpi |
Description
There seems to be a bug in the petsc portfile: --with-fortran does not work, can it be changed to --without-fortran? The problem is, openmpi in macports is built without fortran. This changeset broke the build: http://trac.macports.org/changeset/49721
Change History (5)
comment:1 Changed 15 years ago by mf2k (Frank Schima)
Cc: | mww@… added |
---|---|
Owner: | changed from macports-tickets@… to mmoll@… |
Port: | petsc openmpi added |
comment:2 Changed 15 years ago by mamoll (Mark Moll)
comment:3 Changed 15 years ago by root42
That's too bad. Because we did not have anything installed, the openmpi was pulled in as a dependency, which in turn failed to build, since it was not of the +g95 flavour. So no chance of specifiying variants in dependencies? Should be fixed IMHO, if this can break the build of a port. Note: We did not have g95 nor gcc42 nor gcc43 installed prior to this.
Regards,
Arne
comment:4 Changed 15 years ago by mamoll (Mark Moll)
Resolution: | → duplicate |
---|---|
Status: | new → closed |
This is one of the oldest bugs. See http://trac.macports.org/wiki/FAQ#IsitpossibletohaveadependencyonaspecificvariantofanotherportE.g.postgresql8server and http://trac.macports.org/ticket/126.
Closing this ticket, duplicate of #126.
petsc works as long as you have built openmpi with Fortran support through one of these variants: g95, gcc42, or gcc43. For scientific computing-related ports such as petsc it does not seem unreasonable to me that a Fortan compiler is installed. However, it is currently impossible to depend on a variant of a port. Adding a dependency on a fortran compiler to the petsc port wouldn't work, since the Fortran support also needs to be built into openmpi.