This Message Is From an External Sender
This message came from outside your organization.
On Mon, 1 Jul 2024, meator wrote:
> On 7/1/24 5:23 PM, Stefano Zampini wrote: > > This is unfortunate. I assume that patching the buildsystem or some > > other trickery will be necessary if what you're saying is true. > > > > I admit it is not so standard, but Satish always opposed this. > > No need to hack the buildsystem, just use COPTFLAGS, CXXOPTFLAGS and > > FOPTFLAGS > > I will try using *OPTFLAGS, thanks! > > > I don't think you should use Makefile.user. That is there as a sort of > > template/placeholder. The extra variables are not included in a standard > > usage of pkgconfig, so I don't think this issue is "severe" > > This is not a choice I get to make. The users of my package may choose to use > PETSc however they want. And /usr/share/petsc/Makefile.user or > /usr/share/petsc/CMakeLists.txt are officially supported ways of using PETSc, > so deciding to ignore these use cases and leaving junk flags in > /usr/lib/pkgconfig/petsc.pc is not tolerable for me. Usually configure flags get saved and reused by examples - and in some cases might be required by applications. [for ex: if there is -m32 flag (gcc) somewhere. Other example: /MD for MS compiler] So your "objection" appears to be not based on real requirements. And for your desired use case: "I need to specify flags to PETSc build - and they should not bleed to dependent pkgs" I've indicated one mode for this. i.e - do not specify them with configure - only list them with make. ./configure make CFLAGS='only-for-petsc-library-build' Satish
