[petsc-dev] fix conflict between triangle and spai

[Satish Balay]

And I see:

$ git branch -r
  origin/HEAD -> origin/master
  origin/maint-1.3
  origin/master

master has triangle-1.6 changes. However petsc is installing and using triangle-1.3 i.e maint.

Matt - any idea why this discrepancy?

Satish

On Tue, 28 Apr 2020, Satish Balay via petsc-dev wrote:

> I get the following error with triangle+spai build with gcc-10 [this is on linux though - ignore osx in PETSC_ARCH]
> 
>      CLINKER arch-ci-osx-cxx-pkgs-opt/lib/libpetsc.so.3.13.0
> /usr/bin/ld: /home/balay/petsc/arch-ci-osx-cxx-pkgs-opt/lib/libtriangle.a(triangle.o):/home/balay/petsc/arch-ci-osx-cxx-pkgs-opt/externalpackages/Triangle/src/triangle.c:667: multiple definition of `epsilon'; /home/balay/petsc/arch-ci-osx-cxx-pkgs-opt/lib/libspai.a(spai.o):(.bss+0x188): first defined here
> 
> 
> Matt,
> 
> Any objection to this change to triangle?
> 
> diff --git a/src/triangle.c b/src/triangle.c
> index 8e915a1..76f6d84 100644
> --- a/src/triangle.c
> +++ b/src/triangle.c
> @@ -664,7 +664,7 @@ long samples;                /* Number of random samples for point location. */
>  unsigned long randomseed;                     /* Current random number seed. */
>  
>  REAL splitter;       /* Used to split REAL factors for exact multiplication. */
> -REAL epsilon;                             /* Floating-point machine epsilon. */
> +static REAL epsilon;                      /* Floating-point machine epsilon. */
>  REAL resulterrbound;
>  REAL ccwerrboundA, ccwerrboundB, ccwerrboundC;
>  REAL iccerrboundA, iccerrboundB, iccerrboundC;
> 
> Satish
>