[petsc-users] Fortran stubs 3.22.4

[Barry Smith]

  My changes do not affect the 3.22.* series at all.     Your user likely used the main branch for which this problem would have occurred, but should now be fixed and will be fine in the 2.23.* series



> On Mar 26, 2025, at 2:40 PM, Satish Balay <balay.anl at fastmail.org> wrote:
> 
> Hm - works in 3.22.2 but not 3.22.4? That is strange.
> 
> balay at p1 /home/balay/tmp
> $ nm -Ao petsc-3.22.*/arch*/lib/libpetsc.so |grep matgetrow_
> petsc-3.22.2/arch-linux-c-debug/lib/libpetsc.so:0000000001b44665 T matgetrow_
> petsc-3.22.4/arch-linux-c-debug/lib/libpetsc.so:0000000001b46f9f T matgetrow_
> balay at p1 /home/balay/tmp
> $ nm -Ao petsc-3.22.*/arch*/lib/libpetsc.so |grep matrestorerow_
> petsc-3.22.2/arch-linux-c-debug/lib/libpetsc.so:0000000001b44a2d T matrestorerow_
> petsc-3.22.4/arch-linux-c-debug/lib/libpetsc.so:0000000001b47367 T matrestorerow_
> balay at p1 /home/balay/tmp
> 
> 
> I see the symbols in my builds of both the versions. Suggest rechecking build of 3.22.4 [or redo this build] to make sure its identical to the working 3.22.2 build
> 
> Satish
> 
> On Wed, 26 Mar 2025, Frank Bramkamp wrote:
> 
>> Dear PETSc Team,
>> 
>> 
>> a user of our computer center reported a PETSC issue.
>> He uses petsc version 3.22.4 with gcc compiler
>> 
>> The problem arises with MatGetRow and MatRestoreRow.
>> PETSC compiled properly but his Fortran application then cannot find those
>> routines. It seems to work in version 3.22.2.  For me it also worked in 3.21.1
>> calling those routines from fortran.
>> 
>> It looks that the fortran stubs might have been broken for those two routines in version 3.22.4
>> 
>> I will also test 3.22.4 now myself to confirm it or I can probably test a fixed version as well.
>> 
>> 
>> Greetings, Frank Bramkamp
>> 
>> 
>> 
>> 
>> 
>