[Patch, fortran] Document more environment variables in gfortran.texi

[Tobias Burnus]

Hi,

Tobias Burnus wrote:
> I just found this in an old patch of mine. We were missing quite a lot
> of environment variables in the documentation.
>   
Of which the most don't work. I retract this patch.

> +* GFORTRAN_MEM_INIT:: Default initialization of allocated memory
> +* GFORTRAN_MEM_CHECK:: Report on exit still allocated memory
> +* GFORTRAN_SIGHUP:: Behavior on SIGHUP
> +* GFORTRAN_SIGINT:: Behavior on SIGINT
> +* GFORTRAN_FPU_ROUND:: Floating point rounding
> +* GFORTRAN_FPU_PRECISION:: Precision of intermediate results
>   
These don't seem to work. Fixing SIGHUP/SIGINT might be possible, for
the rest I'm not sure.
The memory checking makes more sense as front-end flag as in the library
we may have performance problems.
For the rounding, we have -mfp-rounding-mode (which does not influence
the IO rounding in the library, though).

They are still part of g95, cf. http://ftp.g95.org/G95Manual.pdf, and
seem to be supported there. If one keeps SIG*, one might consider adding
something like DUMP and DUMP-QUIT there, similarly to g95.

Any idea what we should do? Having them resting in the code, not working
and not document does not make sense. They should be removed or there
should be a PR for documenting and making them valid.

> +* GFORTRAN_ERROR_DUMPCORE:: Dump core on run-time errors
> +* GFORTRAN_ERROR_BACKTRACE:: Show backtrace on run-time errors
>   
These are rather recent and are ok. I will later submit a patch for
these only.

> +is not to core dump unless the @option{-fcoredump} compile option
>   
The option is called -fdump-core ...


Tobias