On Mon, Jun 20, 2011, Francesco Pietra wrote:
> I am trying to install ambertools.1.5 (Debian GNU-Linux amd64 wheezy,
> gcc 4.6.1; gfortran 4.6.1 ) in a separate folder, to be followed by
> installation of ambertools.1.2/amber.10.
>
> ./configure gnu
>
> external idec,random
> 1
> Error: Return type mismatch of function 'random' at (1) (REAL(8)/REAL(4))
> _amg1r5.f:3029.45:
Thanks for the report. There are lots of problems with the amg1r5.f file.
Basically, someone (not sure if it was an Amber developer or not, but it looks
like it) added "implicit double precision" flags in lots of places, but many
of the functions are still labelled "real function cgeps", etc. Apparently,
this sort of thing hasn't caused compiler errors before gfortran 4.6.1, but
seems clearly wrong; it's not clear how what should be happening, or how this
code gives correct answers. (Compiling with -Wall and gfortran 4.4.5 doesn't
show any of the problems that gfortran 4.6.1 seems to find, so it looks like
the behavior of gfortran has changed(?).
Also, the "random" function may conflict with intrinsics, and the code for
it is god-awful (a technical term you can find in advanced treatises).
I'm cc-ing this to the developers list and the Irvine people, since I don't
understand how this part of the code works. I don't have any immediate
workaround, other than trying an older compiler. I'll set up gcc 4.6 on my
machine and run some tests. Of course, the whole thing may be related to
something different than the gfortran version, but it still looks to be like
the code is wrong.
Thanks for the report....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jun 20 2011 - 05:00:03 PDT
