Hello all,
I guess I got side-tracked about the exact nature of the problem -
recognising
those lines that can possibly contain dependency information - as opposed
to the problem of determining comment lines ;).
I am for easy, and having to specify this is difficult for cmake
developers and
for cmake users. The only thing cmake is parsing the files for is
to create
depend information, that is it. I have not yet seen a case where
the cC*dD starting a
line would cause incorrect depend information and still be valid
fortran code.
Lines starting with d or D in fixed form source are regarded as comment,
unless
you explicitly want to have them compiled. I am not sure whether they
are part
of the Fortran 90/95 standard, but it was a popular extension for
FORTRAN 77.
That said, look at the following source fragment:
C Include the debug code only if we really want it ....
D INCLUDE 'debug_code.inc'
If all lines starting with d or D would be regarded as comment, CMake would
miss this dependency.
Note: this feature of conditionally compiled source code does not exist with
the free form.
OK. That statement about the limited nature of the parsing is
reassuring so
would somebody please commit the solution, i.e.,
<fixed_fmt>^[cC*dD].*\n { return EOSTMT; }
to:
^[cC*dD].*\n { return EOSTMT; }
that Bryan Walsh has already worked out?
Please reduce that to:
^[cC*].*\n { return EOSTMT; }
and you will not miss any dependencies as far as I can tell.
(I was wondering about false positives, like when you have
"INCLUDE 'filename'" or "USE module" in a literal string,
but it should be relatively easy to detect them. The following
fragment could cause a false dependency:
write(*,*) 'Usage: &
& use mymodule &
& only mysub'
unless you are careful with the REs.)
Regards,
Arjen
_______________________________________________
CMake mailing list
[email protected]
http://www.cmake.org/mailman/listinfo/cmake
