support for F90/F95 in SCons
From: Chris Murray (murpup_at_comcast.net)
Date: 08/19/04
- Next message: Kamaraju Kusumanchi: "Re: f90 array question"
- Previous message: Richard Maine: "Re: variable length strings"
- Next in thread: Janne Blomqvist: "Re: support for F90/F95 in SCons"
- Reply: Janne Blomqvist: "Re: support for F90/F95 in SCons"
- Reply: Andy Vaught: "Re: support for F90/F95 in SCons"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 18 Aug 2004 21:56:42 -0400
Hi all,
Just wanted to let everyone know that v0.96 of SCons, a software
construction tool designed to replace make, has just been released. It
now has support for F90/F95 and the automated dependency checking
necessary to associate particular set of .mod files with a particular
source file such that source files will get built in the proper order
without the user having to explicitly set up (or run some script to set
up) dependencies between an object file and some set of source files.
The software can be downloaded from http://www.scons.org
Here are some of the features and limitations associated with the F90
dependency scanner
1) It IS designed to be able to deal with multiple USE statements per
line EXCEPT there is a design flaw if that multi-USE line is commented
out, ala:
! USE mod_a; USE mod_b
In that case, it will still try to create a dependency for the second
USE statement. You could bypass this by adding an exclamation after the
semicolon
2) It IS designed to deal with multiple MODULE statements per source
file, although I have not tested that out. The project I needed this
functionality for only has 1 module per file.
3) The assumption is that the compiler will create a module file for
each MODULE statement and the filename will be all lowercase. Most
compilers follow this convention, although I noticed that some on the
MAC (xlf, absoft) don't. Not sure how this will play out (I think MAC
is case-neutral, but now that it has a UNIX layer underneath, I don't
know). I know there is nothing in the standard that forces a compiler to
use .mod files, but this is how every compiler that I know of works.
4) The default module file extension is .mod, but that can be overridden.
5) It probably won't set up the dependencies properly if a MODULE is
brought in via an INCLUDE file, especially when the USE statement that
references it is in the source file that contains the INCLUDE statement.
6) By rights, the dependency scanner is supposed to be able to handle
recursive include's, so if an included file contains a USE statement, it
SHOULD be able to properly pick up the dependency, but that
functionality has not yet been tested.
I think SCons holds a lot of promise in terms of being able to make the
build process easier. I hope there are some brave souls out there
willing to give the software a try. If you do, feel free to post your
experiences here or on the SCons mailing list, or contribute some
patches to make the F90 support even better.
Chris
- Next message: Kamaraju Kusumanchi: "Re: f90 array question"
- Previous message: Richard Maine: "Re: variable length strings"
- Next in thread: Janne Blomqvist: "Re: support for F90/F95 in SCons"
- Reply: Janne Blomqvist: "Re: support for F90/F95 in SCons"
- Reply: Andy Vaught: "Re: support for F90/F95 in SCons"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|