On behalf of Jeff Squyres (discussion moved from ompi-core)
I marginally favor having the .c in SVN (generated from a recent flex)
so that I don't have to see warnings every day in the MTT output. :-)
However, I am *only* in favor of this as long as the makefile
dependencies can be set right such that it will *not* auto-generate a
new copy for developers that would lead to an endless sequence of
developers accidentally committing new lex.c files. This doesn't
happen now because the lex.c files are svn:ignore'd.
George: can you setup those dependencies right? If so, I'd be in
favor of committing the generated .c (probably with some comments in
the Makefile.am or somewhere that we'll see them for when someone
modifies the .l someday).
On Jan 26, 2008, at 2:00 PM, George Bosilca wrote:
> In the file directory in the carto framework the carto_file_lex.c is
> included in the svn. This is the only place where this happens. It
> be a good idea, but I think we should be consistent everywhere.
> Therefore, the first question is: should we generate the .c from
> the .l
> every time, or should we provide directly the .c in the svn ? In the
> we choose the second option, I prefer if we use one of the latest (and
> greatest) versions of flex, as it generate more "-pedantic" friendly
> and the code is actually even capable on compiling on Windows.