Re: [gfortran,patch] Fix PR libfortran/23262 (on mingw32)

[Tim Prince <tprince at myrealbox dot com>]

François-Xavier Coudert wrote:

> > it may not be appropriate to override the user option --
> > because, if we do, a Cygwin user may not get the line ending type that
> > they want.
> >    
>
> I do agree with that. On the other hand, libgfortran is incompatible
> with CRLF switch (that is, today, if a user specifies CRLF, gfortran
> (indepedently of where it was compiled) will not work for him. Trivial
> codes might appear to work, but any code using (for example) rewind or
> sequential direct access will be broken (see PR 23262 and duplicates).
>
> I see no way to have the Cygwin switch influence the behaviour of the
> library. On the other hand, we could add a compiler switch, -fcrlf,
> which would default to -fno-crlf on non-windows systems, to -fcrlf on
> mingw and either (choice of the binaries maker) on cygwin.
>
> How do people feel about that?
>
>  
>
>  
I never have cared for the text mount options in cygwin.  The only 
accommodation for crlf which makes sense to me in Fortran is to treat 
<cr> and <crlf> the same at end of record, for formatted files only, as 
many other compilers do.  It is extra important that gfortran be able to 
make files which are neutral between Windows and linux.