Mail Archives: djgpp/2013/05/25/14:45:11
Hi,
On Saturday, May 25, 2013 11:25:39 AM UTC-5, Eli Zaretskii wrote:
> > Date: Sat, 25 May 2013 08:56:19 -0700 (PDT)
>
> > From: Georg Potthast <dosusb AT nospamforgooglemail DOT moc>
> >
> > As long as you do not access the hardware directly it is really
> > easy to port a DJGPP program to MinGW.
>
> Did you actually try that? My experience in porting to MinGW is
> exactly the opposite: DJGPP has a lot of Posix-like features that
> MinGW sorely lacks. So porting to MinGW using a DJGPP port as a
> starting point will generally give you a broken port, and in many
> cases will simply refuse to compile or link.
IIRC, MinGW isn't explicitly targeting POSIX, only "native" Windows,
hence the avoidance of any Cygwin .DLL (also for licensing?) and
using "known .DLL" MSVCRT.DLL instead (despite bugs, e.g. tmpfile).
Actually, it's only the old MSVC 6 runtime, not newer, that they're
using, IIRC. Though a lot of MinGW usage is also because MSVC
doesn't really support C99, which MinGW seems to prefer.
> Besides, starting with DJGPP will automatically lose the advanced
> features you can have with Windows: networking, threads, parallel
> processes, etc.
If you're targeting all of that anyways, you're not really targeting
pure ISO C, are you? Almost better to "just use Java" (or Modula-3).
> DJ's suggestion to use MinGW is still valid, of course (although
> MinGW still doesn't support generation of 64-bit executables; you
> need to go to semi-official MinGW64 snapshots). But please don't
> underestimate the efforts required for porting a non-trivial
> package to MinGW. Heck, even running a configure script is a
> challenge, and requires an installation of yet another environment
> (MSYS).
Blame those who refuse to code anything outside of POSIX. Obviously
AutoTools was never expected to work on systems without a native
POSIX shell. It might be (barely) wrong to say outside developers
don't care about Windows, but clearly it's not first priority target.
- Raw text -