From: pgarceau AT teleport DOT com (Paul Garceau) Subject: Re: Mingw32 Futures 21 Feb 1998 15:40:05 -0800 Message-ID: <199802210941.BAA04960.cygnus.gnu-win32@smtp3.teleport.com> References: <00b401bd3d04$b4b0efa0$fa173185 AT gbird0 DOT fu DOT is DOT saga-u DOT ac DOT jp> Reply-To: pgarceau AT teleport DOT com To: gnu-win32 AT cygnus DOT com Greetings, On 19 Feb 98 at 16:05, the Illustrious Colin Peters wrote: > Hello all, > > I hope those of you who do not use or have any interest in Mingw32 don't > mind me taking up a bit of bandwidth with strictly Mingw32 issues on this > list. On the other hand I hope that people in positions of > power/authority and/or interested users might take a moment to give their > opinions. > > First of all, I'd like to make a little terminology clear. When I talk > about Mingw32 I mean the header files, configuration files, startup code > and support libraries that allow gcc to build applications which use > CRTDLL.DLL (or possibly other MS supplied C run time library DLLs) as > their C run time library. I would call this your basic Mingw32 distribution set. >Many posters refer to "Mingw32 2.8.0" which is > actually the GNU compiler gcc version 2.8.0 built by Jan-Jaap using the > Mingw32 headers and distributed bundled with those headers. I'd call that > Mingw32 gcc 2.8.0. I would call this a Mingw32 "extension". The version referenced here includes all of the "basic Mingw32 distribution" in much the same way that C++ is a subset of C. Since Jan-Jaap has been maintaining and updating this "extension", the original mingw32 distribution is now capable of handling dx3 and OpenGL in both C and C++ forms. This "extended" version has been qualified as part of the FSF packages. It can also co-exist with cygwin32.dll without any apparent problems so far. I am not sure where Mumit Khans' version fits in here, though I understand that EGCS is supposedly considered the ragged edge of cygwin32. > > That is typical of the "problems" I am having right now. Basically, there > are at least three versions of the gcc compiler distribution which either > have or plan to have Mingw32 integrated into them. Jan jaap's Mingw32 gcc > 2.8.0, Mumit Khan's releases of EGCS for Mingw32, Cygwin32 gcc (perhaps > starting with b20) and maybe the FSF version of gcc. > > I have no problem with this at all. I put that code in the public domain > so people could use it. However, I am basically the defacto maintainer of > the Mingw32 source base until someone tells me otherwise, and so my > questions are something like this: > > 1. Mingw32 is basically a C run time library replacement. As I understand > it gcc is usually bundled with the GNU C library (libc and libm) among > other libraries. Cygwin32's newlib is similar (with a more ambitious > goal). Has anyone seriously thought about how this should fit together? > If *I* thought about it who would I need to talk to about implementing it > (newsgroups? mailing lists?)? The most recent information indicates that gcc/++ 2.8.1 will have the mingw32 headers, etc. (basic Mingw32 distribution) completely integrated as well as full compatibility with the Cygwin32.dll by simply including the cygwin32.dll in the distribution. As far as I can tell, Cygwin32 is basically the first cross-platform extension of gcc/c++ since, as far as I can tell, Cygwin32 takes programs originally coded on Unix or Unix-like (Linux) platforms and allows them to be "more-or-less" directly ported to Wintel based machines such as the WinNT4/5 and Win95/98 based platforms with little or no changes. Mingw32 would not exist if Cygwin32 did not have some sort of previous existence prior to the Mingw32 (v0.4) date of availability. Apparently EGCS requires the "basic" Mingw32 distribution as authored by Colin Peters. The "extended" Mingw32 distribution, as authored by Jan-Jaap, requires the "basic" Mingw32 distribution in order to function properly as far as I can tell. Here's a question for Jan-Jaap: What is the status of the Mingw32 extension in regards to gcc/++ 2.8.1? [Disclaimer: I haven't followed EGCS near as closely as I have Mingw32. Even so, it is my impression that EGCS requires both the cygwin32.dll and the crtdll.dll in order to function properly, "out of the box".] Apparently, all of the Mingw32 variations (including the EGCS variation), regardless of "type", "class" or author, revolve around the existence of the cygwin32 distribution. Without Cygwin32, Mingw32 and EGCS would likely not exist in the form(s) that they do now. This would indicate to me that the best solution may be to issue three variations of cygwin32. a) Cygwin32 with Unix layer (Cygwin32) b) Cygwin32 without Unix layer (Minimalist Cygwin32) c) Cygwin32, the ragged edge (EGCS) Comments, feedback? Peace, Paul G. Information Systems Consultant NewDawn Productions http://www.teleport.com/~pgarceau/taliesin - For help on using this list (especially unsubscribing), send a message to "gnu-win32-request AT cygnus DOT com" with one line of text: "help".