Re: _WIN32?

[Eli Zaretskii <eliz@is.elta.co.il>]

[I'm resending this because qmail rejected my original message for some 
obscure reason.]

On 17 Apr 2001, DJ Delorie wrote:

> > 	__GO32__	djgpp I think
> 
> DJGPP V1, also defined by DJGPP V2 for backwards compatibility.
> __DJGPP__ is the current recommended symbol.

I don't like changing code just for the sake of changing, that's why I
consistently left the old __GO32__ symbol unless I had a good reason
for changing it into __DJGPP__ (e.g., if DJGPP v1.x behaved
differently from the current DJGPP versions).  You can never know:
someone, somewhere could still use ye olde v1.x of DJGPP.

> > 	__MSDOS__
> 
> DJGPP, Borland C, older MSC's, and other dos-hosted compilers.  This
> is (was?) the canonical way of detecting any dos-hosted compiler
> (compiler-independent filesystem issues, for example)

In some code I contributed, __MSDOS__ is used instead of __DJGPP__
where the underlying issue is not specific to DJGPP, but rather is
common to all MS-DOS-based compilers.  An example is some problem with
limitations of a 8+3 DOS filesystem, e.g. that file names like
`.gdbinit' are not allowed.

By contrast, I use __DJGPP__ for code that won't (or shouldn't) work
with anything but DJGPP.  An example is a snippet in
mant.c:maintenance_dump_me which uses SIGABRT instead of SIGQUIT, for
the reason explained in a comment there.