Please hand this over to David Gravereau or Jeff Hobbs.
I think I submitted a patch sometime in August, before
I took some time "off-line". I don't find it and the
sources I downloaded from SF yesterday still don't
build with Borland - so I guess the patch didn't make it.
Anyway, the one enclosed in the diff file is better
than the original, as far as patching tcl.h is
concerned (IMHO):
Currently checking for windows compilers is done like this:
#ifndef __WIN32__
... check for other "Win32" symbols or the Borland compiler
... if ok
#define __WIN32__
and - if not already done so -
#define WIN32
which is used elsewhere in the sources.
Now, Borland already #defines __WIN32__, so this block
is skipped, leaving WIN32 #undefined - which will cause
problems later on.
I propose instead to test for any of the symbols,
Windows compilers use to identify the platform - see
the diff file for details. This eliminates an explicit
test for _BORLANDC_ - just like tcl.h contains no
explicit test for MSC_VER at this point anymore.
Best regards
Helmut Giese
Diff file to build Tcl 8.4 with Borland
Logged In: YES
user_id=7549
working on the 8.5 patches instead.