freetype-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ft-devel] Re: bcc and FreeType2 (Re: freetype-2.3.7 -- ftconfig.h for b


From: Antoine Leca
Subject: [ft-devel] Re: bcc and FreeType2 (Re: freetype-2.3.7 -- ftconfig.h for biarch systems)
Date: Tue, 15 Jul 2008 15:11:46 +0200

mpsuzuki wrote:

I welcome your report even if it looks like unsuccessful, because it shows
some important points about Freetype target environments.


> Just I've tried to build FreeType2 for 16bit ELKS,
> with Bruce's C compiler.

[ @Werner: it is definitively a different beast from Borland's bcc, albeit
from the same time frame, i.e. 1990-94 ]

As far as I know, Bruce's [Evans] C compiler is not Ansi-C compliant, so I
guess this is likely to disfunction.


> After the execution of configure in host-building mode,
> I received many errors about the broken preprocessed
> lines.

Sure. Freetype2 makes extensive use of several ANSI extensions (with respect
to original K&R) in the preprocessor area, so these errors are to be
expected.

You might mitigate it if you preprocess the raw source with a compliant cpp
filter; I remember such a setup (using Venema's unproto as pre-filter) was
build for ELKS years ago, in order to make it able to somewhat process ansi
stuff.
Another possibility would be to replace Bruce's preprocessor with the
preprocessor ucpp from T. Pornin (main site http://pornin.nerim.net/ucpp
seems dead, but the code is widely available, for example in any *BSD
repository like ftp://ftp.freebsd.org/pub/FreeBSD/distfiles/ucpp-1.3.tar.gz)
which can be build as a library, and is known to be 16-bit compliant;
perhaps this has already been done, I do not know.

I remember David strongly advised (from experience) _against_ trying to hack
pre-Ansi compilers to work with Freetype, so I can only provide hints, and I
must add I have _no_ experience with the hack suggested above.

As said David, your best chances are first to put your hands at some 16-bit
Ansi compiler.
Unfortunately the obvious (Watcom, Borland, DigitalMars) targets DOS.
Xenix286 C compiler (from Microsoft) seems a bad idea to begin with (not
free, difficult to setup these days). Minix 2 (http://minix1.woodhull.com/
is the best repository) is another option if you prefer Unix command line,
it can run under vmware or similar and does have an ANSI-compliant 16-bit
compiler; Linux (natively) reads or writes Minix floppies or floppy images,
so it is possible to exchange datas. Of course, you will have to severely
prune Freetype, since under Minix-86 you are limited to 64 K *as a whole*
(code _and_ data, library _and_ main program); furthermore Minix is strictly
text-limited, X is not possible. Forget about Minix 3, they dropped 16-bit.
The same ACK compiler used by Minix has been open sourced
(http://tack.sf.net/) and it seems [no personal experience here] to targets
16-bit 8086 and to be usable as cross-compiler hosted eg. by Linux, but I
fear developement is somewhat stalled; furthermore it misses the target
environment altogether.


> I guess FreeType2 developers don't want to rewrite the
> source for such broken cpp.

You are correct: as I wrote, David clearly stated that _no_ effort would be
made to try to support anything which is not Ansi compliant.


Again, thanks for the time you spent in these investigations, and to have
reported them to the community.

Antoine





reply via email to

[Prev in Thread] Current Thread [Next in Thread]