Author: Christophe Theron
Date: 06:46:14 08/02/98
Go up one level in this thread
On August 01, 1998 at 18:32:18, Dan Homan wrote:
>On August 01, 1998 at 10:34:52, Christophe Theron wrote:
>>On July 31, 1998 at 19:08:10, g.müller wrote:
>>
>>>Who works under Borland c++ 5.02 with Chess Programming please contact me.
>>>I work with V 3.1 and now I have big problems to transport source to V 5.02!!
>>>My platform is win 98 with PII 400 and 128 MB.
>>
>>I have NEVER been able to port Chess Tiger under Borland C++ 5.02 because of
>>horrible compiler and runtime library bugs.
>>
>I use Borland C++ 5.02 to make my Win32 executable and have had no
>trouble with it. My program compiled straight away with no trouble.
>I also compile my program under DJGPP and Digital C++. I have not
>tried compiling 16 bit DOS code in Borland 5.02, but 32 bit code
>(at least mine) seems to give no trouble as a Win32 console app.
>Borland c++ 5.02 seems to build a faster executable than DJGPP but
>I can't compare to Digital C++ as that is on an Alpha.
>
> - Dan
>
>P.S. I agree that there are better choices than Borland 5.02, but it
>is all I have for making Win32 apps.
I really hate to say it, because I have been a long time user of Borland
products, but I was *very* disappointed by BC5.02.
I am wondering how you can work with this compiler.
BC5.02 includes 2 compilers. The first one is the fast usual Borland compiler,
and the second is an optimizing compiler provided by Intel.
I tried to compile Chess Tiger with both. Here is what happened:
* The usual fast Borland compiler crashed when compiling a very long function
in Chess Tiger's positional evaluator. Something like a stack overflow I think.
It is true that I could have broken this function into smaller parts, but I had
previoulsy compiled this without any problem with 5 other compilers, including
Borland 4.0 and 4.5! And the long function is not a complex one. It is just
long, but not tricky.
* The Intel optimizing compiler produced wrong assembly code in the isdigit()
function! isdigit is expanded inline (this is not a call to a run time library),
and the assembly code was doing one indirection in excess, thus causing a page
fault!
After having these problems, I was seriously wondering if any of the Borland
testers had ever really compiled anything with their new stuff.
I didn't give up at that stage in fact. I downloaded 2 patches from Borland's
internet site. It took me at least 2 hours. I applied the patches, but the
problems were still there.
Then I tried to completely uninstall the compiler, clean my disk, and install
again. It didn't change anything.
After 2 or 3 lost days, I decided to throw out the compiler. I had already lost
too much time with it. You can say I could have simplified my long function. The
problem is that my trust in Borland is completely gone now. I already had
problems with BC3.1 (the compiler crashes very often), more problems with BC4
and Windows 95, and then too much trouble with BC5.02. It looks like every new
version is worse than the previous one. Borland is not going to have me as a
customer again.
Christophe
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.