From:	sime AT fly DOT cc DOT fer DOT hr (S. Mikecin)
Newsgroups:	comp.os.msdos.djgpp
Subject:	Re: QEMM & DJGPP new/delete bug?
Date:	26 Nov 1996 12:14:12 GMT
Organization:	FER, Croatia
Lines:	86
Distribution:	inet
Message-ID:	<57emuk$qog@bagan.srce.hr>
References:	<32969DB7 DOT 3128 AT nlc DOT net DOT au>
Reply-To:	sime AT fly DOT cc DOT etf DOT hr
NNTP-Posting-Host:	fly.cc.etf.hr
Mime-Version:	1.0
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp

shaman AT nlc DOT net DOT au wrote:
: I've come across an interesting "bug". I don't know what is causing it
: exactly. Actually, I wouldn't have a clue what is causing it. Is this a
: bug in my code? A problem with DJGPP? My memory manager? I've seen this
: before, but this is the first time I've seen it happen with a small
: enough piece of code to nail it to new and delete. If it's a bug in
: DJGPP, is it fixed in 2.01? PLEASE REPLY BY EMAIL.

: WHAT I'M DOING: Allocating large amounts of ram (4megs+) for zbuffers
: and frame buffers.

: THE SYMPTOMS: Simple test program runs okay in a totally clean boot and
: CWSPMI, but crashes if run more than a couple of times under QEMM 8.0
: and CWSDPMI. QDPMI is not loaded. QEMM is in "ON" mode, not AUTO. It
: crashes faster if I also run ramdisk (~3M) and smartdrv(~2M). I have an
: AMD486DX2/80, DOS 6.22 and 16 megs of ram. It also can't allocate any
: more than half of the available ram, even though I'm not running WIN95,
: and I'm not allocating it in one chunk. (I'm allocating in just over 1
: meg pieces).

: METHOD: I wrote simple classes for the BUFFER and ZBUFFER types, which
: are self allocating arrays of unsigned longs. (Source for ctors and
: dtors below). The program allocates space for them using new, then
: erases them using delete as would be normal.

: SOURCE: 

: // typedef unsigned long ulong; 
: // ulong*	data;
: // ulong**	ytable;

: BUFFER::BUFFER( ulong w, ulong h )
: {
: 	_width = w;	_height = h;

: 	data 	= new ulong	[_width*_height]; // Alloc memory for bitmap
: 	ytable 	= new ulong*	[_height];	  // Then allocate for y table.

:     // precalc y lookup table.

: 	ulong* prev = data;

:     for ( int i = 0; i < _height; i++ )
:     {
:     	ytable[i] = prev;
:         prev += _width;		// increment pointer for next line.
:     }

: }

: ZBUFFER::ZBUFFER() is exactly the same as BUFFER::BUFFER()

: BUFFER::~BUFFER()
: {
: 	delete ytable;
: 	delete data;
: }

: ZBUFFER::~ZBUFFER is the same.


: This is the part in the test program:

: 	...
: 	BUFFER* test = new BUFFER( 1024, 1024 );		// 1 meg.
:     	delete test;
: 	...

: -- 
:                                        . . .   the Lord Shaman

:       
: ------------------------------------------------------------------
:           http://www.nlc.net.au/~shaman  or  mailto:shaman AT nlc DOT net DOT au
:       
: ------------------------------------------------------------------

I have QEMM 8.0 & CWSDPMI v3, ON 486DX4/120 with 12Megs of RAM
(2xDiskCache+1xRAMDisk) and didn't have any problems with it. I allocated
all memory and more (virtual). Actually my QEMM was patched with
800to801.exe so the version is actually 8.01.
--

                                          <E-mail:     sime AT fly DOT cc DOT fer DOT hr    >
                                          <2nd E-mail: simun DOT mikecin AT fer DOT hr  >
                                          <URL:    http://fly.cc.fer.hr/~sime>

- Raw text -