From: Dave Dribin <drib AT enteract DOT com>
Newsgroups: comp.os.msdos.djgpp
Subject: Allegro 3.12: page fault in generate_optimized_palette()
Date: 6 Nov 1999 06:06:48 GMT
Organization: EnterAct L.L.C. Turbo-Elite News Server
Lines: 94
Message-ID: <800gho$aut$1@eve.enteract.com>
NNTP-Posting-Host: 207.229.143.41
User-Agent: tin/pre-1.4-19990624 ("Dawnrazor") (UNIX) (FreeBSD/3.3-STABLE (i386))
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Hello,

I'm getting a page fault in generate_optimized_palette() using Allegro
3.12 (more info on what I'm trying to do below):

----- crash -----
Exiting due to signal SIGSEGV
Page fault at eip=0009e991, error=0004
eax=000000a8 ebx=00000000 ecx=fffe1a34 edx=00251a34 esi=00270000 edi=0026fffc
ebp=00171178 esp=0017116c program=C:\SRC\GLAUNCH\_gl32.exe
cs: sel=0137  base=104c0000  limit=0026ffff
ds: sel=013f  base=104c0000  limit=0026ffff
es: sel=013f  base=104c0000  limit=0026ffff
fs: sel=013f  base=104c0000  limit=0026ffff
gs: sel=014f  base=00000000  limit=0010ffff
ss: sel=013f  base=104c0000  limit=0026ffff
App stack: [001727cc..000f27cc]  Exceptn stack: [000f26ac..000f076c]

Call frame traceback EIPs:
  0x0009e991 ___dj_movedata+37
  0x000268f1 _generate_optimized_palette+2713, line 375 of quantize.c
  0x00001749 show(char *)+245, line 47 of glmain32.cc
  0x00001a8f _main+143, line 147 of glmain32.cc
  0x00092306 ___crt1_startup+174
-----

Looking at ling 375 of my modified quantize.c, it's a memcpy():

----- quantize.c + fprintf()'s -----
   367     for (i=0; i<rsvdcnt; i++) {
   368        for (j=rsvdcnt; j<rsvdcnt+100; j++) {     /* +100 should be enough */
   369               fprintf(stderr,
   370                       "i = %d, j = %d, max = %d, [j]=%08x [i]=%08x [j+1]=%08x\n",
   371                       i, j, max, common[j], common[i], common[j+1]);
   372           if ((tmp = compare_cols(common[i], common[j])) == 0) {
   373               fprintf(stderr, "%08x %08x %d\n", &common[j], &common[j+1],
   374                      (max - j) * 4);
   375              memcpy(&common[j], &common[j + 1], (max - j) * 4);
   376              j--;
   377              max--;
   378           }
   379        }
   380     }
-----

Here's the output from the prints I added:

-----
i = 0, j = 40, max = 42, [j]=0000208e [i]=00000000 [j+1]=0000630e
i = 0, j = 41, max = 42, [j]=0000630e [i]=00000000 [j+1]=00000000
i = 0, j = 42, max = 42, [j]=00000000 [i]=00000000 [j+1]=00000000
00251a34 00251a38 0
i = 0, j = 42, max = 41, [j]=00000000 [i]=00000000 [j+1]=00000000
00251a34 00251a38 -4
Shutting down Allegro
Exiting due to signal SIGSEGV
-----

As you can see (max - j) ends up being negative and causing havoc on
the memcpy().  Is this a known problem?  Am I using the function
wrong?

Here's what I'm trying to do:

I'm running at a depth of 8-bits.  I'd like to reserve about 8 or 9
colors in the palette for stuff that I need.  Yet I still would like
to load in and display a 256-color PCX file.  I cannot just use the
palette from load_bitmap(), 'cause it clobbers the colors I setup in
the palette.  I need to somehow merge the 2 palettes keeping only the
colors that both of them need.  After reading the documentation on
generate_optimized_palette(), this souned perfect.  Fist problem,
g_o_p() doesn't work on 8-bit images.  Doh!  So I get this idea of
converting the PCX image to a 16-bit image, use g_o_p() on the new
bitmap, and then bring it back to 8-bit:

    bmp = load_bitmap(name, pal);
    select_palette(pal);
    bmp16 = create_bitmap_ex(16, bmp->w, bmp->h);
    blit(bmp, bmp16, 0, 0, 0, 0, bmp->w, bmp->h);
    unselect_palette();
    i = generate_optimized_palette(bmp16, pal, my_reserved_colors);
    set_palette(pal);
    blit(bmp16, bmp, 0, 0, 0, 0, bmp->w, bmp->h);
    /* Blit somewhere on screen... */

This works just fine and dandy.....  until one of the colors I reserve
is black!  Then I get the page fault as above.  Can I not reserve
black?  Any ideas?  Is there an easier way to do this?  Can I do it at
all? :)

Thanx for any help,

-Dave