mirror of
https://github.com/DrBeef/Raze.git
synced 2024-12-14 22:51:42 +00:00
e979aa1ff4
git-svn-id: https://svn.eduke32.com/eduke32@6489 1a8010ca-5511-0410-912e-c29ae57300e0
1960 lines
55 KiB
C++
1960 lines
55 KiB
C++
// "Build Engine & Tools" Copyright (c) 1993-1997 Ken Silverman
|
|
// Ken Silverman's official web site: "http://www.advsys.net/ken"
|
|
// See the included license file "BUILDLIC.TXT" for license info.
|
|
//
|
|
// This file has been modified from Ken Silverman's original release
|
|
// by Jonathon Fowler (jf@jonof.id.au)
|
|
|
|
#include "compat.h"
|
|
|
|
#ifdef CACHE1D_COMPRESS_ONLY
|
|
// Standalone libcache1d.so containing only the compression/decompression
|
|
// functions.
|
|
# define C1D_STATIC
|
|
#else
|
|
// cache1d.o for EDuke32
|
|
# define C1D_STATIC static
|
|
|
|
#ifdef _WIN32
|
|
// for FILENAME_CASE_CHECK
|
|
# define NEED_SHELLAPI_H
|
|
# include "windows_inc.h"
|
|
#endif
|
|
#include "cache1d.h"
|
|
#include "pragmas.h"
|
|
#include "baselayer.h"
|
|
|
|
#ifdef WITHKPLIB
|
|
#include "kplib.h"
|
|
|
|
//Insert '|' in front of filename
|
|
//Doing this tells kzopen to load the file only if inside a .ZIP file
|
|
static intptr_t kzipopen(const char *filnam)
|
|
{
|
|
uint32_t i;
|
|
char newst[BMAX_PATH+8];
|
|
|
|
newst[0] = '|';
|
|
for (i=0; filnam[i] && (i < sizeof(newst)-2); i++) newst[i+1] = filnam[i];
|
|
newst[i+1] = 0;
|
|
return kzopen(newst);
|
|
}
|
|
|
|
#endif
|
|
|
|
char *kpzbuf = NULL;
|
|
int32_t kpzbufsiz;
|
|
|
|
int32_t kpzbufloadfil(int32_t const handle)
|
|
{
|
|
int32_t const leng = kfilelength(handle);
|
|
if (leng > kpzbufsiz)
|
|
{
|
|
kpzbuf = (char *) Xrealloc(kpzbuf, leng+1);
|
|
kpzbufsiz = leng;
|
|
if (!kpzbuf)
|
|
return 0;
|
|
}
|
|
|
|
kpzbuf[leng] = 0; // FIXME: buf[leng] read in kpegrend(), see BUF_LENG_READ
|
|
kread(handle, kpzbuf, leng);
|
|
|
|
return leng;
|
|
}
|
|
|
|
int32_t kpzbufload(char const * const filnam)
|
|
{
|
|
int32_t const handle = kopen4load(filnam, 0);
|
|
if (handle < 0)
|
|
return 0;
|
|
|
|
int32_t const leng = kpzbufloadfil(handle);
|
|
|
|
kclose(handle);
|
|
|
|
return leng;
|
|
}
|
|
|
|
// This module keeps track of a standard linear cacheing system.
|
|
// To use this module, here's all you need to do:
|
|
//
|
|
// Step 1: Allocate a nice BIG buffer, like from 1MB-4MB and
|
|
// Call initcache(int32_t cachestart, int32_t cachesize) where
|
|
//
|
|
// cachestart = (intptr_t)(pointer to start of BIG buffer)
|
|
// cachesize = length of BIG buffer
|
|
//
|
|
// Step 2: Call allocache(intptr_t *bufptr, int32_t bufsiz, char *lockptr)
|
|
// whenever you need to allocate a buffer, where:
|
|
//
|
|
// *bufptr = pointer to multi-byte pointer to buffer
|
|
// Confused? Using this method, cache2d can remove
|
|
// previously allocated things from the cache safely by
|
|
// setting the multi-byte pointer to 0.
|
|
// bufsiz = number of bytes to allocate
|
|
// *lockptr = pointer to locking char which tells whether
|
|
// the region can be removed or not. If *lockptr = 0 then
|
|
// the region is not locked else its locked.
|
|
//
|
|
// Step 3: If you need to remove everything from the cache, or every
|
|
// unlocked item from the cache, you can call uninitcache();
|
|
// Call uninitcache(0) to remove all unlocked items, or
|
|
// Call uninitcache(1) to remove everything.
|
|
// After calling uninitcache, it is still ok to call allocache
|
|
// without first calling initcache.
|
|
|
|
#define MAXCACHEOBJECTS 9216
|
|
|
|
#if !defined DEBUG_ALLOCACHE_AS_MALLOC
|
|
static int32_t cachesize = 0;
|
|
static char zerochar = 0;
|
|
static intptr_t cachestart = 0;
|
|
static int32_t agecount = 0;
|
|
static int32_t lockrecip[200];
|
|
|
|
int32_t cacnum = 0;
|
|
cactype cac[MAXCACHEOBJECTS];
|
|
#endif
|
|
|
|
char toupperlookup[256] =
|
|
{
|
|
0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07,0x08,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,
|
|
0x10,0x11,0x12,0x13,0x14,0x15,0x16,0x17,0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f,
|
|
0x20,0x21,0x22,0x23,0x24,0x25,0x26,0x27,0x28,0x29,0x2a,0x2b,0x2c,0x2d,0x2e,0x2f,
|
|
0x30,0x31,0x32,0x33,0x34,0x35,0x36,0x37,0x38,0x39,0x3a,0x3b,0x3c,0x3d,0x3e,0x3f,
|
|
0x40,0x41,0x42,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,
|
|
0x50,0x51,0x52,0x53,0x54,0x55,0x56,0x57,0x58,0x59,0x5a,0x5b,0x5c,0x5d,0x5e,0x5f,
|
|
0x60,0x41,0x42,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,
|
|
0x50,0x51,0x52,0x53,0x54,0x55,0x56,0x57,0x58,0x59,0x5a,0x7b,0x7c,0x7d,0x7e,0x7f,
|
|
0x80,0x81,0x82,0x83,0x84,0x85,0x86,0x87,0x88,0x89,0x8a,0x8b,0x8c,0x8d,0x8e,0x8f,
|
|
0x90,0x91,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0x9b,0x9c,0x9d,0x9e,0x9f,
|
|
0xa0,0xa1,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xab,0xac,0xad,0xae,0xaf,
|
|
0xb0,0xb1,0xb2,0xb3,0xb4,0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xbb,0xbc,0xbd,0xbe,0xbf,
|
|
0xc0,0xc1,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xcb,0xcc,0xcd,0xce,0xcf,
|
|
0xd0,0xd1,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xdb,0xdc,0xdd,0xde,0xdf,
|
|
0xe0,0xe1,0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xeb,0xec,0xed,0xee,0xef,
|
|
0xf0,0xf1,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa,0xfb,0xfc,0xfd,0xfe,0xff
|
|
};
|
|
|
|
static void reportandexit(const char *errormessage);
|
|
|
|
|
|
void initcache(intptr_t dacachestart, int32_t dacachesize)
|
|
{
|
|
#ifndef DEBUG_ALLOCACHE_AS_MALLOC
|
|
int32_t i;
|
|
|
|
for (i=1; i<200; i++)
|
|
lockrecip[i] = tabledivide32_noinline(1<<28, 200-i);
|
|
|
|
// The following code was relocated here from engine.c, since this
|
|
// function is only ever called once (from there), and it seems to
|
|
// really belong here:
|
|
//
|
|
// initcache((FP_OFF(pic)+15)&0xfffffff0,(cachesize-((-FP_OFF(pic))&15))&0xfffffff0);
|
|
//
|
|
// I'm not sure why it's necessary, but the code is making sure the
|
|
// cache starts on a multiple of 16 bytes? -- SA
|
|
|
|
//printf("BEFORE: cachestart = %x, cachesize = %d\n", dacachestart, dacachesize);
|
|
cachestart = ((uintptr_t)dacachestart+15)&~(uintptr_t)0xf;
|
|
cachesize = (dacachesize-(((uintptr_t)(dacachestart))&0xf))&~(uintptr_t)0xf;
|
|
//printf("AFTER : cachestart = %x, cachesize = %d\n", cachestart, cachesize);
|
|
|
|
cac[0].leng = cachesize;
|
|
cac[0].lock = &zerochar;
|
|
cacnum = 1;
|
|
|
|
initprintf("Initialized %.1fM cache\n", (float)(dacachesize/1024.f/1024.f));
|
|
#else
|
|
UNREFERENCED_PARAMETER(dacachestart);
|
|
UNREFERENCED_PARAMETER(dacachesize);
|
|
#endif
|
|
}
|
|
|
|
#ifdef DEBUG_ALLOCACHE_AS_MALLOC
|
|
void allocache(intptr_t *newhandle, int32_t newbytes, char *newlockptr)
|
|
{
|
|
UNREFERENCED_PARAMETER(newlockptr);
|
|
|
|
*newhandle = (intptr_t)Xmalloc(newbytes);
|
|
}
|
|
#else
|
|
static inline void inc_and_check_cacnum(void)
|
|
{
|
|
if (++cacnum > MAXCACHEOBJECTS)
|
|
reportandexit("Too many objects in cache! (cacnum > MAXCACHEOBJECTS)");
|
|
}
|
|
|
|
void allocache(intptr_t *newhandle, int32_t newbytes, char *newlockptr)
|
|
{
|
|
int32_t i, z, bestz=0, bestval, besto=0, o1, sucklen, suckz;
|
|
|
|
//printf(" ==> asking for %d bytes, ", newbytes);
|
|
// Make all requests a multiple of 16 bytes
|
|
newbytes = (newbytes+15)&0xfffffff0;
|
|
//printf("allocated %d bytes\n", newbytes);
|
|
|
|
if ((unsigned)newbytes > (unsigned)cachesize)
|
|
{
|
|
Bprintf("Cachesize: %d\n",cachesize);
|
|
Bprintf("*Newhandle: 0x%" PRIxPTR ", Newbytes: %d, *Newlock: %d\n",(intptr_t)newhandle,newbytes,*newlockptr);
|
|
reportandexit("BUFFER TOO BIG TO FIT IN CACHE!");
|
|
}
|
|
|
|
if (*newlockptr == 0)
|
|
{
|
|
reportandexit("ALLOCACHE CALLED WITH LOCK OF 0!");
|
|
}
|
|
|
|
//Find best place
|
|
bestval = 0x7fffffff; o1 = cachesize;
|
|
for (z=cacnum-1; z>=0; z--)
|
|
{
|
|
int32_t zz, o2, daval;
|
|
|
|
o1 -= cac[z].leng;
|
|
o2 = o1+newbytes;
|
|
|
|
if (o2 > cachesize)
|
|
continue;
|
|
|
|
daval = 0;
|
|
for (i=o1,zz=z; i<o2; i+=cac[zz++].leng)
|
|
{
|
|
if (*cac[zz].lock == 0)
|
|
continue;
|
|
|
|
if (*cac[zz].lock >= 200)
|
|
{
|
|
daval = 0x7fffffff;
|
|
break;
|
|
}
|
|
|
|
// Potential for eviction increases with
|
|
// - smaller item size
|
|
// - smaller lock byte value (but in [1 .. 199])
|
|
daval += mulscale32(cac[zz].leng+65536, lockrecip[*cac[zz].lock]);
|
|
if (daval >= bestval)
|
|
break;
|
|
}
|
|
|
|
if (daval < bestval)
|
|
{
|
|
bestval = daval; besto = o1; bestz = z;
|
|
if (bestval == 0) break;
|
|
}
|
|
}
|
|
|
|
//printf("%d %d %d\n",besto,newbytes,*newlockptr);
|
|
|
|
if (bestval == 0x7fffffff)
|
|
reportandexit("CACHE SPACE ALL LOCKED UP!");
|
|
|
|
//Suck things out
|
|
for (sucklen=-newbytes,suckz=bestz; sucklen<0; sucklen+=cac[suckz++].leng)
|
|
if (*cac[suckz].lock)
|
|
*cac[suckz].hand = 0;
|
|
|
|
//Remove all blocks except 1
|
|
suckz -= bestz+1;
|
|
cacnum -= suckz;
|
|
|
|
Bmemmove(&cac[bestz], &cac[bestz+suckz], (cacnum-bestz)*sizeof(cactype));
|
|
cac[bestz].hand = newhandle;
|
|
*newhandle = cachestart + besto;
|
|
cac[bestz].leng = newbytes;
|
|
cac[bestz].lock = newlockptr;
|
|
|
|
//Add new empty block if necessary
|
|
if (sucklen <= 0)
|
|
return;
|
|
|
|
bestz++;
|
|
if (bestz == cacnum)
|
|
{
|
|
inc_and_check_cacnum();
|
|
|
|
cac[bestz].leng = sucklen;
|
|
cac[bestz].lock = &zerochar;
|
|
return;
|
|
}
|
|
|
|
if (*cac[bestz].lock == 0)
|
|
{
|
|
cac[bestz].leng += sucklen;
|
|
return;
|
|
}
|
|
|
|
inc_and_check_cacnum();
|
|
|
|
for (z=cacnum-1; z>bestz; z--)
|
|
cac[z] = cac[z-1];
|
|
cac[bestz].leng = sucklen;
|
|
cac[bestz].lock = &zerochar;
|
|
}
|
|
#endif
|
|
|
|
void agecache(void)
|
|
{
|
|
#ifndef DEBUG_ALLOCACHE_AS_MALLOC
|
|
bssize_t cnt = (cacnum>>4);
|
|
|
|
if (agecount >= cacnum)
|
|
agecount = cacnum-1;
|
|
|
|
if (agecount < 0 || !cnt)
|
|
return;
|
|
|
|
for (; cnt>=0; cnt--)
|
|
{
|
|
// If we have pointer to lock char and it's in [2 .. 199], decrease.
|
|
if (cac[agecount].lock && (((*cac[agecount].lock)-2)&255) < 198)
|
|
(*cac[agecount].lock)--;
|
|
|
|
agecount--;
|
|
if (agecount < 0)
|
|
agecount = cacnum-1;
|
|
}
|
|
#endif
|
|
}
|
|
|
|
static void reportandexit(const char *errormessage)
|
|
{
|
|
#ifndef DEBUG_ALLOCACHE_AS_MALLOC
|
|
int32_t i, j;
|
|
|
|
//setvmode(0x3);
|
|
j = 0;
|
|
for (i=0; i<cacnum; i++)
|
|
{
|
|
Bprintf("%d- ",i);
|
|
if (cac[i].hand) Bprintf("ptr: 0x%" PRIxPTR ", ",*cac[i].hand);
|
|
else Bprintf("ptr: NULL, ");
|
|
Bprintf("leng: %d, ",cac[i].leng);
|
|
if (cac[i].lock) Bprintf("lock: %d\n",*cac[i].lock);
|
|
else Bprintf("lock: NULL\n");
|
|
j += cac[i].leng;
|
|
}
|
|
Bprintf("Cachesize = %d\n",cachesize);
|
|
Bprintf("Cacnum = %d\n",cacnum);
|
|
Bprintf("Cache length sum = %d\n",j);
|
|
#endif
|
|
initprintf("ERROR: %s\n",errormessage);
|
|
Bexit(1);
|
|
}
|
|
|
|
#include <errno.h>
|
|
|
|
typedef struct _searchpath
|
|
{
|
|
struct _searchpath *next;
|
|
char *path;
|
|
size_t pathlen; // to save repeated calls to strlen()
|
|
int32_t user;
|
|
} searchpath_t;
|
|
static searchpath_t *searchpathhead = NULL;
|
|
static size_t maxsearchpathlen = 0;
|
|
int32_t pathsearchmode = 0;
|
|
|
|
char *listsearchpath(int32_t initp)
|
|
{
|
|
static searchpath_t *sp;
|
|
|
|
if (initp)
|
|
sp = searchpathhead;
|
|
else if (sp != NULL)
|
|
sp = sp->next;
|
|
|
|
return sp ? sp->path : NULL;
|
|
}
|
|
|
|
int32_t addsearchpath_user(const char *p, int32_t user)
|
|
{
|
|
struct Bstat st;
|
|
char *s;
|
|
searchpath_t *srch;
|
|
char *path = Xstrdup(p);
|
|
|
|
if (path[Bstrlen(path)-1] == '\\')
|
|
path[Bstrlen(path)-1] = 0; // hack for stat() returning ENOENT on paths ending in a backslash
|
|
|
|
if (Bstat(path, &st) < 0)
|
|
{
|
|
Bfree(path);
|
|
if (errno == ENOENT) return -2;
|
|
return -1;
|
|
}
|
|
if (!(st.st_mode & BS_IFDIR))
|
|
{
|
|
Bfree(path);
|
|
return -1;
|
|
}
|
|
|
|
srch = (searchpath_t *)Xmalloc(sizeof(searchpath_t));
|
|
|
|
srch->next = searchpathhead;
|
|
srch->pathlen = Bstrlen(path)+1;
|
|
srch->path = (char *)Xmalloc(srch->pathlen + 1);
|
|
|
|
Bstrcpy(srch->path, path);
|
|
for (s=srch->path; *s; s++) { }
|
|
s--;
|
|
|
|
if (s<srch->path || toupperlookup[*s] != '/')
|
|
Bstrcat(srch->path, "/");
|
|
|
|
searchpathhead = srch;
|
|
if (srch->pathlen > maxsearchpathlen)
|
|
maxsearchpathlen = srch->pathlen;
|
|
|
|
Bcorrectfilename(srch->path,0);
|
|
|
|
srch->user = user;
|
|
|
|
initprintf("Using %s for game data\n", srch->path);
|
|
|
|
Bfree(path);
|
|
return 0;
|
|
}
|
|
|
|
int32_t removesearchpath(const char *p)
|
|
{
|
|
searchpath_t *srch;
|
|
char *s;
|
|
char *path = (char *)Xmalloc(Bstrlen(p) + 2);
|
|
|
|
Bstrcpy(path, p);
|
|
|
|
if (path[Bstrlen(path)-1] == '\\')
|
|
path[Bstrlen(path)-1] = 0;
|
|
|
|
for (s=path; *s; s++) { }
|
|
s--;
|
|
|
|
if (s<path || toupperlookup[*s] != '/')
|
|
Bstrcat(path, "/");
|
|
|
|
Bcorrectfilename(path,0);
|
|
|
|
for (srch = searchpathhead; srch; srch = srch->next)
|
|
{
|
|
if (!Bstrncmp(path, srch->path, srch->pathlen))
|
|
{
|
|
// initprintf("Removing %s from path stack\n", path);
|
|
|
|
if (srch == searchpathhead)
|
|
searchpathhead = srch->next;
|
|
else
|
|
{
|
|
searchpath_t *sp;
|
|
|
|
for (sp = searchpathhead; sp; sp = sp->next)
|
|
{
|
|
if (sp->next == srch)
|
|
{
|
|
// initprintf("matched %s\n", srch->path);
|
|
sp->next = srch->next;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
Bfree(srch->path);
|
|
Bfree(srch);
|
|
break;
|
|
}
|
|
}
|
|
|
|
Bfree(path);
|
|
return 0;
|
|
}
|
|
|
|
void removesearchpaths_withuser(int32_t usermask)
|
|
{
|
|
searchpath_t *next;
|
|
|
|
for (searchpath_t *srch = searchpathhead; srch; srch = next)
|
|
{
|
|
next = srch->next;
|
|
|
|
if (srch->user & usermask)
|
|
{
|
|
|
|
if (srch == searchpathhead)
|
|
searchpathhead = srch->next;
|
|
else
|
|
{
|
|
searchpath_t *sp;
|
|
|
|
for (sp = searchpathhead; sp; sp = sp->next)
|
|
{
|
|
if (sp->next == srch)
|
|
{
|
|
sp->next = srch->next;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
Bfree(srch->path);
|
|
Bfree(srch);
|
|
}
|
|
}
|
|
}
|
|
|
|
int32_t findfrompath(const char *fn, char **where)
|
|
{
|
|
// pathsearchmode == 0: tests current dir and then the dirs of the path stack
|
|
// pathsearchmode == 1: tests fn without modification, then like for pathsearchmode == 0
|
|
|
|
if (pathsearchmode)
|
|
{
|
|
// test unmolested filename first
|
|
if (access(fn, F_OK) >= 0)
|
|
{
|
|
*where = Xstrdup(fn);
|
|
return 0;
|
|
}
|
|
#ifndef _WIN32
|
|
else
|
|
{
|
|
char *tfn = Bstrtolower(Xstrdup(fn));
|
|
|
|
if (access(tfn, F_OK) >= 0)
|
|
{
|
|
*where = tfn;
|
|
return 0;
|
|
}
|
|
|
|
Bstrupr(tfn);
|
|
|
|
if (access(tfn, F_OK) >= 0)
|
|
{
|
|
*where = tfn;
|
|
return 0;
|
|
}
|
|
|
|
Bfree(tfn);
|
|
}
|
|
#endif
|
|
}
|
|
|
|
char const *cpfn;
|
|
|
|
for (cpfn = fn; toupperlookup[*cpfn] == '/'; cpfn++) { }
|
|
char *ffn = Xstrdup(cpfn);
|
|
|
|
Bcorrectfilename(ffn,0); // compress relative paths
|
|
|
|
int32_t allocsiz = max(maxsearchpathlen, 2); // "./" (aka. curdir)
|
|
allocsiz += strlen(ffn);
|
|
allocsiz += 1; // a nul
|
|
|
|
char *pfn = (char *)Xmalloc(allocsiz);
|
|
|
|
strcpy(pfn, "./");
|
|
strcat(pfn, ffn);
|
|
if (access(pfn, F_OK) >= 0)
|
|
{
|
|
*where = pfn;
|
|
Bfree(ffn);
|
|
return 0;
|
|
}
|
|
|
|
for (searchpath_t *sp = searchpathhead; sp; sp = sp->next)
|
|
{
|
|
char *tfn = Xstrdup(ffn);
|
|
|
|
strcpy(pfn, sp->path);
|
|
strcat(pfn, ffn);
|
|
//initprintf("Trying %s\n", pfn);
|
|
if (access(pfn, F_OK) >= 0)
|
|
{
|
|
*where = pfn;
|
|
Bfree(ffn);
|
|
Bfree(tfn);
|
|
return 0;
|
|
}
|
|
|
|
#ifndef _WIN32
|
|
//Check with all lowercase
|
|
strcpy(pfn, sp->path);
|
|
Bstrtolower(tfn);
|
|
strcat(pfn, tfn);
|
|
if (access(pfn, F_OK) >= 0)
|
|
{
|
|
*where = pfn;
|
|
Bfree(ffn);
|
|
Bfree(tfn);
|
|
return 0;
|
|
}
|
|
|
|
//Check again with uppercase
|
|
strcpy(pfn, sp->path);
|
|
Bstrupr(tfn);
|
|
strcat(pfn, tfn);
|
|
if (access(pfn, F_OK) >= 0)
|
|
{
|
|
*where = pfn;
|
|
Bfree(ffn);
|
|
Bfree(tfn);
|
|
return 0;
|
|
}
|
|
#endif
|
|
Bfree(tfn);
|
|
}
|
|
|
|
Bfree(pfn); Bfree(ffn);
|
|
return -1;
|
|
}
|
|
|
|
#if defined(_WIN32) && defined(DEBUGGINGAIDS)
|
|
# define FILENAME_CASE_CHECK
|
|
#endif
|
|
|
|
static int32_t openfrompath_internal(const char *fn, char **where, int32_t flags, int32_t mode)
|
|
{
|
|
if (findfrompath(fn, where) < 0)
|
|
return -1;
|
|
|
|
return Bopen(*where, flags, mode);
|
|
}
|
|
|
|
int32_t openfrompath(const char *fn, int32_t flags, int32_t mode)
|
|
{
|
|
char *pfn = NULL;
|
|
|
|
int32_t h = openfrompath_internal(fn, &pfn, flags, mode);
|
|
|
|
Bfree(pfn);
|
|
|
|
return h;
|
|
}
|
|
|
|
BFILE *fopenfrompath(const char *fn, const char *mode)
|
|
{
|
|
int32_t fh;
|
|
BFILE *h;
|
|
int32_t bmode = 0, smode = 0;
|
|
const char *c;
|
|
|
|
for (c=mode; c[0];)
|
|
{
|
|
if (c[0] == 'r' && c[1] == '+') { bmode = BO_RDWR; smode = BS_IREAD|BS_IWRITE; c+=2; }
|
|
else if (c[0] == 'r') { bmode = BO_RDONLY; smode = BS_IREAD; c+=1; }
|
|
else if (c[0] == 'w' && c[1] == '+') { bmode = BO_RDWR|BO_CREAT|BO_TRUNC; smode = BS_IREAD|BS_IWRITE; c+=2; }
|
|
else if (c[0] == 'w') { bmode = BO_WRONLY|BO_CREAT|BO_TRUNC; smode = BS_IREAD|BS_IWRITE; c+=2; }
|
|
else if (c[0] == 'a' && c[1] == '+') { bmode = BO_RDWR|BO_CREAT; smode=BS_IREAD|BS_IWRITE; c+=2; }
|
|
else if (c[0] == 'a') { bmode = BO_WRONLY|BO_CREAT; smode=BS_IREAD|BS_IWRITE; c+=1; }
|
|
else if (c[0] == 'b') { bmode |= BO_BINARY; c+=1; }
|
|
else if (c[1] == 't') { bmode |= BO_TEXT; c+=1; }
|
|
else c++;
|
|
}
|
|
fh = openfrompath(fn,bmode,smode);
|
|
if (fh < 0) return NULL;
|
|
|
|
h = fdopen(fh,mode);
|
|
if (!h) close(fh);
|
|
|
|
return h;
|
|
}
|
|
|
|
#define MAXGROUPFILES 8 // Warning: Fix groupfil if this is changed
|
|
#define MAXOPENFILES 64 // Warning: Fix filehan if this is changed
|
|
|
|
enum {
|
|
GRP_RESERVED_ID_START = 254,
|
|
|
|
GRP_ZIP = GRP_RESERVED_ID_START,
|
|
GRP_FILESYSTEM = GRP_RESERVED_ID_START + 1,
|
|
};
|
|
|
|
EDUKE32_STATIC_ASSERT(MAXGROUPFILES <= GRP_RESERVED_ID_START);
|
|
|
|
int32_t numgroupfiles = 0;
|
|
static int32_t gnumfiles[MAXGROUPFILES];
|
|
static intptr_t groupfil[MAXGROUPFILES] = {-1,-1,-1,-1,-1,-1,-1,-1};
|
|
static int32_t groupfilpos[MAXGROUPFILES];
|
|
static uint8_t groupfilgrp[MAXGROUPFILES];
|
|
static char *gfilelist[MAXGROUPFILES];
|
|
static char *groupname[MAXGROUPFILES];
|
|
static int32_t *gfileoffs[MAXGROUPFILES];
|
|
|
|
static uint8_t filegrp[MAXOPENFILES];
|
|
static int32_t filepos[MAXOPENFILES];
|
|
static intptr_t filehan[MAXOPENFILES] =
|
|
{
|
|
-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
|
|
-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
|
|
-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
|
|
-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1
|
|
};
|
|
|
|
#ifdef WITHKPLIB
|
|
static char filenamsav[MAXOPENFILES][260];
|
|
static int32_t kzcurhand = -1;
|
|
|
|
int32_t cache1d_file_fromzip(int32_t fil)
|
|
{
|
|
return (filegrp[fil] == GRP_ZIP);
|
|
}
|
|
#endif
|
|
|
|
static int32_t kopen_internal(const char *filename, char **lastpfn, char searchfirst, char checkcase, char tryzip, int32_t newhandle, uint8_t *arraygrp, intptr_t *arrayhan, int32_t *arraypos);
|
|
static int32_t kread_grp(int32_t handle, void *buffer, int32_t leng);
|
|
static int32_t klseek_grp(int32_t handle, int32_t offset, int32_t whence);
|
|
static void kclose_grp(int32_t handle);
|
|
|
|
int initgroupfile(const char *filename)
|
|
{
|
|
char buf[70];
|
|
|
|
// translate all backslashes (0x5c) to forward slashes (0x2f)
|
|
toupperlookup[0x5c] = 0x2f;
|
|
|
|
if (filename == NULL)
|
|
return -1;
|
|
|
|
// Technically you should be able to load more zips even if your GRPs are maxed out,
|
|
// but this system is already enough of a disaster.
|
|
if (numgroupfiles >= MAXGROUPFILES)
|
|
return -1;
|
|
|
|
char *zfn = NULL;
|
|
|
|
if (kopen_internal(filename, &zfn, 0, 0, 0, numgroupfiles, groupfilgrp, groupfil, groupfilpos) < 0)
|
|
return -1;
|
|
|
|
#ifdef WITHKPLIB
|
|
// check if ZIP
|
|
if (zfn)
|
|
{
|
|
kread_grp(numgroupfiles, buf, 4);
|
|
if (buf[0] == 0x50 && buf[1] == 0x4B && buf[2] == 0x03 && buf[3] == 0x04)
|
|
{
|
|
kclose_grp(numgroupfiles);
|
|
|
|
kzaddstack(zfn);
|
|
Bfree(zfn);
|
|
return MAXGROUPFILES;
|
|
}
|
|
klseek_grp(numgroupfiles,0,BSEEK_SET);
|
|
|
|
Bfree(zfn);
|
|
}
|
|
#else
|
|
Bfree(zfn);
|
|
#endif
|
|
|
|
// check if GRP
|
|
kread_grp(numgroupfiles,buf,16);
|
|
if (!Bmemcmp(buf, "KenSilverman", 12))
|
|
{
|
|
gnumfiles[numgroupfiles] = B_LITTLE32(*((int32_t *)&buf[12]));
|
|
|
|
gfilelist[numgroupfiles] = (char *)Xmalloc(gnumfiles[numgroupfiles]<<4);
|
|
gfileoffs[numgroupfiles] = (int32_t *)Xmalloc((gnumfiles[numgroupfiles]+1)<<2);
|
|
|
|
kread_grp(numgroupfiles,gfilelist[numgroupfiles],gnumfiles[numgroupfiles]<<4);
|
|
|
|
int32_t j = (gnumfiles[numgroupfiles]+1)<<4, k;
|
|
for (bssize_t i=0; i<gnumfiles[numgroupfiles]; i++)
|
|
{
|
|
k = B_LITTLE32(*((int32_t *)&gfilelist[numgroupfiles][(i<<4)+12]));
|
|
gfilelist[numgroupfiles][(i<<4)+12] = 0;
|
|
gfileoffs[numgroupfiles][i] = j;
|
|
j += k;
|
|
}
|
|
gfileoffs[numgroupfiles][gnumfiles[numgroupfiles]] = j;
|
|
groupname[numgroupfiles] = Xstrdup(filename);
|
|
return numgroupfiles++;
|
|
}
|
|
klseek_grp(numgroupfiles, 0, BSEEK_SET);
|
|
|
|
// check if SSI
|
|
// this performs several checks because there is no "SSI" magic
|
|
int32_t version;
|
|
kread_grp(numgroupfiles, &version, 4);
|
|
version = B_LITTLE32(version);
|
|
while (version == 1 || version == 2) // if
|
|
{
|
|
char zerobuf[70];
|
|
Bmemset(zerobuf, 0, 70);
|
|
|
|
int32_t numfiles;
|
|
kread_grp(numgroupfiles, &numfiles, 4);
|
|
numfiles = B_LITTLE32(numfiles);
|
|
|
|
uint8_t temp, temp2;
|
|
|
|
// get the string length
|
|
kread_grp(numgroupfiles, &temp, 1);
|
|
if (temp > 31) // 32 bytes allocated for the string
|
|
break;
|
|
// seek to the end of the string
|
|
klseek_grp(numgroupfiles, temp, BSEEK_CUR);
|
|
// verify everything remaining is a null terminator
|
|
temp = 32 - temp;
|
|
kread_grp(numgroupfiles, buf, temp);
|
|
if (Bmemcmp(buf, zerobuf, temp))
|
|
break;
|
|
|
|
if (version == 2)
|
|
{
|
|
// get the string length
|
|
kread_grp(numgroupfiles, &temp, 1);
|
|
if (temp > 11) // 12 bytes allocated for the string
|
|
break;
|
|
// seek to the end of the string
|
|
klseek_grp(numgroupfiles, temp, BSEEK_CUR);
|
|
// verify everything remaining is a null terminator
|
|
temp = 12 - temp;
|
|
kread_grp(numgroupfiles, buf, temp);
|
|
if (Bmemcmp(buf, zerobuf, temp))
|
|
break;
|
|
}
|
|
|
|
temp2 = 0;
|
|
for (uint8_t i=0;i<3;i++)
|
|
{
|
|
// get the string length
|
|
kread_grp(numgroupfiles, &temp, 1);
|
|
if (temp > 70) // 70 bytes allocated for the string
|
|
{
|
|
temp2 = 1;
|
|
break;
|
|
}
|
|
// seek to the end of the string
|
|
klseek_grp(numgroupfiles, temp, BSEEK_CUR);
|
|
// verify everything remaining is a null terminator
|
|
temp = 70 - temp;
|
|
if (temp == 0)
|
|
continue;
|
|
kread_grp(numgroupfiles, buf, temp);
|
|
temp2 |= Bmemcmp(buf, zerobuf, temp);
|
|
}
|
|
if (temp2)
|
|
break;
|
|
|
|
// Passed all the tests: read data.
|
|
|
|
gnumfiles[numgroupfiles] = numfiles;
|
|
|
|
gfilelist[numgroupfiles] = (char *)Xmalloc(gnumfiles[numgroupfiles]<<4);
|
|
gfileoffs[numgroupfiles] = (int32_t *)Xmalloc((gnumfiles[numgroupfiles]+1)<<2);
|
|
|
|
int32_t j = (version == 2 ? 267 : 254) + (numfiles * 121), k;
|
|
for (bssize_t i = 0; i < numfiles; i++)
|
|
{
|
|
// get the string length
|
|
kread_grp(numgroupfiles, &temp, 1);
|
|
if (temp > 12)
|
|
temp = 12;
|
|
// read the file name
|
|
kread_grp(numgroupfiles, &gfilelist[numgroupfiles][i<<4], temp);
|
|
gfilelist[numgroupfiles][(i<<4)+temp] = 0;
|
|
|
|
// skip to the end of the 12 bytes
|
|
klseek_grp(numgroupfiles, 12-temp, BSEEK_CUR);
|
|
|
|
// get the file size
|
|
kread_grp(numgroupfiles, &k, 4);
|
|
k = B_LITTLE32(k);
|
|
|
|
// record the offset of the file in the SSI
|
|
gfileoffs[numgroupfiles][i] = j;
|
|
j += k;
|
|
|
|
// skip unknown data
|
|
klseek_grp(numgroupfiles, 104, BSEEK_CUR);
|
|
}
|
|
gfileoffs[numgroupfiles][gnumfiles[numgroupfiles]] = j;
|
|
groupname[numgroupfiles] = Xstrdup(filename);
|
|
return numgroupfiles++;
|
|
}
|
|
|
|
kclose_grp(numgroupfiles);
|
|
return -1;
|
|
}
|
|
|
|
void uninitgroupfile(void)
|
|
{
|
|
int32_t i;
|
|
|
|
for (i=numgroupfiles-1; i>=0; i--)
|
|
if (groupfil[i] != -1)
|
|
{
|
|
DO_FREE_AND_NULL(gfilelist[i]);
|
|
DO_FREE_AND_NULL(gfileoffs[i]);
|
|
DO_FREE_AND_NULL(groupname[i]);
|
|
|
|
Bclose(groupfil[i]);
|
|
groupfil[i] = -1;
|
|
}
|
|
numgroupfiles = 0;
|
|
|
|
// JBF 20040111: "close" any files open in groups
|
|
for (i=0; i<MAXOPENFILES; i++)
|
|
{
|
|
if (filegrp[i] < GRP_RESERVED_ID_START) // JBF 20040130: not external or ZIPped
|
|
filehan[i] = -1;
|
|
}
|
|
}
|
|
|
|
#ifdef FILENAME_CASE_CHECK
|
|
// See
|
|
// http://stackoverflow.com/questions/74451/getting-actual-file-name-with-proper-casing-on-windows
|
|
// for relevant discussion.
|
|
|
|
static char fnbuf[BMAX_PATH];
|
|
int fnofs;
|
|
|
|
int32_t (*check_filename_casing_fn)(void) = NULL;
|
|
|
|
// -1: failure, 0: match, 1: mismatch
|
|
static int32_t check_filename_mismatch(const char * const filename, int ofs)
|
|
{
|
|
if (!GetShortPathNameA(filename, fnbuf, BMAX_PATH)) return -1;
|
|
if (!GetLongPathNameA(fnbuf, fnbuf, BMAX_PATH)) return -1;
|
|
|
|
fnofs = ofs;
|
|
|
|
int len = Bstrlen(fnbuf+ofs);
|
|
|
|
char const * const fn = filename+ofs;
|
|
|
|
if (!Bstrncmp(fnbuf+ofs, fn, len))
|
|
return 0;
|
|
|
|
char * const tfn = Bstrtolower(Xstrdup(fn));
|
|
|
|
if (!Bstrncmp(fnbuf+ofs, tfn, len))
|
|
{
|
|
Bfree(tfn);
|
|
return 0;
|
|
}
|
|
|
|
Bstrupr(tfn);
|
|
|
|
if (!Bstrncmp(fnbuf+ofs, tfn, len))
|
|
{
|
|
Bfree(tfn);
|
|
return 0;
|
|
}
|
|
|
|
Bfree(tfn);
|
|
|
|
return 1;
|
|
}
|
|
#endif
|
|
|
|
static int32_t kopen_internal(const char *filename, char **lastpfn, char searchfirst, char checkcase, char tryzip, int32_t newhandle, uint8_t *arraygrp, intptr_t *arrayhan, int32_t *arraypos)
|
|
{
|
|
int32_t fil;
|
|
if (searchfirst == 0 && (fil = openfrompath_internal(filename, lastpfn, BO_BINARY|BO_RDONLY, BS_IREAD)) >= 0)
|
|
{
|
|
#ifdef FILENAME_CASE_CHECK
|
|
if (checkcase && check_filename_casing_fn && check_filename_casing_fn())
|
|
{
|
|
int32_t status;
|
|
char *cp, *lastslash;
|
|
|
|
// convert all slashes to backslashes because SHGetFileInfo()
|
|
// complains else!
|
|
lastslash = *lastpfn;
|
|
for (cp=*lastpfn; *cp; cp++)
|
|
if (*cp=='/')
|
|
{
|
|
*cp = '\\';
|
|
lastslash = cp;
|
|
}
|
|
if (lastslash != *lastpfn)
|
|
lastslash++;
|
|
|
|
status = check_filename_mismatch(*lastpfn, lastslash-*lastpfn);
|
|
|
|
if (status == -1)
|
|
{
|
|
// initprintf("SHGetFileInfo failed with error code %lu\n", GetLastError());
|
|
}
|
|
else if (status == 1)
|
|
{
|
|
initprintf("warning: case mismatch: passed \"%s\", real \"%s\"\n",
|
|
lastslash, fnbuf+fnofs);
|
|
}
|
|
}
|
|
#else
|
|
UNREFERENCED_PARAMETER(checkcase);
|
|
#endif
|
|
arraygrp[newhandle] = GRP_FILESYSTEM;
|
|
arrayhan[newhandle] = fil;
|
|
arraypos[newhandle] = 0;
|
|
return newhandle;
|
|
}
|
|
|
|
for (; toupperlookup[*filename] == '/'; filename++) { }
|
|
|
|
#ifdef WITHKPLIB
|
|
if (tryzip)
|
|
{
|
|
intptr_t i;
|
|
if ((kzcurhand != newhandle) && (kztell() >= 0))
|
|
{
|
|
if (kzcurhand >= 0) arraypos[kzcurhand] = kztell();
|
|
kzclose();
|
|
kzcurhand = -1;
|
|
}
|
|
if (searchfirst != 1 && (i = kzipopen(filename)) != 0)
|
|
{
|
|
kzcurhand = newhandle;
|
|
arraygrp[newhandle] = GRP_ZIP;
|
|
arrayhan[newhandle] = i;
|
|
arraypos[newhandle] = 0;
|
|
strcpy(filenamsav[newhandle],filename);
|
|
return newhandle;
|
|
}
|
|
}
|
|
#else
|
|
UNREFERENCED_PARAMETER(tryzip);
|
|
#endif
|
|
|
|
for (bssize_t k = searchfirst != 1 ? numgroupfiles-1 : 0; k >= 0; --k)
|
|
{
|
|
if (groupfil[k] < 0)
|
|
continue;
|
|
|
|
for (bssize_t i = gnumfiles[k]-1; i >= 0; --i)
|
|
{
|
|
char const * const gfileptr = (char *)&gfilelist[k][i<<4];
|
|
|
|
unsigned int j;
|
|
for (j = 0; j < 13; ++j)
|
|
{
|
|
if (!filename[j]) break;
|
|
if (toupperlookup[filename[j]] != toupperlookup[gfileptr[j]])
|
|
goto gnumfiles_continue;
|
|
}
|
|
if (j<13 && gfileptr[j]) continue; // JBF: because e1l1.map might exist before e1l1
|
|
if (j==13 && filename[j]) continue; // JBF: long file name
|
|
|
|
arraygrp[newhandle] = k;
|
|
arrayhan[newhandle] = i;
|
|
arraypos[newhandle] = 0;
|
|
return newhandle;
|
|
|
|
gnumfiles_continue: ;
|
|
}
|
|
}
|
|
|
|
return -1;
|
|
}
|
|
|
|
void krename(int32_t crcval, int32_t filenum, const char *newname)
|
|
{
|
|
Bstrncpy((char *)&gfilelist[crcval][filenum<<4], newname, 12);
|
|
}
|
|
|
|
char const * kfileparent(int32_t const handle)
|
|
{
|
|
int32_t const groupnum = filegrp[handle];
|
|
|
|
if ((unsigned)groupnum >= MAXGROUPFILES || groupfil[groupnum] == -1)
|
|
return NULL;
|
|
|
|
return groupname[groupnum];
|
|
}
|
|
|
|
int32_t kopen4load(const char *filename, char searchfirst)
|
|
{
|
|
int32_t newhandle = MAXOPENFILES-1;
|
|
|
|
if (filename==NULL)
|
|
return -1;
|
|
|
|
while (filehan[newhandle] != -1)
|
|
{
|
|
newhandle--;
|
|
if (newhandle < 0)
|
|
{
|
|
Bprintf("TOO MANY FILES OPEN IN FILE GROUPING SYSTEM!");
|
|
Bexit(0);
|
|
}
|
|
}
|
|
|
|
char *lastpfn = NULL;
|
|
|
|
int32_t h = kopen_internal(filename, &lastpfn, searchfirst, 1, 1, newhandle, filegrp, filehan, filepos);
|
|
|
|
Bfree(lastpfn);
|
|
|
|
return h;
|
|
}
|
|
|
|
int32_t kread_internal(int32_t handle, void *buffer, int32_t leng, uint8_t *arraygrp, intptr_t *arrayhan, int32_t *arraypos)
|
|
{
|
|
int32_t filenum = arrayhan[handle];
|
|
int32_t groupnum = arraygrp[handle];
|
|
|
|
if (groupnum == GRP_FILESYSTEM) return Bread(filenum,buffer,leng);
|
|
#ifdef WITHKPLIB
|
|
else if (groupnum == GRP_ZIP)
|
|
{
|
|
if (kzcurhand != handle)
|
|
{
|
|
if (kztell() >= 0) { arraypos[kzcurhand] = kztell(); kzclose(); }
|
|
kzcurhand = handle;
|
|
kzipopen(filenamsav[handle]);
|
|
kzseek(arraypos[handle],SEEK_SET);
|
|
}
|
|
return kzread(buffer,leng);
|
|
}
|
|
#endif
|
|
|
|
if (EDUKE32_PREDICT_FALSE(groupfil[groupnum] == -1))
|
|
return 0;
|
|
|
|
int32_t rootgroupnum = groupnum;
|
|
int32_t i = 0;
|
|
while (groupfilgrp[rootgroupnum] != GRP_FILESYSTEM)
|
|
{
|
|
i += gfileoffs[groupfilgrp[rootgroupnum]][groupfil[rootgroupnum]];
|
|
rootgroupnum = groupfilgrp[rootgroupnum];
|
|
}
|
|
if (EDUKE32_PREDICT_TRUE(groupfil[rootgroupnum] != -1))
|
|
{
|
|
i += gfileoffs[groupnum][filenum]+arraypos[handle];
|
|
if (i != groupfilpos[rootgroupnum])
|
|
{
|
|
Blseek(groupfil[rootgroupnum],i,BSEEK_SET);
|
|
groupfilpos[rootgroupnum] = i;
|
|
}
|
|
leng = min(leng,(gfileoffs[groupnum][filenum+1]-gfileoffs[groupnum][filenum])-arraypos[handle]);
|
|
leng = Bread(groupfil[rootgroupnum],buffer,leng);
|
|
arraypos[handle] += leng;
|
|
groupfilpos[rootgroupnum] += leng;
|
|
return leng;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
int32_t klseek_internal(int32_t handle, int32_t offset, int32_t whence, uint8_t *arraygrp, intptr_t *arrayhan, int32_t *arraypos)
|
|
{
|
|
int32_t i, groupnum;
|
|
|
|
groupnum = arraygrp[handle];
|
|
|
|
if (groupnum == GRP_FILESYSTEM) return Blseek(arrayhan[handle],offset,whence);
|
|
#ifdef WITHKPLIB
|
|
else if (groupnum == GRP_ZIP)
|
|
{
|
|
if (kzcurhand != handle)
|
|
{
|
|
if (kztell() >= 0) { arraypos[kzcurhand] = kztell(); kzclose(); }
|
|
kzcurhand = handle;
|
|
kzipopen(filenamsav[handle]);
|
|
kzseek(arraypos[handle],SEEK_SET);
|
|
}
|
|
return kzseek(offset,whence);
|
|
}
|
|
#endif
|
|
|
|
if (groupfil[groupnum] != -1)
|
|
{
|
|
switch (whence)
|
|
{
|
|
case BSEEK_SET:
|
|
arraypos[handle] = offset; break;
|
|
case BSEEK_END:
|
|
i = arrayhan[handle];
|
|
arraypos[handle] = (gfileoffs[groupnum][i+1]-gfileoffs[groupnum][i])+offset;
|
|
break;
|
|
case BSEEK_CUR:
|
|
arraypos[handle] += offset; break;
|
|
}
|
|
return arraypos[handle];
|
|
}
|
|
return -1;
|
|
}
|
|
|
|
int32_t kfilelength_internal(int32_t handle, uint8_t *arraygrp, intptr_t *arrayhan, int32_t *arraypos)
|
|
{
|
|
int32_t i, groupnum;
|
|
|
|
groupnum = arraygrp[handle];
|
|
if (groupnum == GRP_FILESYSTEM)
|
|
{
|
|
// return (filelength(arrayhan[handle]))
|
|
return Bfilelength(arrayhan[handle]);
|
|
}
|
|
#ifdef WITHKPLIB
|
|
else if (groupnum == GRP_ZIP)
|
|
{
|
|
if (kzcurhand != handle)
|
|
{
|
|
if (kztell() >= 0) { arraypos[kzcurhand] = kztell(); kzclose(); }
|
|
kzcurhand = handle;
|
|
kzipopen(filenamsav[handle]);
|
|
kzseek(arraypos[handle],SEEK_SET);
|
|
}
|
|
return kzfilelength();
|
|
}
|
|
#endif
|
|
i = arrayhan[handle];
|
|
return gfileoffs[groupnum][i+1]-gfileoffs[groupnum][i];
|
|
}
|
|
|
|
int32_t ktell_internal(int32_t handle, uint8_t *arraygrp, intptr_t *arrayhan, int32_t *arraypos)
|
|
{
|
|
int32_t groupnum = arraygrp[handle];
|
|
|
|
if (groupnum == GRP_FILESYSTEM) return Blseek(arrayhan[handle],0,BSEEK_CUR);
|
|
#ifdef WITHKPLIB
|
|
else if (groupnum == GRP_ZIP)
|
|
{
|
|
if (kzcurhand != handle)
|
|
{
|
|
if (kztell() >= 0) { arraypos[kzcurhand] = kztell(); kzclose(); }
|
|
kzcurhand = handle;
|
|
kzipopen(filenamsav[handle]);
|
|
kzseek(arraypos[handle],SEEK_SET);
|
|
}
|
|
return kztell();
|
|
}
|
|
#endif
|
|
if (groupfil[groupnum] != -1)
|
|
return arraypos[handle];
|
|
return -1;
|
|
}
|
|
|
|
void kclose_internal(int32_t handle, uint8_t *arraygrp, intptr_t *arrayhan)
|
|
{
|
|
if (handle < 0) return;
|
|
if (arraygrp[handle] == GRP_FILESYSTEM) Bclose(arrayhan[handle]);
|
|
#ifdef WITHKPLIB
|
|
else if (arraygrp[handle] == GRP_ZIP)
|
|
{
|
|
kzclose();
|
|
kzcurhand = -1;
|
|
}
|
|
#endif
|
|
arrayhan[handle] = -1;
|
|
}
|
|
|
|
int32_t kread(int32_t handle, void *buffer, int32_t leng)
|
|
{
|
|
return kread_internal(handle, buffer, leng, filegrp, filehan, filepos);
|
|
}
|
|
int32_t klseek(int32_t handle, int32_t offset, int32_t whence)
|
|
{
|
|
return klseek_internal(handle, offset, whence, filegrp, filehan, filepos);
|
|
}
|
|
int32_t kfilelength(int32_t handle)
|
|
{
|
|
return kfilelength_internal(handle, filegrp, filehan, filepos);
|
|
}
|
|
int32_t ktell(int32_t handle)
|
|
{
|
|
return ktell_internal(handle, filegrp, filehan, filepos);
|
|
}
|
|
void kclose(int32_t handle)
|
|
{
|
|
return kclose_internal(handle, filegrp, filehan);
|
|
}
|
|
|
|
static int32_t kread_grp(int32_t handle, void *buffer, int32_t leng)
|
|
{
|
|
return kread_internal(handle, buffer, leng, groupfilgrp, groupfil, groupfilpos);
|
|
}
|
|
static int32_t klseek_grp(int32_t handle, int32_t offset, int32_t whence)
|
|
{
|
|
return klseek_internal(handle, offset, whence, groupfilgrp, groupfil, groupfilpos);
|
|
}
|
|
static void kclose_grp(int32_t handle)
|
|
{
|
|
return kclose_internal(handle, groupfilgrp, groupfil);
|
|
}
|
|
|
|
static int32_t klistaddentry(CACHE1D_FIND_REC **rec, const char *name, int32_t type, int32_t source)
|
|
{
|
|
CACHE1D_FIND_REC *r = NULL, *attach = NULL;
|
|
|
|
if (*rec)
|
|
{
|
|
int32_t insensitive, v;
|
|
CACHE1D_FIND_REC *last = NULL;
|
|
|
|
for (attach = *rec; attach; last = attach, attach = attach->next)
|
|
{
|
|
if (type == CACHE1D_FIND_DRIVE) continue; // we just want to get to the end for drives
|
|
#ifdef _WIN32
|
|
insensitive = 1;
|
|
#else
|
|
if (source == CACHE1D_SOURCE_GRP || attach->source == CACHE1D_SOURCE_GRP)
|
|
insensitive = 1;
|
|
else if (source == CACHE1D_SOURCE_ZIP || attach->source == CACHE1D_SOURCE_ZIP)
|
|
insensitive = 1;
|
|
else
|
|
{
|
|
extern int16_t editstatus; // XXX
|
|
insensitive = !editstatus;
|
|
}
|
|
// ^ in the game, don't show file list case-sensitive
|
|
#endif
|
|
if (insensitive) v = Bstrcasecmp(name, attach->name);
|
|
else v = Bstrcmp(name, attach->name);
|
|
|
|
// sorted list
|
|
if (v > 0) continue; // item to add is bigger than the current one
|
|
// so look for something bigger than us
|
|
if (v < 0) // item to add is smaller than the current one
|
|
{
|
|
attach = NULL; // so wedge it between the current item and the one before
|
|
break;
|
|
}
|
|
|
|
// matched
|
|
if (source >= attach->source) return 1; // item to add is of lower priority
|
|
r = attach;
|
|
break;
|
|
}
|
|
|
|
// wasn't found in the list, so attach to the end
|
|
if (!attach) attach = last;
|
|
}
|
|
|
|
if (r)
|
|
{
|
|
r->type = type;
|
|
r->source = source;
|
|
return 0;
|
|
}
|
|
|
|
r = (CACHE1D_FIND_REC *)Xmalloc(sizeof(CACHE1D_FIND_REC)+strlen(name)+1);
|
|
|
|
r->name = (char *)r + sizeof(CACHE1D_FIND_REC); strcpy(r->name, name);
|
|
r->type = type;
|
|
r->source = source;
|
|
r->usera = r->userb = NULL;
|
|
|
|
if (!attach) // we are the first item
|
|
{
|
|
r->prev = NULL;
|
|
r->next = *rec;
|
|
if (*rec)(*rec)->prev = r;
|
|
*rec = r;
|
|
}
|
|
else
|
|
{
|
|
r->prev = attach;
|
|
r->next = attach->next;
|
|
if (attach->next) attach->next->prev = r;
|
|
attach->next = r;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
void klistfree(CACHE1D_FIND_REC *rec)
|
|
{
|
|
CACHE1D_FIND_REC *n;
|
|
|
|
while (rec)
|
|
{
|
|
n = rec->next;
|
|
Bfree(rec);
|
|
rec = n;
|
|
}
|
|
}
|
|
|
|
CACHE1D_FIND_REC *klistpath(const char *_path, const char *mask, int32_t type)
|
|
{
|
|
CACHE1D_FIND_REC *rec = NULL;
|
|
char *path;
|
|
|
|
// pathsearchmode == 0: enumerates a path in the virtual filesystem
|
|
// pathsearchmode == 1: enumerates the system filesystem path passed in
|
|
|
|
path = Xstrdup(_path);
|
|
|
|
// we don't need any leading dots and slashes or trailing slashes either
|
|
{
|
|
int32_t i,j;
|
|
for (i=0; path[i] == '.' || toupperlookup[path[i]] == '/';) i++;
|
|
for (j=0; (path[j] = path[i]); j++,i++) ;
|
|
while (j>0 && toupperlookup[path[j-1]] == '/') j--;
|
|
path[j] = 0;
|
|
//initprintf("Cleaned up path = \"%s\"\n",path);
|
|
}
|
|
|
|
if (*path && (type & CACHE1D_FIND_DIR))
|
|
{
|
|
if (klistaddentry(&rec, "..", CACHE1D_FIND_DIR, CACHE1D_SOURCE_CURDIR) < 0) goto failure;
|
|
}
|
|
|
|
if (!(type & CACHE1D_OPT_NOSTACK)) // current directory and paths in the search stack
|
|
{
|
|
searchpath_t *search = NULL;
|
|
BDIR *dir;
|
|
struct Bdirent *dirent;
|
|
|
|
static const char *const CUR_DIR = "./";
|
|
// Adjusted for the following "autoload" dir fix - NY00123
|
|
const char *d = pathsearchmode ? _path : CUR_DIR;
|
|
int32_t stackdepth = CACHE1D_SOURCE_CURDIR;
|
|
char buf[BMAX_PATH];
|
|
|
|
do
|
|
{
|
|
if (d==CUR_DIR && (type & CACHE1D_FIND_NOCURDIR))
|
|
goto next;
|
|
|
|
strcpy(buf, d);
|
|
if (!pathsearchmode)
|
|
{
|
|
// Fix for "autoload" dir in multi-user environments - NY00123
|
|
strcat(buf, path);
|
|
if (*path) strcat(buf, "/");
|
|
}
|
|
|
|
dir = Bopendir(buf);
|
|
if (dir)
|
|
{
|
|
while ((dirent = Breaddir(dir)))
|
|
{
|
|
if ((dirent->name[0] == '.' && dirent->name[1] == 0) ||
|
|
(dirent->name[0] == '.' && dirent->name[1] == '.' && dirent->name[2] == 0))
|
|
continue;
|
|
if ((type & CACHE1D_FIND_DIR) && !(dirent->mode & BS_IFDIR)) continue;
|
|
if ((type & CACHE1D_FIND_FILE) && (dirent->mode & BS_IFDIR)) continue;
|
|
if (!Bwildmatch(dirent->name, mask)) continue;
|
|
switch (klistaddentry(&rec, dirent->name,
|
|
(dirent->mode & BS_IFDIR) ? CACHE1D_FIND_DIR : CACHE1D_FIND_FILE,
|
|
stackdepth))
|
|
{
|
|
case -1: goto failure;
|
|
//case 1: initprintf("%s:%s dropped for lower priority\n", d,dirent->name); break;
|
|
//case 0: initprintf("%s:%s accepted\n", d,dirent->name); break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
Bclosedir(dir);
|
|
}
|
|
next:
|
|
if (pathsearchmode)
|
|
break;
|
|
|
|
if (!search)
|
|
{
|
|
search = searchpathhead;
|
|
stackdepth = CACHE1D_SOURCE_PATH;
|
|
}
|
|
else
|
|
{
|
|
search = search->next;
|
|
stackdepth++;
|
|
}
|
|
|
|
if (search)
|
|
d = search->path;
|
|
}
|
|
while (search);
|
|
}
|
|
|
|
#ifdef WITHKPLIB
|
|
if (!(type & CACHE1D_FIND_NOCURDIR)) // TEMP, until we have sorted out fs.listpath() API
|
|
if (!pathsearchmode) // next, zip files
|
|
{
|
|
char buf[BMAX_PATH+4];
|
|
int32_t i, j, ftype;
|
|
strcpy(buf,path);
|
|
if (*path) strcat(buf,"/");
|
|
strcat(buf,mask);
|
|
for (kzfindfilestart(buf); kzfindfile(buf);)
|
|
{
|
|
if (buf[0] != '|') continue; // local files we don't need
|
|
|
|
// scan for the end of the string and shift
|
|
// everything left a char in the process
|
|
for (i=1; (buf[i-1]=buf[i]); i++)
|
|
{
|
|
/* do nothing */
|
|
}
|
|
i-=2;
|
|
if (i < 0)
|
|
i = 0;
|
|
|
|
// if there's a slash at the end, this is a directory entry
|
|
if (toupperlookup[buf[i]] == '/') { ftype = CACHE1D_FIND_DIR; buf[i] = 0; }
|
|
else ftype = CACHE1D_FIND_FILE;
|
|
|
|
// skip over the common characters at the beginning of the base path and the zip entry
|
|
for (j=0; buf[j] && path[j]; j++)
|
|
{
|
|
if (toupperlookup[ path[j] ] == toupperlookup[ buf[j] ]) continue;
|
|
break;
|
|
}
|
|
// we've now hopefully skipped the common path component at the beginning.
|
|
// if that's true, we should be staring at a null byte in path and either any character in buf
|
|
// if j==0, or a slash if j>0
|
|
if ((!path[0] && buf[j]) || (!path[j] && toupperlookup[ buf[j] ] == '/'))
|
|
{
|
|
if (j>0) j++;
|
|
|
|
// yep, so now we shift what follows back to the start of buf and while we do that,
|
|
// keep an eye out for any more slashes which would mean this entry has sub-entities
|
|
// and is useless to us.
|
|
for (i = 0; (buf[i] = buf[j]) && toupperlookup[buf[j]] != '/'; i++,j++) ;
|
|
if (toupperlookup[buf[j]] == '/') continue; // damn, try next entry
|
|
}
|
|
else
|
|
{
|
|
// if we're here it means we have a situation where:
|
|
// path = foo
|
|
// buf = foobar...
|
|
// or
|
|
// path = foobar
|
|
// buf = foo...
|
|
// which would mean the entry is higher up in the directory tree and is also useless
|
|
continue;
|
|
}
|
|
|
|
if ((type & CACHE1D_FIND_DIR) && ftype != CACHE1D_FIND_DIR) continue;
|
|
if ((type & CACHE1D_FIND_FILE) && ftype != CACHE1D_FIND_FILE) continue;
|
|
|
|
// the entry is in the clear
|
|
switch (klistaddentry(&rec, buf, ftype, CACHE1D_SOURCE_ZIP))
|
|
{
|
|
case -1:
|
|
goto failure;
|
|
//case 1: initprintf("<ZIP>:%s dropped for lower priority\n", buf); break;
|
|
//case 0: initprintf("<ZIP>:%s accepted\n", buf); break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
#endif
|
|
// then, grp files
|
|
if (!(type & CACHE1D_FIND_NOCURDIR)) // TEMP, until we have sorted out fs.listpath() API
|
|
if (!pathsearchmode && !*path && (type & CACHE1D_FIND_FILE))
|
|
{
|
|
char buf[13];
|
|
int32_t i,j;
|
|
buf[12] = 0;
|
|
for (i=0; i<MAXGROUPFILES; i++)
|
|
{
|
|
if (groupfil[i] == -1) continue;
|
|
for (j=gnumfiles[i]-1; j>=0; j--)
|
|
{
|
|
Bmemcpy(buf,&gfilelist[i][j<<4],12);
|
|
if (!Bwildmatch(buf,mask)) continue;
|
|
switch (klistaddentry(&rec, buf, CACHE1D_FIND_FILE, CACHE1D_SOURCE_GRP))
|
|
{
|
|
case -1:
|
|
goto failure;
|
|
//case 1: initprintf("<GRP>:%s dropped for lower priority\n", workspace); break;
|
|
//case 0: initprintf("<GRP>:%s accepted\n", workspace); break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if (pathsearchmode && (type & CACHE1D_FIND_DRIVE))
|
|
{
|
|
char *drives, *drp;
|
|
drives = Bgetsystemdrives();
|
|
if (drives)
|
|
{
|
|
for (drp=drives; *drp; drp+=strlen(drp)+1)
|
|
{
|
|
if (klistaddentry(&rec, drp, CACHE1D_FIND_DRIVE, CACHE1D_SOURCE_DRIVE) < 0)
|
|
{
|
|
Bfree(drives);
|
|
goto failure;
|
|
}
|
|
}
|
|
Bfree(drives);
|
|
}
|
|
}
|
|
|
|
Bfree(path);
|
|
// XXX: may be NULL if no file was listed, and thus indistinguishable from
|
|
// an error condition.
|
|
return rec;
|
|
failure:
|
|
Bfree(path);
|
|
klistfree(rec);
|
|
return NULL;
|
|
}
|
|
|
|
|
|
#endif // #ifdef CACHE1D_COMPRESS_ONLY / else
|
|
|
|
|
|
//Internal LZW variables
|
|
#define LZWSIZE 16384 //Watch out for shorts!
|
|
#define LZWSIZEPAD (LZWSIZE+(LZWSIZE>>4))
|
|
|
|
// lzwrawbuf LZWSIZE+1 (formerly): see (*) below
|
|
// XXX: lzwrawbuf size increased again :-/
|
|
static char lzwtmpbuf[LZWSIZEPAD], lzwrawbuf[LZWSIZEPAD], lzwcompbuf[LZWSIZEPAD];
|
|
static int16_t lzwbuf2[LZWSIZEPAD], lzwbuf3[LZWSIZEPAD];
|
|
|
|
static int32_t lzwcompress(const char *lzwinbuf, int32_t uncompleng, char *lzwoutbuf);
|
|
static int32_t lzwuncompress(const char *lzwinbuf, int32_t compleng, char *lzwoutbuf);
|
|
|
|
#ifndef CACHE1D_COMPRESS_ONLY
|
|
static int32_t kdfread_func(intptr_t fil, void *outbuf, int32_t length)
|
|
{
|
|
return kread((int32_t)fil, outbuf, length);
|
|
}
|
|
|
|
static void dfwrite_func(intptr_t fp, const void *inbuf, int32_t length)
|
|
{
|
|
Bfwrite(inbuf, length, 1, (BFILE *)fp);
|
|
}
|
|
#else
|
|
# define kdfread_func NULL
|
|
# define dfwrite_func NULL
|
|
#endif
|
|
|
|
// These two follow the argument order of the C functions "read" and "write":
|
|
// handle, buffer, length.
|
|
C1D_STATIC int32_t (*c1d_readfunc)(intptr_t, void *, int32_t) = kdfread_func;
|
|
C1D_STATIC void (*c1d_writefunc)(intptr_t, const void *, int32_t) = dfwrite_func;
|
|
|
|
|
|
////////// COMPRESSED READ //////////
|
|
|
|
static uint32_t decompress_part(intptr_t f, uint32_t *kgoalptr)
|
|
{
|
|
int16_t leng;
|
|
|
|
// Read compressed length first.
|
|
if (c1d_readfunc(f, &leng, 2) != 2)
|
|
return 1;
|
|
leng = B_LITTLE16(leng);
|
|
|
|
if (c1d_readfunc(f,lzwcompbuf, leng) != leng)
|
|
return 1;
|
|
|
|
*kgoalptr = lzwuncompress(lzwcompbuf, leng, lzwrawbuf);
|
|
return 0;
|
|
}
|
|
|
|
// Read from 'f' into 'buffer'.
|
|
C1D_STATIC int32_t c1d_read_compressed(void *buffer, bsize_t dasizeof, bsize_t count, intptr_t f)
|
|
{
|
|
char *ptr = (char *)buffer;
|
|
|
|
if (dasizeof > LZWSIZE)
|
|
{
|
|
count *= dasizeof;
|
|
dasizeof = 1;
|
|
}
|
|
|
|
uint32_t kgoal;
|
|
|
|
if (decompress_part(f, &kgoal))
|
|
return -1;
|
|
|
|
Bmemcpy(ptr, lzwrawbuf, (int32_t)dasizeof);
|
|
|
|
uint32_t k = (int32_t)dasizeof;
|
|
|
|
for (uint32_t i=1; i<count; i++)
|
|
{
|
|
if (k >= kgoal)
|
|
{
|
|
k = decompress_part(f, &kgoal);
|
|
if (k) return -1;
|
|
}
|
|
|
|
uint32_t j = 0;
|
|
|
|
if (dasizeof >= 4)
|
|
{
|
|
for (; j<dasizeof-4; j+=4)
|
|
{
|
|
ptr[j+dasizeof] = ((ptr[j]+lzwrawbuf[j+k])&255);
|
|
ptr[j+1+dasizeof] = ((ptr[j+1]+lzwrawbuf[j+1+k])&255);
|
|
ptr[j+2+dasizeof] = ((ptr[j+2]+lzwrawbuf[j+2+k])&255);
|
|
ptr[j+3+dasizeof] = ((ptr[j+3]+lzwrawbuf[j+3+k])&255);
|
|
}
|
|
}
|
|
|
|
for (; j<dasizeof; j++)
|
|
ptr[j+dasizeof] = ((ptr[j]+lzwrawbuf[j+k])&255);
|
|
|
|
k += dasizeof;
|
|
ptr += dasizeof;
|
|
}
|
|
|
|
return count;
|
|
}
|
|
|
|
int32_t kdfread(void *buffer, bsize_t dasizeof, bsize_t count, int32_t fil)
|
|
{
|
|
return c1d_read_compressed(buffer, dasizeof, count, (intptr_t)fil);
|
|
}
|
|
|
|
|
|
////////// COMPRESSED WRITE //////////
|
|
|
|
static uint32_t compress_part(uint32_t k, intptr_t f)
|
|
{
|
|
const int16_t leng = (int16_t)lzwcompress(lzwrawbuf, k, lzwcompbuf);
|
|
const int16_t swleng = B_LITTLE16(leng);
|
|
|
|
c1d_writefunc(f, &swleng, 2);
|
|
c1d_writefunc(f, lzwcompbuf, leng);
|
|
|
|
return 0;
|
|
}
|
|
|
|
// Write from 'buffer' to 'f'.
|
|
C1D_STATIC void c1d_write_compressed(const void *buffer, bsize_t dasizeof, bsize_t count, intptr_t f)
|
|
{
|
|
char const *ptr = (char const *)buffer;
|
|
|
|
if (dasizeof > LZWSIZE)
|
|
{
|
|
count *= dasizeof;
|
|
dasizeof = 1;
|
|
}
|
|
|
|
Bmemcpy(lzwrawbuf, ptr, (int32_t)dasizeof);
|
|
|
|
uint32_t k = dasizeof;
|
|
if (k > LZWSIZE-dasizeof)
|
|
k = compress_part(k, f);
|
|
|
|
for (uint32_t i=1; i<count; i++)
|
|
{
|
|
uint32_t j = 0;
|
|
|
|
if (dasizeof >= 4)
|
|
{
|
|
for (; j<dasizeof-4; j+=4)
|
|
{
|
|
lzwrawbuf[j+k] = ((ptr[j+dasizeof]-ptr[j])&255);
|
|
lzwrawbuf[j+1+k] = ((ptr[j+1+dasizeof]-ptr[j+1])&255);
|
|
lzwrawbuf[j+2+k] = ((ptr[j+2+dasizeof]-ptr[j+2])&255);
|
|
lzwrawbuf[j+3+k] = ((ptr[j+3+dasizeof]-ptr[j+3])&255);
|
|
}
|
|
}
|
|
|
|
for (; j<dasizeof; j++)
|
|
lzwrawbuf[j+k] = ((ptr[j+dasizeof]-ptr[j])&255);
|
|
|
|
k += dasizeof;
|
|
if (k > LZWSIZE-dasizeof)
|
|
k = compress_part(k, f);
|
|
|
|
ptr += dasizeof;
|
|
}
|
|
|
|
if (k > 0)
|
|
compress_part(k, f);
|
|
}
|
|
|
|
void dfwrite(const void *buffer, bsize_t dasizeof, bsize_t count, BFILE *fil)
|
|
{
|
|
c1d_write_compressed(buffer, dasizeof, count, (intptr_t)fil);
|
|
}
|
|
|
|
////////// CORE COMPRESSION FUNCTIONS //////////
|
|
|
|
static int32_t lzwcompress(const char *lzwinbuf, int32_t uncompleng, char *lzwoutbuf)
|
|
{
|
|
int32_t i, addr, addrcnt, *intptr;
|
|
int32_t bytecnt1, bitcnt, numbits, oneupnumbits;
|
|
int16_t *shortptr;
|
|
|
|
int16_t *const lzwcodehead = lzwbuf2;
|
|
int16_t *const lzwcodenext = lzwbuf3;
|
|
|
|
for (i=255; i>=4; i-=4)
|
|
{
|
|
lzwtmpbuf[i] = i, lzwcodenext[i] = (i+1)&255;
|
|
lzwtmpbuf[i-1] = i-1, lzwcodenext[i-1] = (i) &255;
|
|
lzwtmpbuf[i-2] = i-2, lzwcodenext[i-2] = (i-1)&255;
|
|
lzwtmpbuf[i-3] = i-3, lzwcodenext[i-3] = (i-2)&255;
|
|
lzwcodehead[i] = lzwcodehead[i-1] = lzwcodehead[i-2] = lzwcodehead[i-3] = -1;
|
|
}
|
|
|
|
for (; i>=0; i--)
|
|
{
|
|
lzwtmpbuf[i] = i;
|
|
lzwcodenext[i] = (i+1)&255;
|
|
lzwcodehead[i] = -1;
|
|
}
|
|
|
|
Bmemset(lzwoutbuf, 0, 4+uncompleng+1);
|
|
// clearbuf(lzwoutbuf,((uncompleng+15)+3)>>2,0L);
|
|
|
|
addrcnt = 256; bytecnt1 = 0; bitcnt = (4<<3);
|
|
numbits = 8; oneupnumbits = (1<<8);
|
|
do
|
|
{
|
|
addr = lzwinbuf[bytecnt1];
|
|
do
|
|
{
|
|
int32_t newaddr;
|
|
|
|
if (++bytecnt1 == uncompleng)
|
|
break; // (*) see XXX below
|
|
|
|
if (lzwcodehead[addr] < 0)
|
|
{
|
|
lzwcodehead[addr] = addrcnt;
|
|
break;
|
|
}
|
|
|
|
newaddr = lzwcodehead[addr];
|
|
while (lzwtmpbuf[newaddr] != lzwinbuf[bytecnt1])
|
|
{
|
|
if (lzwcodenext[newaddr] < 0)
|
|
{
|
|
lzwcodenext[newaddr] = addrcnt;
|
|
break;
|
|
}
|
|
newaddr = lzwcodenext[newaddr];
|
|
}
|
|
|
|
if (lzwcodenext[newaddr] == addrcnt)
|
|
break;
|
|
addr = newaddr;
|
|
}
|
|
while (addr >= 0);
|
|
|
|
lzwtmpbuf[addrcnt] = lzwinbuf[bytecnt1]; // XXX: potential oob access of lzwinbuf via (*) above
|
|
lzwcodehead[addrcnt] = -1;
|
|
lzwcodenext[addrcnt] = -1;
|
|
|
|
intptr = (int32_t *)&lzwoutbuf[bitcnt>>3];
|
|
intptr[0] |= B_LITTLE32(addr<<(bitcnt&7));
|
|
bitcnt += numbits;
|
|
if ((addr&((oneupnumbits>>1)-1)) > ((addrcnt-1)&((oneupnumbits>>1)-1)))
|
|
bitcnt--;
|
|
|
|
addrcnt++;
|
|
if (addrcnt > oneupnumbits)
|
|
{ numbits++; oneupnumbits <<= 1; }
|
|
}
|
|
while ((bytecnt1 < uncompleng) && (bitcnt < (uncompleng<<3)));
|
|
|
|
intptr = (int32_t *)&lzwoutbuf[bitcnt>>3];
|
|
intptr[0] |= B_LITTLE32(addr<<(bitcnt&7));
|
|
bitcnt += numbits;
|
|
if ((addr&((oneupnumbits>>1)-1)) > ((addrcnt-1)&((oneupnumbits>>1)-1)))
|
|
bitcnt--;
|
|
|
|
shortptr = (int16_t *)lzwoutbuf;
|
|
shortptr[0] = B_LITTLE16((int16_t)uncompleng);
|
|
|
|
if (((bitcnt+7)>>3) < uncompleng)
|
|
{
|
|
shortptr[1] = B_LITTLE16((int16_t)addrcnt);
|
|
return (bitcnt+7)>>3;
|
|
}
|
|
|
|
// Failed compressing, mark this in the stream.
|
|
shortptr[1] = 0;
|
|
|
|
for (i=0; i<uncompleng-4; i+=4)
|
|
{
|
|
lzwoutbuf[i+4] = lzwinbuf[i];
|
|
lzwoutbuf[i+5] = lzwinbuf[i+1];
|
|
lzwoutbuf[i+6] = lzwinbuf[i+2];
|
|
lzwoutbuf[i+7] = lzwinbuf[i+3];
|
|
}
|
|
|
|
for (; i<uncompleng; i++)
|
|
lzwoutbuf[i+4] = lzwinbuf[i];
|
|
|
|
return uncompleng+4;
|
|
}
|
|
|
|
static int32_t lzwuncompress(const char *lzwinbuf, int32_t compleng, char *lzwoutbuf)
|
|
{
|
|
int32_t currstr, numbits, oneupnumbits;
|
|
int32_t i, bitcnt, outbytecnt;
|
|
|
|
const int16_t *const shortptr = (const int16_t *)lzwinbuf;
|
|
const int32_t strtot = B_LITTLE16(shortptr[1]);
|
|
const int32_t uncompleng = B_LITTLE16(shortptr[0]);
|
|
|
|
if (strtot == 0)
|
|
{
|
|
if (lzwoutbuf==lzwrawbuf && lzwinbuf==lzwcompbuf)
|
|
{
|
|
Bassert((compleng-4)+3+0u < sizeof(lzwrawbuf));
|
|
Bassert((compleng-4)+3+0u < sizeof(lzwcompbuf)-4);
|
|
}
|
|
|
|
Bmemcpy(lzwoutbuf, lzwinbuf+4, (compleng-4)+3);
|
|
return uncompleng;
|
|
}
|
|
|
|
for (i=255; i>=4; i-=4)
|
|
{
|
|
lzwbuf2[i] = lzwbuf3[i] = i;
|
|
lzwbuf2[i-1] = lzwbuf3[i-1] = i-1;
|
|
lzwbuf2[i-2] = lzwbuf3[i-2] = i-2;
|
|
lzwbuf2[i-3] = lzwbuf3[i-3] = i-3;
|
|
}
|
|
|
|
lzwbuf2[i] = lzwbuf3[i] = i;
|
|
lzwbuf2[i-1] = lzwbuf3[i-1] = i-1;
|
|
lzwbuf2[i-2] = lzwbuf3[i-2] = i-2;
|
|
|
|
currstr = 256; bitcnt = (4<<3); outbytecnt = 0;
|
|
numbits = 8; oneupnumbits = (1<<8);
|
|
do
|
|
{
|
|
const int32_t *const intptr = (const int32_t *)&lzwinbuf[bitcnt>>3];
|
|
|
|
int32_t dat = ((B_LITTLE32(intptr[0])>>(bitcnt&7)) & (oneupnumbits-1));
|
|
int32_t leng;
|
|
|
|
bitcnt += numbits;
|
|
if ((dat&((oneupnumbits>>1)-1)) > ((currstr-1)&((oneupnumbits>>1)-1)))
|
|
{ dat &= ((oneupnumbits>>1)-1); bitcnt--; }
|
|
|
|
lzwbuf3[currstr] = dat;
|
|
|
|
for (leng=0; dat>=256; leng++,dat=lzwbuf3[dat])
|
|
lzwtmpbuf[leng] = lzwbuf2[dat];
|
|
|
|
lzwoutbuf[outbytecnt++] = dat;
|
|
|
|
for (i=leng-1; i>=4; i-=4, outbytecnt+=4)
|
|
{
|
|
lzwoutbuf[outbytecnt] = lzwtmpbuf[i];
|
|
lzwoutbuf[outbytecnt+1] = lzwtmpbuf[i-1];
|
|
lzwoutbuf[outbytecnt+2] = lzwtmpbuf[i-2];
|
|
lzwoutbuf[outbytecnt+3] = lzwtmpbuf[i-3];
|
|
}
|
|
|
|
for (; i>=0; i--)
|
|
lzwoutbuf[outbytecnt++] = lzwtmpbuf[i];
|
|
|
|
lzwbuf2[currstr-1] = dat; lzwbuf2[currstr] = dat;
|
|
currstr++;
|
|
if (currstr > oneupnumbits)
|
|
{ numbits++; oneupnumbits <<= 1; }
|
|
}
|
|
while (currstr < strtot);
|
|
|
|
return uncompleng;
|
|
}
|
|
|
|
/*
|
|
* vim:ts=4:sw=4:
|
|
*/
|