static fz_error
pdf_readoldtrailer(pdf_xref *xref, char *buf, int cap)
{
fz_error error;
int len;
char *s;
int n;
int t;
pdf_token_e tok;
int c;
pdf_logxref("load old xref format trailer\n");
fz_readline(xref->file, buf, cap);
if (strncmp(buf, "xref", 4) != 0)
return fz_throw("cannot find xref marker");
while (1)
{
c = fz_peekbyte(xref->file);
if (!(c >= '0' && c <= '9'))
break;
fz_readline(xref->file, buf, cap);
s = buf;
fz_strsep(&s, " "); /* ignore ofs */
if (!s)
return fz_throw("invalid range marker in xref");
len = atoi(fz_strsep(&s, " "));
/* broken pdfs where the section is not on a separate line */
if (s && *s != '\0')
fz_seek(xref->file, -(2 + (int)strlen(s)), 1);
t = fz_tell(xref->file);
if (t < 0)
return fz_throw("cannot tell in file");
fz_seek(xref->file, t + 20 * len, 0);
}
error = pdf_lex(&tok, xref->file, buf, cap, &n);
if (error)
return fz_rethrow(error, "cannot parse trailer");
if (tok != PDF_TTRAILER)
return fz_throw("expected trailer marker");
error = pdf_lex(&tok, xref->file, buf, cap, &n);
if (error)
return fz_rethrow(error, "cannot parse trailer");
if (tok != PDF_TODICT)
return fz_throw("expected trailer dictionary");
error = pdf_parsedict(&xref->trailer, xref, xref->file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse trailer");
return fz_okay;
}
static fz_error
pdf_repairobjstm(pdf_xref *xref, int num, int gen)
{
fz_error error;
fz_obj *obj;
fz_stream *stm;
int tok;
int i, n, count;
char buf[256];
error = pdf_loadobject(&obj, xref, num, gen);
if (error)
return fz_rethrow(error, "cannot load object stream object (%d %d R)", num, gen);
count = fz_toint(fz_dictgets(obj, "N"));
fz_dropobj(obj);
error = pdf_openstream(&stm, xref, num, gen);
if (error)
return fz_rethrow(error, "cannot open object stream object (%d %d R)", num, gen);
for (i = 0; i < count; i++)
{
error = pdf_lex(&tok, stm, buf, sizeof buf, &n);
if (error || tok != PDF_TINT)
{
fz_close(stm);
return fz_rethrow(error, "corrupt object stream (%d %d R)", num, gen);
}
n = atoi(buf);
if (n >= xref->len)
pdf_resizexref(xref, n + 1);
xref->table[n].ofs = num;
xref->table[n].gen = i;
xref->table[n].stmofs = 0;
xref->table[n].obj = nil;
xref->table[n].type = 'o';
error = pdf_lex(&tok, stm, buf, sizeof buf, &n);
if (error || tok != PDF_TINT)
{
fz_close(stm);
return fz_rethrow(error, "corrupt object stream (%d %d R)", num, gen);
}
}
fz_close(stm);
return fz_okay;
}
static float
pdf_extract_font_size(pdf_xref *xref, char *appearance, char **font_name)
{
fz_context *ctx = xref->ctx;
fz_stream *stream = fz_open_memory(ctx, appearance, strlen(appearance));
float font_size = 0;
int tok, len;
*font_name = NULL;
do
{
fz_error error = pdf_lex(&tok, stream, xref->scratch, sizeof(xref->scratch), &len);
if (error || tok == PDF_TOK_EOF)
{
fz_free(ctx, *font_name);
*font_name = NULL;
break;
}
if (tok == PDF_TOK_NAME)
{
fz_free(ctx, *font_name);
*font_name = fz_strdup(ctx, xref->scratch);
}
else if (tok == PDF_TOK_REAL || tok == PDF_TOK_INT)
{
font_size = fz_atof(xref->scratch);
}
} while (tok != PDF_TOK_KEYWORD || strcmp(xref->scratch, "Tf") != 0);
fz_close(stream);
return font_size;
}
fz_error
pdf_parse_stm_obj(fz_obj **op, pdf_xref *xref, fz_stream *file, char *buf, int cap)
{
fz_error error;
int tok;
int len;
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse token in object stream");
switch (tok)
{
case PDF_TOK_OPEN_ARRAY:
error = pdf_parse_array(op, xref, file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse object stream");
break;
case PDF_TOK_OPEN_DICT:
error = pdf_parse_dict(op, xref, file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse object stream");
break;
case PDF_TOK_NAME: *op = fz_new_name(buf); break;
case PDF_TOK_REAL: *op = fz_new_real(fz_atof(buf)); break;
case PDF_TOK_STRING: *op = fz_new_string(buf, len); break;
case PDF_TOK_TRUE: *op = fz_new_bool(1); break;
case PDF_TOK_FALSE: *op = fz_new_bool(0); break;
case PDF_TOK_NULL: *op = fz_new_null(); break;
case PDF_TOK_INT: *op = fz_new_int(atoi(buf)); break;
default: return fz_throw("unknown token in object stream");
}
return fz_okay;
}
pdf_obj *
pdf_parse_stm_obj(pdf_document *doc, fz_stream *file, pdf_lexbuf *buf)
{
pdf_token tok;
fz_context *ctx = file->ctx;
tok = pdf_lex(file, buf);
switch (tok)
{
case PDF_TOK_OPEN_ARRAY:
return pdf_parse_array(doc, file, buf);
case PDF_TOK_OPEN_DICT:
return pdf_parse_dict(doc, file, buf);
case PDF_TOK_NAME: return pdf_new_name(doc, buf->scratch); break;
case PDF_TOK_REAL: return pdf_new_real(doc, buf->f); break;
case PDF_TOK_STRING: return pdf_new_string(doc, buf->scratch, buf->len); break;
case PDF_TOK_TRUE: return pdf_new_bool(doc, 1); break;
case PDF_TOK_FALSE: return pdf_new_bool(doc, 0); break;
case PDF_TOK_NULL: return pdf_new_null(doc); break;
case PDF_TOK_INT: return pdf_new_int(doc, buf->i); break;
default: fz_throw(ctx, FZ_ERROR_GENERIC, "unknown token in object stream");
}
return NULL; /* Stupid MSVC */
}
fz_error
pdf_parsestmobj(fz_obj **op, pdf_xref *xref, fz_stream *file, char *buf, int cap)
{
fz_error error;
pdf_token_e tok;
int len;
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse token in object stream");
switch (tok)
{
case PDF_TOARRAY:
error = pdf_parsearray(op, xref, file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse object stream");
break;
case PDF_TODICT:
error = pdf_parsedict(op, xref, file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse object stream");
break;
case PDF_TNAME: *op = fz_newname(buf); break;
case PDF_TREAL: *op = fz_newreal(atof(buf)); break;
case PDF_TSTRING: *op = fz_newstring(buf, len); break;
case PDF_TTRUE: *op = fz_newbool(1); break;
case PDF_TFALSE: *op = fz_newbool(0); break;
case PDF_TNULL: *op = fz_newnull(); break;
case PDF_TINT: *op = fz_newint(atoi(buf)); break;
default: return fz_throw("unknown token in object stream");
}
return fz_okay;
}
static int
pdf_lex_cmap(fz_stream *file, pdf_lexbuf *buf)
{
int tok = pdf_lex(file, buf);
/* RJW: Lost debugging here: "cannot parse cmap token" */
if (tok == PDF_TOK_KEYWORD)
tok = pdf_cmap_token_from_keyword(buf->scratch);
return tok;
}
static int
pdf_lex_cmap(fz_stream *file, char *buf, int n, int *sl)
{
int tok = pdf_lex(file, buf, n, sl);
/* RJW: Lost debugging here: "cannot parse cmap token" */
if (tok == PDF_TOK_KEYWORD)
tok = pdf_cmap_token_from_keyword(buf);
return tok;
}
static fz_error lexcmap(pdf_token_e *tok, fz_stream *file, char *buf, int n, int *sl)
{
fz_error error;
error = pdf_lex(tok, file, buf, n, sl);
if (error)
return fz_rethrow(error, "cannot parse cmap token");
if (*tok == PDF_TKEYWORD)
*tok = pdf_cmaptokenfromkeyword(buf);
return fz_okay;
}
pdf_obj *
pdf_parse_stm_obj(pdf_document *xref, fz_stream *file, pdf_lexbuf *buf)
{
int tok;
fz_context *ctx = file->ctx;
tok = pdf_lex(file, buf);
/* RJW: "cannot parse token in object stream") */
switch (tok)
{
case PDF_TOK_OPEN_ARRAY:
return pdf_parse_array(xref, file, buf);
/* RJW: "cannot parse object stream" */
case PDF_TOK_OPEN_DICT:
return pdf_parse_dict(xref, file, buf);
/* RJW: "cannot parse object stream" */
case PDF_TOK_NAME:
return fz_new_name(ctx, buf->scratch);
break;
case PDF_TOK_REAL:
return pdf_new_real(ctx, buf->f);
break;
case PDF_TOK_STRING:
return pdf_new_string(ctx, buf->scratch, buf->len);
break;
case PDF_TOK_TRUE:
return pdf_new_bool(ctx, 1);
break;
case PDF_TOK_FALSE:
return pdf_new_bool(ctx, 0);
break;
case PDF_TOK_NULL:
return pdf_new_null(ctx);
break;
case PDF_TOK_INT:
return pdf_new_int(ctx, buf->i);
break;
default:
fz_throw(ctx, "unknown token in object stream");
}
return NULL; /* Stupid MSVC */
}
fz_error *
pdf_parsestmobj(fz_obj **op, fz_stream *file, char *buf, int cap)
{
int tok, len;
tok = pdf_lex(file, buf, cap, &len);
switch (tok)
{
case PDF_TOARRAY: return pdf_parsearray(op, file, buf, cap);
case PDF_TODICT: return pdf_parsedict(op, file, buf, cap);
case PDF_TNAME: return fz_newname(op, buf);
case PDF_TREAL: return fz_newreal(op, atof(buf));
case PDF_TSTRING: return fz_newstring(op, buf, len);
case PDF_TTRUE: return fz_newbool(op, 1);
case PDF_TFALSE: return fz_newbool(op, 0);
case PDF_TNULL: return fz_newnull(op);
case PDF_TINT: return fz_newint(op, atoi(buf));
}
return fz_throw("syntaxerror: corrupt object stream");
}
static fz_error
pdf_readoldxref(fz_obj **trailerp, pdf_xref *xref, char *buf, int cap)
{
fz_error error;
int ofs, len;
char *s;
int n;
pdf_token_e tok;
int i;
int c;
pdf_logxref("load old xref format\n");
fz_readline(xref->file, buf, cap);
if (strncmp(buf, "xref", 4) != 0)
return fz_throw("cannot find xref marker");
while (1)
{
c = fz_peekbyte(xref->file);
if (!(c >= '0' && c <= '9'))
break;
fz_readline(xref->file, buf, cap);
s = buf;
ofs = atoi(fz_strsep(&s, " "));
len = atoi(fz_strsep(&s, " "));
/* broken pdfs where the section is not on a separate line */
if (s && *s != '\0')
{
fz_warn("broken xref section. proceeding anyway.");
fz_seek(xref->file, -(2 + (int)strlen(s)), 1);
}
/* broken pdfs where size in trailer undershoots entries in xref sections */
if (ofs + len > xref->cap)
{
fz_warn("broken xref section, proceeding anyway.");
xref->cap = ofs + len;
xref->table = fz_realloc(xref->table, xref->cap * sizeof(pdf_xrefentry));
}
if ((ofs + len) > xref->len)
{
for (i = xref->len; i < (ofs + len); i++)
{
xref->table[i].ofs = 0;
xref->table[i].gen = 0;
xref->table[i].stmofs = 0;
xref->table[i].obj = nil;
xref->table[i].type = 0;
}
xref->len = ofs + len;
}
for (i = ofs; i < ofs + len; i++)
{
n = fz_read(xref->file, (unsigned char *) buf, 20);
if (n < 0)
return fz_rethrow(n, "cannot read xref table");
if (!xref->table[i].type)
{
s = buf;
/* broken pdfs where line start with white space */
while (*s != '\0' && iswhite(*s))
s++;
xref->table[i].ofs = atoi(s);
xref->table[i].gen = atoi(s + 11);
xref->table[i].type = s[17];
}
}
}
error = pdf_lex(&tok, xref->file, buf, cap, &n);
if (error)
return fz_rethrow(error, "cannot parse trailer");
if (tok != PDF_TTRAILER)
return fz_throw("expected trailer marker");
error = pdf_lex(&tok, xref->file, buf, cap, &n);
if (error)
return fz_rethrow(error, "cannot parse trailer");
if (tok != PDF_TODICT)
return fz_throw("expected trailer dictionary");
error = pdf_parsedict(trailerp, xref, xref->file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse trailer");
return fz_okay;
}
fz_error
pdf_parseindobj(fz_obj **op, pdf_xref *xref,
fz_stream *file, char *buf, int cap,
int *onum, int *ogen, int *ostmofs)
{
fz_error error = fz_okay;
fz_obj *obj = nil;
int num = 0, gen = 0, stmofs;
pdf_token_e tok;
int len;
int a, b;
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
if (tok != PDF_TINT)
return fz_throw("cannot parse indirect object (%d %d R)", num, gen);
num = atoi(buf);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
if (tok != PDF_TINT)
return fz_throw("cannot parse indirect object (%d %d R)", num, gen);
gen = atoi(buf);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
if (tok != PDF_TOBJ)
return fz_throw("cannot parse indirect object (%d %d R)", num, gen);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
switch (tok)
{
case PDF_TOARRAY:
error = pdf_parsearray(&obj, xref, file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
break;
case PDF_TODICT:
error = pdf_parsedict(&obj, xref, file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
break;
case PDF_TNAME: obj = fz_newname(buf); break;
case PDF_TREAL: obj = fz_newreal(atof(buf)); break;
case PDF_TSTRING: obj = fz_newstring(buf, len); break;
case PDF_TTRUE: obj = fz_newbool(1); break;
case PDF_TFALSE: obj = fz_newbool(0); break;
case PDF_TNULL: obj = fz_newnull(); break;
case PDF_TINT:
a = atoi(buf);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
if (tok == PDF_TSTREAM || tok == PDF_TENDOBJ)
{
obj = fz_newint(a);
goto skip;
}
if (tok == PDF_TINT)
{
b = atoi(buf);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
if (tok == PDF_TR)
{
obj = fz_newindirect(a, b, xref);
break;
}
}
return fz_throw("cannot parse indirect object (%d %d R)", num, gen);
case PDF_TENDOBJ:
obj = fz_newnull();
goto skip;
default:
return fz_throw("cannot parse indirect object (%d %d R)", num, gen);
}
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
{
fz_dropobj(obj);
return fz_rethrow(error, "cannot parse indirect object (%d %d R)", num, gen);
}
skip:
if (tok == PDF_TSTREAM)
{
int c = fz_readbyte(file);
while (c == ' ')
c = fz_readbyte(file);
if (c == '\r')
{
c = fz_peekbyte(file);
if (c != '\n')
fz_warn("line feed missing after stream begin marker (%d %d R)", num, gen);
else
fz_readbyte(file);
}
stmofs = fz_tell(file);
}
else if (tok == PDF_TENDOBJ)
{
stmofs = 0;
}
else
{
fz_warn("expected endobj or stream keyword (%d %d R)", num, gen);
stmofs = 0;
}
if (onum) *onum = num;
if (ogen) *ogen = gen;
if (ostmofs) *ostmofs = stmofs;
*op = obj;
return fz_okay;
}
static fz_error
fz_repairobj(fz_stream *file, char *buf, int cap, int *stmofsp, int *stmlenp, fz_obj **encrypt, fz_obj **id)
{
fz_error error;
int tok;
int stmlen;
int len;
int n;
*stmofsp = 0;
*stmlenp = -1;
stmlen = 0;
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot parse object");
if (tok == PDF_TODICT)
{
fz_obj *dict, *obj;
/* Send nil xref so we don't try to resolve references */
error = pdf_parsedict(&dict, nil, file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse object");
obj = fz_dictgets(dict, "Type");
if (fz_isname(obj) && !strcmp(fz_toname(obj), "XRef"))
{
obj = fz_dictgets(dict, "Encrypt");
if (obj)
{
if (*encrypt)
fz_dropobj(*encrypt);
*encrypt = fz_keepobj(obj);
}
obj = fz_dictgets(dict, "ID");
if (obj)
{
if (*id)
fz_dropobj(*id);
*id = fz_keepobj(obj);
}
}
obj = fz_dictgets(dict, "Length");
if (fz_isint(obj))
stmlen = fz_toint(obj);
fz_dropobj(dict);
}
while ( tok != PDF_TSTREAM &&
tok != PDF_TENDOBJ &&
tok != PDF_TERROR &&
tok != PDF_TEOF )
{
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot scan for endobj or stream token");
}
if (tok == PDF_TSTREAM)
{
int c = fz_readbyte(file);
if (c == '\r') {
c = fz_peekbyte(file);
if (c == '\n')
fz_readbyte(file);
}
*stmofsp = fz_tell(file);
if (*stmofsp < 0)
return fz_throw("cannot seek in file");
if (stmlen > 0)
{
fz_seek(file, *stmofsp + stmlen, 0);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
fz_catch(error, "cannot find endstream token, falling back to scanning");
if (tok == PDF_TENDSTREAM)
goto atobjend;
fz_seek(file, *stmofsp, 0);
}
n = fz_read(file, (unsigned char *) buf, 9);
if (n < 0)
return fz_rethrow(n, "cannot read from file");
while (memcmp(buf, "endstream", 9) != 0)
{
c = fz_readbyte(file);
if (c == EOF)
break;
memmove(buf, buf + 1, 8);
buf[8] = c;
}
*stmlenp = fz_tell(file) - *stmofsp - 9;
atobjend:
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
return fz_rethrow(error, "cannot scan for endobj token");
if (tok != PDF_TENDOBJ)
fz_warn("object missing 'endobj' token");
}
return fz_okay;
}
fz_error
pdf_parsedict(fz_obj **op, pdf_xref *xref, fz_stream *file, char *buf, int cap)
{
fz_error error = fz_okay;
fz_obj *dict = nil;
fz_obj *key = nil;
fz_obj *val = nil;
pdf_token_e tok;
int len;
int a, b;
dict = fz_newdict(8);
while (1)
{
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
{
fz_dropobj(dict);
return fz_rethrow(error, "cannot parse dict");
}
skip:
if (tok == PDF_TCDICT)
{
*op = dict;
return fz_okay;
}
/* for BI .. ID .. EI in content streams */
if (tok == PDF_TKEYWORD && !strcmp(buf, "ID"))
{
*op = dict;
return fz_okay;
}
if (tok != PDF_TNAME)
{
fz_dropobj(dict);
return fz_throw("invalid key in dict");;
}
key = fz_newname(buf);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
{
fz_dropobj(dict);
return fz_rethrow(error, "cannot parse dict");
}
switch (tok)
{
case PDF_TOARRAY:
error = pdf_parsearray(&val, xref, file, buf, cap);
if (error)
{
fz_dropobj(key);
fz_dropobj(dict);
return fz_rethrow(error, "cannot parse dict");
}
break;
case PDF_TODICT:
error = pdf_parsedict(&val, xref, file, buf, cap);
if (error)
{
fz_dropobj(key);
fz_dropobj(dict);
return fz_rethrow(error, "cannot parse dict");
}
break;
case PDF_TNAME: val = fz_newname(buf); break;
case PDF_TREAL: val = fz_newreal(atof(buf)); break;
case PDF_TSTRING: val = fz_newstring(buf, len); break;
case PDF_TTRUE: val = fz_newbool(1); break;
case PDF_TFALSE: val = fz_newbool(0); break;
case PDF_TNULL: val = fz_newnull(); break;
case PDF_TINT:
/* 64-bit to allow for numbers > INT_MAX and overflow */
a = (int) strtoll(buf, 0, 10);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
{
fz_dropobj(key);
fz_dropobj(dict);
return fz_rethrow(error, "cannot parse dict");
}
if (tok == PDF_TCDICT || tok == PDF_TNAME ||
(tok == PDF_TKEYWORD && !strcmp(buf, "ID")))
{
val = fz_newint(a);
fz_dictput(dict, key, val);
fz_dropobj(val);
fz_dropobj(key);
goto skip;
}
if (tok == PDF_TINT)
{
b = atoi(buf);
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
{
fz_dropobj(key);
fz_dropobj(dict);
return fz_rethrow(error, "cannot parse dict");
}
if (tok == PDF_TR)
{
val = fz_newindirect(a, b, xref);
break;
}
}
fz_dropobj(key);
fz_dropobj(dict);
return fz_throw("invalid indirect reference in dict");
default:
return fz_throw("unknown token in dict");
}
fz_dictput(dict, key, val);
fz_dropobj(val);
fz_dropobj(key);
}
}
fz_error
pdf_parsearray(fz_obj **op, pdf_xref *xref, fz_stream *file, char *buf, int cap)
{
fz_error error = fz_okay;
fz_obj *ary = nil;
fz_obj *obj = nil;
int a = 0, b = 0, n = 0;
pdf_token_e tok;
int len;
ary = fz_newarray(4);
while (1)
{
error = pdf_lex(&tok, file, buf, cap, &len);
if (error)
{
fz_dropobj(ary);
return fz_rethrow(error, "cannot parse array");
}
if (tok != PDF_TINT && tok != PDF_TR)
{
if (n > 0)
{
obj = fz_newint(a);
fz_arraypush(ary, obj);
fz_dropobj(obj);
}
if (n > 1)
{
obj = fz_newint(b);
fz_arraypush(ary, obj);
fz_dropobj(obj);
}
n = 0;
}
if (tok == PDF_TINT && n == 2)
{
obj = fz_newint(a);
fz_arraypush(ary, obj);
fz_dropobj(obj);
a = b;
n --;
}
switch (tok)
{
case PDF_TCARRAY:
*op = ary;
return fz_okay;
case PDF_TINT:
if (n == 0)
a = atoi(buf);
if (n == 1)
b = atoi(buf);
n ++;
break;
case PDF_TR:
if (n != 2)
{
fz_dropobj(ary);
return fz_throw("cannot parse indirect reference in array");
}
obj = fz_newindirect(a, b, xref);
fz_arraypush(ary, obj);
fz_dropobj(obj);
n = 0;
break;
case PDF_TOARRAY:
error = pdf_parsearray(&obj, xref, file, buf, cap);
if (error)
{
fz_dropobj(ary);
return fz_rethrow(error, "cannot parse array");
}
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
case PDF_TODICT:
error = pdf_parsedict(&obj, xref, file, buf, cap);
if (error)
{
fz_dropobj(ary);
return fz_rethrow(error, "cannot parse array");
}
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
case PDF_TNAME:
obj = fz_newname(buf);
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
case PDF_TREAL:
obj = fz_newreal(atof(buf));
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
case PDF_TSTRING:
obj = fz_newstring(buf, len);
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
case PDF_TTRUE:
obj = fz_newbool(1);
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
case PDF_TFALSE:
obj = fz_newbool(0);
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
case PDF_TNULL:
obj = fz_newnull();
fz_arraypush(ary, obj);
fz_dropobj(obj);
break;
default:
fz_dropobj(ary);
return fz_throw("cannot parse token in array");
}
}
}
fz_error
pdf_repairxref(pdf_xref *xref, char *buf, int bufsize)
{
fz_error error;
fz_obj *dict, *obj;
fz_obj *length;
fz_obj *encrypt = nil;
fz_obj *id = nil;
fz_obj *root = nil;
fz_obj *info = nil;
struct entry *list = nil;
int listlen;
int listcap;
int maxnum = 0;
int num = 0;
int gen = 0;
int tmpofs, numofs = 0, genofs = 0;
int stmlen, stmofs = 0;
int tok;
int next;
int i, n;
pdf_logxref("repairxref %p\n", xref);
fz_seek(xref->file, 0, 0);
listlen = 0;
listcap = 1024;
list = fz_calloc(listcap, sizeof(struct entry));
/* look for '%PDF' version marker within first kilobyte of file */
n = fz_read(xref->file, (unsigned char *)buf, MAX(bufsize, 1024));
if (n < 0)
{
error = fz_rethrow(n, "cannot read from file");
goto cleanup;
}
fz_seek(xref->file, 0, 0);
for (i = 0; i < n - 4; i++)
{
if (memcmp(buf + i, "%PDF", 4) == 0)
{
fz_seek(xref->file, i, 0);
break;
}
}
while (1)
{
tmpofs = fz_tell(xref->file);
if (tmpofs < 0)
{
error = fz_throw("cannot tell in file");
goto cleanup;
}
error = pdf_lex(&tok, xref->file, buf, bufsize, &n);
if (error)
{
fz_catch(error, "ignoring the rest of the file");
break;
}
if (tok == PDF_TINT)
{
numofs = genofs;
num = gen;
genofs = tmpofs;
gen = atoi(buf);
}
if (tok == PDF_TOBJ)
{
error = fz_repairobj(xref->file, buf, bufsize, &stmofs, &stmlen, &encrypt, &id);
if (error)
{
error = fz_rethrow(error, "cannot parse object (%d %d R)", num, gen);
goto cleanup;
}
pdf_logxref("found object: (%d %d R)\n", num, gen);
if (listlen + 1 == listcap)
{
listcap = (listcap * 3) / 2;
list = fz_realloc(list, listcap, sizeof(struct entry));
}
list[listlen].num = num;
list[listlen].gen = gen;
list[listlen].ofs = numofs;
list[listlen].stmofs = stmofs;
list[listlen].stmlen = stmlen;
listlen ++;
if (num > maxnum)
maxnum = num;
}
/* trailer dictionary */
if (tok == PDF_TODICT)
{
error = pdf_parsedict(&dict, xref, xref->file, buf, bufsize);
if (error)
{
error = fz_rethrow(error, "cannot parse object");
goto cleanup;
}
obj = fz_dictgets(dict, "Encrypt");
if (obj)
{
if (encrypt)
fz_dropobj(encrypt);
encrypt = fz_keepobj(obj);
}
obj = fz_dictgets(dict, "ID");
if (obj)
{
if (id)
fz_dropobj(id);
id = fz_keepobj(obj);
}
obj = fz_dictgets(dict, "Root");
if (obj)
{
if (root)
fz_dropobj(root);
root = fz_keepobj(obj);
}
obj = fz_dictgets(dict, "Info");
if (obj)
{
if (info)
fz_dropobj(info);
info = fz_keepobj(obj);
}
fz_dropobj(dict);
}
if (tok == PDF_TERROR)
fz_readbyte(xref->file);
if (tok == PDF_TEOF)
break;
}
/* make xref reasonable */
pdf_resizexref(xref, maxnum + 1);
for (i = 0; i < listlen; i++)
{
xref->table[list[i].num].type = 'n';
xref->table[list[i].num].ofs = list[i].ofs;
xref->table[list[i].num].gen = list[i].gen;
xref->table[list[i].num].stmofs = list[i].stmofs;
/* corrected stream length */
if (list[i].stmlen >= 0)
{
pdf_logxref("correct stream length %d %d = %d\n",
list[i].num, list[i].gen, list[i].stmlen);
error = pdf_loadobject(&dict, xref, list[i].num, list[i].gen);
if (error)
{
error = fz_rethrow(error, "cannot load stream object (%d %d R)", list[i].num, list[i].gen);
goto cleanup;
}
length = fz_newint(list[i].stmlen);
fz_dictputs(dict, "Length", length);
fz_dropobj(length);
fz_dropobj(dict);
}
}
xref->table[0].type = 'f';
xref->table[0].ofs = 0;
xref->table[0].gen = 65535;
xref->table[0].stmofs = 0;
xref->table[0].obj = nil;
next = 0;
for (i = xref->len - 1; i >= 0; i--)
{
if (xref->table[i].type == 'f')
{
xref->table[i].ofs = next;
if (xref->table[i].gen < 65535)
xref->table[i].gen ++;
next = i;
}
}
/* create a repaired trailer, Root will be added later */
xref->trailer = fz_newdict(5);
obj = fz_newint(maxnum + 1);
fz_dictputs(xref->trailer, "Size", obj);
fz_dropobj(obj);
if (root)
{
fz_dictputs(xref->trailer, "Root", root);
fz_dropobj(root);
}
if (info)
{
fz_dictputs(xref->trailer, "Info", info);
fz_dropobj(info);
}
if (encrypt)
{
if (fz_isindirect(encrypt))
{
/* create new reference with non-nil xref pointer */
obj = fz_newindirect(fz_tonum(encrypt), fz_togen(encrypt), xref);
fz_dropobj(encrypt);
encrypt = obj;
}
fz_dictputs(xref->trailer, "Encrypt", encrypt);
fz_dropobj(encrypt);
}
if (id)
{
if (fz_isindirect(id))
{
/* create new reference with non-nil xref pointer */
obj = fz_newindirect(fz_tonum(id), fz_togen(id), xref);
fz_dropobj(id);
id = obj;
}
fz_dictputs(xref->trailer, "ID", id);
fz_dropobj(id);
}
fz_free(list);
return fz_okay;
cleanup:
if (encrypt) fz_dropobj(encrypt);
if (id) fz_dropobj(id);
if (root) fz_dropobj(root);
if (info) fz_dropobj(info);
fz_free(list);
return error; /* already rethrown */
}
pdf_obj *
pdf_parse_ind_obj(fz_context *ctx, pdf_document *doc,
fz_stream *file, pdf_lexbuf *buf,
int *onum, int *ogen, fz_off_t *ostmofs, int *try_repair)
{
pdf_obj *obj = NULL;
int num = 0, gen = 0;
fz_off_t stm_ofs;
pdf_token tok;
fz_off_t a, b;
fz_var(obj);
tok = pdf_lex(ctx, file, buf);
if (tok != PDF_TOK_INT)
{
if (try_repair)
*try_repair = 1;
fz_throw(ctx, FZ_ERROR_GENERIC, "expected object number");
}
num = buf->i;
tok = pdf_lex(ctx, file, buf);
if (tok != PDF_TOK_INT)
{
if (try_repair)
*try_repair = 1;
fz_throw(ctx, FZ_ERROR_GENERIC, "expected generation number (%d ? obj)", num);
}
gen = buf->i;
tok = pdf_lex(ctx, file, buf);
if (tok != PDF_TOK_OBJ)
{
if (try_repair)
*try_repair = 1;
fz_throw(ctx, FZ_ERROR_GENERIC, "expected 'obj' keyword (%d %d ?)", num, gen);
}
tok = pdf_lex(ctx, file, buf);
switch (tok)
{
case PDF_TOK_OPEN_ARRAY:
obj = pdf_parse_array(ctx, doc, file, buf);
break;
case PDF_TOK_OPEN_DICT:
obj = pdf_parse_dict(ctx, doc, file, buf);
break;
case PDF_TOK_NAME: obj = pdf_new_name(ctx, doc, buf->scratch); break;
case PDF_TOK_REAL: obj = pdf_new_real(ctx, doc, buf->f); break;
case PDF_TOK_STRING: obj = pdf_new_string(ctx, doc, buf->scratch, buf->len); break;
case PDF_TOK_TRUE: obj = pdf_new_bool(ctx, doc, 1); break;
case PDF_TOK_FALSE: obj = pdf_new_bool(ctx, doc, 0); break;
case PDF_TOK_NULL: obj = pdf_new_null(ctx, doc); break;
case PDF_TOK_INT:
a = buf->i;
tok = pdf_lex(ctx, file, buf);
if (tok == PDF_TOK_STREAM || tok == PDF_TOK_ENDOBJ)
{
obj = pdf_new_int_offset(ctx, doc, a);
goto skip;
}
if (tok == PDF_TOK_INT)
{
b = buf->i;
tok = pdf_lex(ctx, file, buf);
if (tok == PDF_TOK_R)
{
obj = pdf_new_indirect(ctx, doc, a, b);
break;
}
}
fz_throw(ctx, FZ_ERROR_GENERIC, "expected 'R' keyword (%d %d R)", num, gen);
case PDF_TOK_ENDOBJ:
obj = pdf_new_null(ctx, doc);
goto skip;
default:
fz_throw(ctx, FZ_ERROR_GENERIC, "syntax error in object (%d %d R)", num, gen);
}
fz_try(ctx)
{
tok = pdf_lex(ctx, file, buf);
}
fz_catch(ctx)
{
pdf_drop_obj(ctx, obj);
fz_rethrow(ctx);
}
skip:
if (tok == PDF_TOK_STREAM)
{
int c = fz_read_byte(ctx, file);
while (c == ' ')
c = fz_read_byte(ctx, file);
if (c == '\r')
{
c = fz_peek_byte(ctx, file);
if (c != '\n')
fz_warn(ctx, "line feed missing after stream begin marker (%d %d R)", num, gen);
else
fz_read_byte(ctx, file);
}
stm_ofs = fz_tell(ctx, file);
}
else if (tok == PDF_TOK_ENDOBJ)
{
stm_ofs = 0;
}
else
{
fz_warn(ctx, "expected 'endobj' or 'stream' keyword (%d %d R)", num, gen);
stm_ofs = 0;
}
if (onum) *onum = num;
if (ogen) *ogen = gen;
if (ostmofs) *ostmofs = stm_ofs;
return obj;
}
fz_error
pdf_loadobjstm(pdf_xref *xref, int oid, int gen, char *buf, int cap)
{
fz_error error;
fz_stream *stm;
fz_obj *objstm;
int *oidbuf;
int *ofsbuf;
fz_obj *obj;
int first;
int count;
int i, n;
pdf_token_e tok;
pdf_logxref("loadobjstm (%d %d R)\n", oid, gen);
error = pdf_loadobject(&objstm, xref, oid, gen);
if (error)
return fz_rethrow(error, "cannot load object stream object");
count = fz_toint(fz_dictgets(objstm, "N"));
first = fz_toint(fz_dictgets(objstm, "First"));
pdf_logxref(" count %d\n", count);
oidbuf = fz_malloc(count * sizeof(int));
if (!oidbuf)
{
error = fz_rethrow(-1, "out of memory: object id buffer");
goto cleanupobj;
}
ofsbuf = fz_malloc(count * sizeof(int));
if (!ofsbuf)
{
error = fz_rethrow(-1, "out of memory: offset buffer");
goto cleanupoid;
}
error = pdf_openstream(&stm, xref, oid, gen);
if (error)
{
error = fz_rethrow(error, "cannot open object stream");
goto cleanupofs;
}
for (i = 0; i < count; i++)
{
error = pdf_lex(&tok, stm, buf, cap, &n);
if (error || tok != PDF_TINT)
{
error = fz_rethrow(error, "corrupt object stream");
goto cleanupstm;
}
oidbuf[i] = atoi(buf);
error = pdf_lex(&tok, stm, buf, cap, &n);
if (error || tok != PDF_TINT)
{
error = fz_rethrow(error, "corrupt object stream");
goto cleanupstm;
}
ofsbuf[i] = atoi(buf);
}
error = fz_seek(stm, first, 0);
if (error)
{
error = fz_rethrow(error, "cannot seek in object stream");
goto cleanupstm;
}
for (i = 0; i < count; i++)
{
/* FIXME: seek to first + ofsbuf[i] */
error = pdf_parsestmobj(&obj, xref, stm, buf, cap);
if (error)
{
error = fz_rethrow(error, "cannot parse object %d in stream", i);
goto cleanupstm;
}
if (oidbuf[i] < 1 || oidbuf[i] >= xref->len)
{
fz_dropobj(obj);
error = fz_throw("object id (%d 0 R) out of range (0..%d)", oidbuf[i], xref->len - 1);
goto cleanupstm;
}
if (xref->table[oidbuf[i]].obj)
fz_dropobj(xref->table[oidbuf[i]].obj);
xref->table[oidbuf[i]].obj = obj;
}
fz_dropstream(stm);
fz_free(ofsbuf);
fz_free(oidbuf);
fz_dropobj(objstm);
return fz_okay;
cleanupstm:
fz_dropstream(stm);
cleanupofs:
fz_free(ofsbuf);
cleanupoid:
fz_free(oidbuf);
cleanupobj:
fz_dropobj(objstm);
return error; /* already rethrown */
}
pdf_obj *
pdf_parse_array(pdf_document *xref, fz_stream *file, pdf_lexbuf *buf)
{
pdf_obj *ary = NULL;
pdf_obj *obj = NULL;
int a = 0, b = 0, n = 0;
int tok;
fz_context *ctx = file->ctx;
pdf_obj *op;
fz_var(obj);
ary = pdf_new_array(ctx, 4);
fz_try(ctx)
{
while (1)
{
tok = pdf_lex(file, buf);
if (tok != PDF_TOK_INT && tok != PDF_TOK_R)
{
if (n > 0)
{
obj = pdf_new_int(ctx, a);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
}
if (n > 1)
{
obj = pdf_new_int(ctx, b);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
}
n = 0;
}
if (tok == PDF_TOK_INT && n == 2)
{
obj = pdf_new_int(ctx, a);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
a = b;
n --;
}
switch (tok)
{
case PDF_TOK_CLOSE_ARRAY:
op = ary;
goto end;
case PDF_TOK_INT:
if (n == 0)
a = buf->i;
if (n == 1)
b = buf->i;
n ++;
break;
case PDF_TOK_R:
if (n != 2)
fz_throw(ctx, "cannot parse indirect reference in array");
obj = pdf_new_indirect(ctx, a, b, xref);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
n = 0;
break;
case PDF_TOK_OPEN_ARRAY:
obj = pdf_parse_array(xref, file, buf);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
case PDF_TOK_OPEN_DICT:
obj = pdf_parse_dict(xref, file, buf);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
case PDF_TOK_NAME:
obj = fz_new_name(ctx, buf->scratch);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
case PDF_TOK_REAL:
obj = pdf_new_real(ctx, buf->f);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
case PDF_TOK_STRING:
obj = pdf_new_string(ctx, buf->scratch, buf->len);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
case PDF_TOK_TRUE:
obj = pdf_new_bool(ctx, 1);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
case PDF_TOK_FALSE:
obj = pdf_new_bool(ctx, 0);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
case PDF_TOK_NULL:
obj = pdf_new_null(ctx);
pdf_array_push(ary, obj);
pdf_drop_obj(obj);
obj = NULL;
break;
default:
fz_throw(ctx, "cannot parse token in array");
}
}
end:
{}
}
fz_catch(ctx)
{
pdf_drop_obj(obj);
pdf_drop_obj(ary);
fz_throw(ctx, "cannot parse array");
}
return op;
}
fz_error *
pdf_parsearray(fz_obj **op, fz_stream *file, char *buf, int cap)
{
fz_error *error = nil;
fz_obj *ary = nil;
fz_obj *obj = nil;
int a = 0, b = 0, n = 0;
int tok, len;
error = fz_newarray(op, 4);
if (error) return error;
ary = *op;
while (1)
{
tok = pdf_lex(file, buf, cap, &len);
if (tok != PDF_TINT && tok != PDF_TR)
{
if (n > 0)
{
error = fz_newint(&obj, a);
if (error) goto cleanup;
error = fz_arraypush(ary, obj);
if (error) goto cleanup;
fz_dropobj(obj);
obj = nil;
}
if (n > 1)
{
error = fz_newint(&obj, b);
if (error) goto cleanup;
error = fz_arraypush(ary, obj);
if (error) goto cleanup;
fz_dropobj(obj);
obj = nil;
}
n = 0;
}
if (tok == PDF_TINT && n == 2)
{
error = fz_newint(&obj, a);
if (error) goto cleanup;
error = fz_arraypush(ary, obj);
if (error) goto cleanup;
fz_dropobj(obj);
obj = nil;
a = b;
n --;
}
switch (tok)
{
case PDF_TCARRAY:
return nil;
case PDF_TINT:
if (n == 0)
a = atoi(buf);
if (n == 1)
b = atoi(buf);
n ++;
break;
case PDF_TR:
if (n != 2)
goto cleanup;
error = fz_newindirect(&obj, a, b);
if (error) goto cleanup;
n = 0;
break;
case PDF_TOARRAY: error = pdf_parsearray(&obj, file, buf, cap); break;
case PDF_TODICT: error = pdf_parsedict(&obj, file, buf, cap); break;
case PDF_TNAME: error = fz_newname(&obj, buf); break;
case PDF_TREAL: error = fz_newreal(&obj, atof(buf)); break;
case PDF_TSTRING: error = fz_newstring(&obj, buf, len); break;
case PDF_TTRUE: error = fz_newbool(&obj, 1); break;
case PDF_TFALSE: error = fz_newbool(&obj, 0); break;
case PDF_TNULL: error = fz_newnull(&obj); break;
default: goto cleanup;
}
if (error) goto cleanup;
if (obj)
{
error = fz_arraypush(ary, obj);
if (error) goto cleanup;
fz_dropobj(obj);
}
obj = nil;
}
cleanup:
if (obj) fz_dropobj(obj);
if (ary) fz_dropobj(ary);
if (error) return error;
return fz_throw("syntaxerror: corrupt array");
}
fz_error *
pdf_parseindobj(fz_obj **op, fz_stream *file, char *buf, int cap,
int *ooid, int *ogid, int *ostmofs)
{
fz_error *error = nil;
fz_obj *obj = nil;
int oid = 0, gid = 0, stmofs;
int tok, len;
int a, b;
tok = pdf_lex(file, buf, cap, &len);
if (tok != PDF_TINT)
goto cleanup;
oid = atoi(buf);
tok = pdf_lex(file, buf, cap, &len);
if (tok != PDF_TINT)
goto cleanup;
gid = atoi(buf);
tok = pdf_lex(file, buf, cap, &len);
if (tok != PDF_TOBJ)
goto cleanup;
tok = pdf_lex(file, buf, cap, &len);
switch (tok)
{
case PDF_TOARRAY: error = pdf_parsearray(&obj, file, buf, cap); break;
case PDF_TODICT: error = pdf_parsedict(&obj, file, buf, cap); break;
case PDF_TNAME: error = fz_newname(&obj, buf); break;
case PDF_TREAL: error = fz_newreal(&obj, atof(buf)); break;
case PDF_TSTRING: error = fz_newstring(&obj, buf, len); break;
case PDF_TTRUE: error = fz_newbool(&obj, 1); break;
case PDF_TFALSE: error = fz_newbool(&obj, 0); break;
case PDF_TNULL: error = fz_newnull(&obj); break;
case PDF_TINT:
a = atoi(buf);
tok = pdf_lex(file, buf, cap, &len);
if (tok == PDF_TSTREAM || tok == PDF_TENDOBJ)
{
error = fz_newint(&obj, a);
if (error) goto cleanup;
goto skip;
}
if (tok == PDF_TINT)
{
b = atoi(buf);
tok = pdf_lex(file, buf, cap, &len);
if (tok == PDF_TR)
{
error = fz_newindirect(&obj, a, b);
break;
}
}
goto cleanup;
default:
goto cleanup;
}
if (error) goto cleanup;
tok = pdf_lex(file, buf, cap, &len);
skip:
if (tok == PDF_TSTREAM)
{
int c = fz_readbyte(file);
if (c == '\r')
{
c = fz_peekbyte(file);
if (c != '\n')
fz_warn("syntaxerror: DOS format line ending after stream keyword (%d %d)\n", oid, gid);
else
c = fz_readbyte(file);
}
stmofs = fz_tell(file);
}
else if (tok == PDF_TENDOBJ)
stmofs = 0;
else
goto cleanup;
if (ooid) *ooid = oid;
if (ogid) *ogid = gid;
if (ostmofs) *ostmofs = stmofs;
*op = obj;
return nil;
cleanup:
if (obj) fz_dropobj(obj);
if (error) return error;
return fz_throw("syntaxerror: corrupt indirect object (%d %d)", oid, gid);
}
static fz_error
pdf_loadobjstm(pdf_xref *xref, int num, int gen, char *buf, int cap)
{
fz_error error;
fz_stream *stm;
fz_obj *objstm;
int *numbuf;
int *ofsbuf;
fz_obj *obj;
int first;
int count;
int i, n;
pdf_token_e tok;
pdf_logxref("loadobjstm (%d %d R)\n", num, gen);
error = pdf_loadobject(&objstm, xref, num, gen);
if (error)
return fz_rethrow(error, "cannot load object stream object (%d %d R)", num, gen);
count = fz_toint(fz_dictgets(objstm, "N"));
first = fz_toint(fz_dictgets(objstm, "First"));
pdf_logxref("\tcount %d\n", count);
numbuf = fz_malloc(count * sizeof(int));
ofsbuf = fz_malloc(count * sizeof(int));
error = pdf_openstream(&stm, xref, num, gen);
if (error)
{
error = fz_rethrow(error, "cannot open object stream (%d %d R)", num, gen);
goto cleanupbuf;
}
for (i = 0; i < count; i++)
{
error = pdf_lex(&tok, stm, buf, cap, &n);
if (error || tok != PDF_TINT)
{
error = fz_rethrow(error, "corrupt object stream (%d %d R)", num, gen);
goto cleanupstm;
}
numbuf[i] = atoi(buf);
error = pdf_lex(&tok, stm, buf, cap, &n);
if (error || tok != PDF_TINT)
{
error = fz_rethrow(error, "corrupt object stream (%d %d R)", num, gen);
goto cleanupstm;
}
ofsbuf[i] = atoi(buf);
}
fz_seek(stm, first, 0);
for (i = 0; i < count; i++)
{
fz_seek(stm, first + ofsbuf[i], 0);
error = pdf_parsestmobj(&obj, xref, stm, buf, cap);
if (error)
{
error = fz_rethrow(error, "cannot parse object %d in stream (%d %d R)", i, num, gen);
goto cleanupstm;
}
if (numbuf[i] < 1 || numbuf[i] >= xref->len)
{
fz_dropobj(obj);
error = fz_throw("object id (%d 0 R) out of range (0..%d)", numbuf[i], xref->len - 1);
goto cleanupstm;
}
if (xref->table[numbuf[i]].type == 'o' && xref->table[numbuf[i]].ofs == num)
{
if (xref->table[numbuf[i]].obj)
fz_dropobj(xref->table[numbuf[i]].obj);
xref->table[numbuf[i]].obj = obj;
}
else
{
fz_dropobj(obj);
}
}
fz_close(stm);
fz_free(ofsbuf);
fz_free(numbuf);
fz_dropobj(objstm);
return fz_okay;
cleanupstm:
fz_close(stm);
cleanupbuf:
fz_free(ofsbuf);
fz_free(numbuf);
fz_dropobj(objstm);
return error; /* already rethrown */
}
pdf_obj *
pdf_parse_dict(pdf_document *doc, fz_stream *file, pdf_lexbuf *buf)
{
pdf_obj *dict;
pdf_obj *key = NULL;
pdf_obj *val = NULL;
pdf_token tok;
int a, b;
fz_context *ctx = file->ctx;
dict = pdf_new_dict(doc, 8);
fz_var(key);
fz_var(val);
fz_try(ctx)
{
while (1)
{
tok = pdf_lex(file, buf);
skip:
if (tok == PDF_TOK_CLOSE_DICT)
break;
/* for BI .. ID .. EI in content streams */
if (tok == PDF_TOK_KEYWORD && !strcmp(buf->scratch, "ID"))
break;
if (tok != PDF_TOK_NAME)
fz_throw(ctx, FZ_ERROR_GENERIC, "invalid key in dict");
key = pdf_new_name(doc, buf->scratch);
tok = pdf_lex(file, buf);
switch (tok)
{
case PDF_TOK_OPEN_ARRAY:
val = pdf_parse_array(doc, file, buf);
break;
case PDF_TOK_OPEN_DICT:
val = pdf_parse_dict(doc, file, buf);
break;
case PDF_TOK_NAME: val = pdf_new_name(doc, buf->scratch); break;
case PDF_TOK_REAL: val = pdf_new_real(doc, buf->f); break;
case PDF_TOK_STRING: val = pdf_new_string(doc, buf->scratch, buf->len); break;
case PDF_TOK_TRUE: val = pdf_new_bool(doc, 1); break;
case PDF_TOK_FALSE: val = pdf_new_bool(doc, 0); break;
case PDF_TOK_NULL: val = pdf_new_null(doc); break;
case PDF_TOK_INT:
/* 64-bit to allow for numbers > INT_MAX and overflow */
a = buf->i;
tok = pdf_lex(file, buf);
if (tok == PDF_TOK_CLOSE_DICT || tok == PDF_TOK_NAME ||
(tok == PDF_TOK_KEYWORD && !strcmp(buf->scratch, "ID")))
{
val = pdf_new_int(doc, a);
pdf_dict_put(dict, key, val);
pdf_drop_obj(val);
val = NULL;
pdf_drop_obj(key);
key = NULL;
goto skip;
}
if (tok == PDF_TOK_INT)
{
b = buf->i;
tok = pdf_lex(file, buf);
if (tok == PDF_TOK_R)
{
val = pdf_new_indirect(doc, a, b);
break;
}
}
fz_throw(ctx, FZ_ERROR_GENERIC, "invalid indirect reference in dict");
default:
fz_throw(ctx, FZ_ERROR_GENERIC, "unknown token in dict");
}
pdf_dict_put(dict, key, val);
pdf_drop_obj(val);
val = NULL;
pdf_drop_obj(key);
key = NULL;
}
}
fz_catch(ctx)
{
pdf_drop_obj(dict);
pdf_drop_obj(key);
pdf_drop_obj(val);
fz_rethrow_message(ctx, "cannot parse dict");
}
return dict;
}
static fz_error
pdf_read_old_xref(fz_obj **trailerp, pdf_xref *xref, char *buf, int cap)
{
fz_error error;
int ofs, len;
char *s;
int n;
int tok;
int i;
int c;
fz_read_line(xref->file, buf, cap);
if (strncmp(buf, "xref", 4) != 0)
return fz_throw("cannot find xref marker");
while (1)
{
c = fz_peek_byte(xref->file);
if (!(c >= '0' && c <= '9'))
break;
fz_read_line(xref->file, buf, cap);
s = buf;
ofs = atoi(fz_strsep(&s, " "));
len = atoi(fz_strsep(&s, " "));
/* broken pdfs where the section is not on a separate line */
if (s && *s != '\0')
{
fz_warn("broken xref section. proceeding anyway.");
fz_seek(xref->file, -(2 + (int)strlen(s)), 1);
}
/* broken pdfs where size in trailer undershoots entries in xref sections */
if (ofs + len > xref->len)
{
fz_warn("broken xref section, proceeding anyway.");
pdf_resize_xref(xref, ofs + len);
}
for (i = ofs; i < ofs + len; i++)
{
n = fz_read(xref->file, (unsigned char *) buf, 20);
if (n < 0)
return fz_rethrow(n, "cannot read xref table");
if (!xref->table[i].type)
{
s = buf;
/* broken pdfs where line start with white space */
while (*s != '\0' && iswhite(*s))
s++;
xref->table[i].ofs = atoi(s);
xref->table[i].gen = atoi(s + 11);
xref->table[i].type = s[17];
if (s[17] != 'f' && s[17] != 'n' && s[17] != 'o')
return fz_throw("unexpected xref type: %#x (%d %d R)", s[17], i, xref->table[i].gen);
}
}
}
error = pdf_lex(&tok, xref->file, buf, cap, &n);
if (error)
return fz_rethrow(error, "cannot parse trailer");
if (tok != PDF_TOK_TRAILER)
return fz_throw("expected trailer marker");
error = pdf_lex(&tok, xref->file, buf, cap, &n);
if (error)
return fz_rethrow(error, "cannot parse trailer");
if (tok != PDF_TOK_OPEN_DICT)
return fz_throw("expected trailer dictionary");
error = pdf_parse_dict(trailerp, xref, xref->file, buf, cap);
if (error)
return fz_rethrow(error, "cannot parse trailer");
return fz_okay;
}
fz_error *
pdf_parsedict(fz_obj **op, fz_stream *file, char *buf, int cap)
{
fz_error *error = nil;
fz_obj *dict = nil;
fz_obj *key = nil;
fz_obj *val = nil;
int tok, len;
int a, b;
error = fz_newdict(op, 8);
if (error) return error;
dict = *op;
while (1)
{
tok = pdf_lex(file, buf, cap, &len);
skip:
if (tok == PDF_TCDICT)
return nil;
/* for BI .. ID .. EI in content streams */
if (tok == PDF_TKEYWORD && !strcmp(buf, "ID"))
return nil;
if (tok != PDF_TNAME)
goto cleanup;
error = fz_newname(&key, buf);
if (error) goto cleanup;
tok = pdf_lex(file, buf, cap, &len);
switch (tok)
{
case PDF_TOARRAY: error = pdf_parsearray(&val, file, buf, cap); break;
case PDF_TODICT: error = pdf_parsedict(&val, file, buf, cap); break;
case PDF_TNAME: error = fz_newname(&val, buf); break;
case PDF_TREAL: error = fz_newreal(&val, atof(buf)); break;
case PDF_TSTRING: error = fz_newstring(&val, buf, len); break;
case PDF_TTRUE: error = fz_newbool(&val, 1); break;
case PDF_TFALSE: error = fz_newbool(&val, 0); break;
case PDF_TNULL: error = fz_newnull(&val); break;
case PDF_TINT:
a = atoi(buf);
tok = pdf_lex(file, buf, cap, &len);
if (tok == PDF_TCDICT || tok == PDF_TNAME ||
(tok == PDF_TKEYWORD && !strcmp(buf, "ID")))
{
error = fz_newint(&val, a);
if (error) goto cleanup;
error = fz_dictput(dict, key, val);
if (error) goto cleanup;
fz_dropobj(val);
fz_dropobj(key);
key = val = nil;
goto skip;
}
if (tok == PDF_TINT)
{
b = atoi(buf);
tok = pdf_lex(file, buf, cap, &len);
if (tok == PDF_TR)
{
error = fz_newindirect(&val, a, b);
break;
}
}
goto cleanup;
default:
goto cleanup;
}
if (error) goto cleanup;
error = fz_dictput(dict, key, val);
if (error) goto cleanup;
fz_dropobj(val);
fz_dropobj(key);
key = val = nil;
}
cleanup:
if (key) fz_dropobj(key);
if (val) fz_dropobj(val);
if (dict) fz_dropobj(dict);
if (error) return error;
return fz_throw("syntaxerror: corrupt dictionary");
}
pdf_obj *
pdf_parse_dict(pdf_document *xref, fz_stream *file, pdf_lexbuf *buf)
{
pdf_obj *dict;
pdf_obj *key = NULL;
pdf_obj *val = NULL;
int tok;
int a, b;
fz_context *ctx = file->ctx;
dict = pdf_new_dict(ctx, 8);
fz_var(key);
fz_var(val);
fz_try(ctx)
{
while (1)
{
tok = pdf_lex(file, buf);
skip:
if (tok == PDF_TOK_CLOSE_DICT)
break;
/* for BI .. ID .. EI in content streams */
if (tok == PDF_TOK_KEYWORD && !strcmp(buf->scratch, "ID"))
break;
if (tok != PDF_TOK_NAME)
fz_throw(ctx, "invalid key in dict");
key = fz_new_name(ctx, buf->scratch);
tok = pdf_lex(file, buf);
switch (tok)
{
case PDF_TOK_OPEN_ARRAY:
/* cf. http://code.google.com/p/sumatrapdf/issues/detail?id=1643 */
fz_try(ctx)
{
val = pdf_parse_array(xref, file, buf);
}
fz_catch(ctx)
{
fz_warn(ctx, "ignoring broken array for '%s'", pdf_to_name(key));
pdf_drop_obj(key);
val = key = NULL;
do
tok = pdf_lex(file, buf);
while (tok != PDF_TOK_CLOSE_DICT && tok != PDF_TOK_CLOSE_ARRAY &&
tok != PDF_TOK_EOF && tok != PDF_TOK_OPEN_ARRAY && tok != PDF_TOK_OPEN_DICT);
if (tok == PDF_TOK_CLOSE_DICT)
goto skip;
if (tok == PDF_TOK_CLOSE_ARRAY)
continue;
fz_throw(ctx, "cannot make sense of broken array after all");
}
break;
case PDF_TOK_OPEN_DICT:
val = pdf_parse_dict(xref, file, buf);
break;
case PDF_TOK_NAME:
val = fz_new_name(ctx, buf->scratch);
break;
case PDF_TOK_REAL:
val = pdf_new_real(ctx, buf->f);
break;
case PDF_TOK_STRING:
val = pdf_new_string(ctx, buf->scratch, buf->len);
break;
case PDF_TOK_TRUE:
val = pdf_new_bool(ctx, 1);
break;
case PDF_TOK_FALSE:
val = pdf_new_bool(ctx, 0);
break;
case PDF_TOK_NULL:
val = pdf_new_null(ctx);
break;
case PDF_TOK_INT:
/* 64-bit to allow for numbers > INT_MAX and overflow */
a = buf->i;
tok = pdf_lex(file, buf);
if (tok == PDF_TOK_CLOSE_DICT || tok == PDF_TOK_NAME ||
(tok == PDF_TOK_KEYWORD && !strcmp(buf->scratch, "ID")))
{
val = pdf_new_int(ctx, a);
fz_dict_put(dict, key, val);
pdf_drop_obj(val);
val = NULL;
pdf_drop_obj(key);
key = NULL;
goto skip;
}
if (tok == PDF_TOK_INT)
{
b = buf->i;
tok = pdf_lex(file, buf);
if (tok == PDF_TOK_R)
{
val = pdf_new_indirect(ctx, a, b, xref);
break;
}
}
fz_throw(ctx, "invalid indirect reference in dict");
default:
fz_throw(ctx, "unknown token in dict");
}
fz_dict_put(dict, key, val);
pdf_drop_obj(val);
val = NULL;
pdf_drop_obj(key);
key = NULL;
}
}
fz_catch(ctx)
{
pdf_drop_obj(dict);
pdf_drop_obj(key);
pdf_drop_obj(val);
fz_throw(ctx, "cannot parse dict");
}
return dict;
}
static fz_error
parsecode(pdf_function *func, fz_stream *stream, int *codeptr)
{
fz_error error;
char buf[64];
int len;
pdf_token_e tok;
int opptr, elseptr;
int a, b, mid, cmp;
memset(buf, 0, sizeof(buf));
while (1)
{
error = pdf_lex(&tok, stream, buf, sizeof buf, &len);
if (error)
return fz_rethrow(error, "calculator function lexical error");
switch(tok)
{
case PDF_TEOF:
return fz_throw("truncated calculator function");
case PDF_TINT:
error = resizecode(func, *codeptr);
if (error)
return fz_rethrow(error, "resize calculator function code");
func->u.p.code[*codeptr].type = PSINT;
func->u.p.code[*codeptr].u.i = atoi(buf);
++*codeptr;
break;
case PDF_TREAL:
error = resizecode(func, *codeptr);
if (error)
return fz_rethrow(error, "resize calculator function code");
func->u.p.code[*codeptr].type = PSREAL;
func->u.p.code[*codeptr].u.f = atof(buf);
++*codeptr;
break;
case PDF_TOBRACE:
opptr = *codeptr;
*codeptr += 3;
error = resizecode(func, opptr + 2);
if (error)
return fz_rethrow(error, "resize calculator function code");
error = parsecode(func, stream, codeptr);
if (error)
return fz_rethrow(error, "error in 'if' branch");
error = pdf_lex(&tok, stream, buf, sizeof buf, &len);
if (error)
return fz_rethrow(error, "calculator function syntax error");
if (tok == PDF_TOBRACE)
{
elseptr = *codeptr;
error = parsecode(func, stream, codeptr);
if (error)
return fz_rethrow(error, "error in 'else' branch");
error = pdf_lex(&tok, stream, buf, sizeof buf, &len);
if (error)
return fz_rethrow(error, "calculator function syntax error");
}
else
{
elseptr = -1;
}
if (tok == PDF_TKEYWORD)
{
if (!strcmp(buf, "if"))
{
if (elseptr >= 0)
return fz_throw("too many branches for 'if'");
func->u.p.code[opptr].type = PSOPERATOR;
func->u.p.code[opptr].u.op = PSOIF;
func->u.p.code[opptr+2].type = PSBLOCK;
func->u.p.code[opptr+2].u.block = *codeptr;
}
else if (!strcmp(buf, "ifelse"))
{
if (elseptr < 0)
return fz_throw("not enough branches for 'ifelse'");
func->u.p.code[opptr].type = PSOPERATOR;
func->u.p.code[opptr].u.op = PSOIFELSE;
func->u.p.code[opptr+1].type = PSBLOCK;
func->u.p.code[opptr+1].u.block = elseptr;
func->u.p.code[opptr+2].type = PSBLOCK;
func->u.p.code[opptr+2].u.block = *codeptr;
}
else
{
return fz_throw("unknown keyword in 'if-else' context: '%s'", buf);
}
}
else
{
return fz_throw("missing keyword in 'if-else' context");
}
break;
case PDF_TCBRACE:
error = resizecode(func, *codeptr);
if (error)
return fz_rethrow(error, "resize calculator function code");
func->u.p.code[*codeptr].type = PSOPERATOR;
func->u.p.code[*codeptr].u.op = PSORETURN;
++*codeptr;
return fz_okay;
case PDF_TKEYWORD:
cmp = -1;
a = -1;
b = sizeof(psopnames) / sizeof(psopnames[0]);
while (b - a > 1)
{
mid = (a + b) / 2;
cmp = strcmp(buf, psopnames[mid]);
if (cmp > 0)
a = mid;
else if (cmp < 0)
b = mid;
else
a = b = mid;
}
if (cmp != 0)
return fz_throw("unknown operator: '%s'", buf);
error = resizecode(func, *codeptr);
if (error)
return fz_rethrow(error, "resize calculator function code");
func->u.p.code[*codeptr].type = PSOPERATOR;
func->u.p.code[*codeptr].u.op = a;
++*codeptr;
break;
default:
return fz_throw("calculator function syntax error");
}
}
}
pdf_obj *
pdf_parse_ind_obj(pdf_document *xref,
fz_stream *file, pdf_lexbuf *buf,
int *onum, int *ogen, int *ostmofs)
{
pdf_obj *obj = NULL;
int num = 0, gen = 0, stm_ofs;
int tok;
int a, b;
fz_context *ctx = file->ctx;
fz_var(obj);
tok = pdf_lex(file, buf);
/* RJW: cannot parse indirect object (%d %d R)", num, gen */
if (tok != PDF_TOK_INT)
fz_throw(ctx, "expected object number (%d %d R)", num, gen);
num = buf->i;
tok = pdf_lex(file, buf);
/* RJW: "cannot parse indirect object (%d %d R)", num, gen */
if (tok != PDF_TOK_INT)
fz_throw(ctx, "expected generation number (%d %d R)", num, gen);
gen = buf->i;
tok = pdf_lex(file, buf);
/* RJW: "cannot parse indirect object (%d %d R)", num, gen */
if (tok != PDF_TOK_OBJ)
fz_throw(ctx, "expected 'obj' keyword (%d %d R)", num, gen);
tok = pdf_lex(file, buf);
/* RJW: "cannot parse indirect object (%d %d R)", num, gen */
switch (tok)
{
case PDF_TOK_OPEN_ARRAY:
obj = pdf_parse_array(xref, file, buf);
/* RJW: "cannot parse indirect object (%d %d R)", num, gen */
break;
case PDF_TOK_OPEN_DICT:
obj = pdf_parse_dict(xref, file, buf);
/* RJW: "cannot parse indirect object (%d %d R)", num, gen */
break;
case PDF_TOK_NAME:
obj = fz_new_name(ctx, buf->scratch);
break;
case PDF_TOK_REAL:
obj = pdf_new_real(ctx, buf->f);
break;
case PDF_TOK_STRING:
obj = pdf_new_string(ctx, buf->scratch, buf->len);
break;
case PDF_TOK_TRUE:
obj = pdf_new_bool(ctx, 1);
break;
case PDF_TOK_FALSE:
obj = pdf_new_bool(ctx, 0);
break;
case PDF_TOK_NULL:
obj = pdf_new_null(ctx);
break;
case PDF_TOK_INT:
a = buf->i;
tok = pdf_lex(file, buf);
/* "cannot parse indirect object (%d %d R)", num, gen */
if (tok == PDF_TOK_STREAM || tok == PDF_TOK_ENDOBJ)
{
obj = pdf_new_int(ctx, a);
goto skip;
}
if (tok == PDF_TOK_INT)
{
b = buf->i;
tok = pdf_lex(file, buf);
/* RJW: "cannot parse indirect object (%d %d R)", num, gen); */
if (tok == PDF_TOK_R)
{
obj = pdf_new_indirect(ctx, a, b, xref);
break;
}
}
fz_throw(ctx, "expected 'R' keyword (%d %d R)", num, gen);
case PDF_TOK_ENDOBJ:
obj = pdf_new_null(ctx);
goto skip;
default:
fz_throw(ctx, "syntax error in object (%d %d R)", num, gen);
}
fz_try(ctx)
{
tok = pdf_lex(file, buf);
}
fz_catch(ctx)
{
pdf_drop_obj(obj);
fz_throw(ctx, "cannot parse indirect object (%d %d R)", num, gen);
}
skip:
if (tok == PDF_TOK_STREAM)
{
int c = fz_read_byte(file);
while (c == ' ')
c = fz_read_byte(file);
if (c == '\r')
{
c = fz_peek_byte(file);
if (c != '\n')
fz_warn(ctx, "line feed missing after stream begin marker (%d %d R)", num, gen);
else
fz_read_byte(file);
}
stm_ofs = fz_tell(file);
}
else if (tok == PDF_TOK_ENDOBJ)
{
stm_ofs = 0;
}
else
{
fz_warn(ctx, "expected 'endobj' or 'stream' keyword (%d %d R)", num, gen);
stm_ofs = 0;
}
if (onum) *onum = num;
if (ogen) *ogen = gen;
if (ostmofs) *ostmofs = stm_ofs;
return obj;
}
int
pdf_repair_obj(fz_context *ctx, pdf_document *doc, pdf_lexbuf *buf, fz_off_t *stmofsp, int *stmlenp, pdf_obj **encrypt, pdf_obj **id, pdf_obj **page, fz_off_t *tmpofs, pdf_obj **root)
{
fz_stream *file = doc->file;
pdf_token tok;
int stm_len;
*stmofsp = 0;
if (stmlenp)
*stmlenp = -1;
stm_len = 0;
/* On entry to this function, we know that we've just seen
* '<int> <int> obj'. We expect the next thing we see to be a
* pdf object. Regardless of the type of thing we meet next
* we only need to fully parse it if it is a dictionary. */
tok = pdf_lex(ctx, file, buf);
if (tok == PDF_TOK_OPEN_DICT)
{
pdf_obj *dict, *obj;
fz_try(ctx)
{
dict = pdf_parse_dict(ctx, doc, file, buf);
}
fz_catch(ctx)
{
fz_rethrow_if(ctx, FZ_ERROR_TRYLATER);
/* Don't let a broken object at EOF overwrite a good one */
if (file->eof)
fz_rethrow(ctx);
/* Silently swallow the error */
dict = pdf_new_dict(ctx, NULL, 2);
}
/* We must be careful not to try to resolve any indirections
* here. We have just read dict, so we know it to be a non
* indirected dictionary. Before we look at any values that
* we get back from looking up in it, we need to check they
* aren't indirected. */
if (encrypt || id || root)
{
obj = pdf_dict_get(ctx, dict, PDF_NAME_Type);
if (!pdf_is_indirect(ctx, obj) && pdf_name_eq(ctx, obj, PDF_NAME_XRef))
{
if (encrypt)
{
obj = pdf_dict_get(ctx, dict, PDF_NAME_Encrypt);
if (obj)
{
pdf_drop_obj(ctx, *encrypt);
*encrypt = pdf_keep_obj(ctx, obj);
}
}
if (id)
{
obj = pdf_dict_get(ctx, dict, PDF_NAME_ID);
if (obj)
{
pdf_drop_obj(ctx, *id);
*id = pdf_keep_obj(ctx, obj);
}
}
if (root)
*root = pdf_keep_obj(ctx, pdf_dict_get(ctx, dict, PDF_NAME_Root));
}
}
obj = pdf_dict_get(ctx, dict, PDF_NAME_Length);
if (!pdf_is_indirect(ctx, obj) && pdf_is_int(ctx, obj))
stm_len = pdf_to_int(ctx, obj);
if (doc->file_reading_linearly && page)
{
obj = pdf_dict_get(ctx, dict, PDF_NAME_Type);
if (!pdf_is_indirect(ctx, obj) && pdf_name_eq(ctx, obj, PDF_NAME_Page))
{
pdf_drop_obj(ctx, *page);
*page = pdf_keep_obj(ctx, dict);
}
}
pdf_drop_obj(ctx, dict);
}
