diff --git a/pdf.c b/pdf.c
index 3ec9e81fdf8c554019b3cc02fc3bb61531a0a34e..fe155c14f2fb0985436f531439af76707f0c498e 100644
--- a/pdf.c
+++ b/pdf.c
@@ -1,9 +1,11 @@
/* beginnings of a PDF parser in hammer
* pesco 2019
* pompolic 2020
+ * Paul Vines 2020
*/
#include <string.h> /* strncmp(), memset() */
+#include <stdlib.h> /* exit() */
#include <hammer/hammer.h>
#include <hammer/glue.h>
@@ -18,6 +20,13 @@
#define IN(STR) h_in((const uint8_t *)(STR), sizeof(STR) - 1)
#define NOT_IN(STR) h_not_in((const uint8_t *)(STR), sizeof(STR) - 1)
+#ifdef LOG
+#define VIOL(P,VIOL) h_action(h_sequence(P, h_tell(), NULL), act_viol, VIOL)
+#else
+#define VIOL(P,VIOL) P
+#endif
+
+
/*
* some helpers
@@ -27,6 +36,7 @@ HParser *p_fail;
HParser *p_epsilon;
HParser *p_return_0;
HParser *p_return_1;
+uint8_t strictness = 0;
/* a combinator to parse a given character but return a different value */
@@ -411,6 +421,31 @@ act_nat(const HParseResult *p, void *u)
#define act_xroff act_nat
#define act_xrgen act_nat
+HParser *p_violsev;
+HParsedToken *
+act_viol(const HParseResult *p, void *viol)
+{
+ uint severity;
+ uint offset;
+ HParseResult *severity_parse;
+ viol = (uint8_t *) viol;
+ severity_parse = h_parse(p_violsev, viol, strlen((char *)viol));
+ if (!severity_parse) {
+ fprintf(stderr, "Severity for violaiton %s not assigned!\n", (char *)viol);
+ severity = 99999;
+ }
+ else {
+ severity = severity_parse->ast->seq->elements[0]->uint;
+ }
+ offset = p->ast->seq->elements[1]->uint / 8;
+ fprintf(stderr, "VIOLATION[%d]@%d (0x%x): %s\n", severity, offset, offset, (char *) viol);
+ if (strictness && severity > strictness) {
+ exit(1);
+ }
+ /* Just return the parse AST, drop the h_tell */
+ return (HParsedToken *) p->ast->seq->elements[0];
+}
+
bool
validate_pnat(HParseResult *p, void *u)
{
@@ -856,8 +891,9 @@ init_parser(struct Env *aux)
/* dictionaries */
H_RULE(dopen, LIT("<<"));
H_RULE(dclose, LIT(">>"));
- H_RULE(k_v, CHX(SEQ(name, wel,ws, obj),
- SEQ(name, CHX(name,dobj))));
+ H_RULE(k_v, CHX(CHX(SEQ(name, wel,ws, obj),
+ SEQ(name, CHX(name,dobj))),
+ VIOL(SEQ(name, wel,ws), "Key with no value (severity=2)")));
H_ARULE(dict_, h_middle(dopen, MANY_WS(k_v), dclose));
// XXX this allows, for instance, "<<<<" to be parsed as "<< <<". ok?
// XXX validate: dict keys must be unique
@@ -880,8 +916,24 @@ init_parser(struct Env *aux)
h_bind_indirect(array, array_);
/* streams */
- H_RULE(stmbeg, SEQ(dict, ws, LIT("stream"), OPT(cr), lf));
- H_RULE(stmend, SEQ(OPT(eol), LIT("endstream")));
+ H_RULE(stmbeg, SEQ(dict, OPT(ws), LIT("stream"), OPT(cr), lf));
+ H_RULE(stmend, CHX(SEQ(eol, LIT("endstream")),
+ VIOL(LIT("ndstream"), "Stream length >1-too-long (severity=10)"),
+ VIOL(SEQ(h_many1(wchar), LIT("endstream")),
+ "No newline before endstream (severity=7)"),
+ VIOL(LIT("endstream"), "Stream length 1-too-long (severity=9)"),
+ VIOL(SEQ(OPT(h_ch_range(0, 255)), OPT(eol), LIT("endstream")),
+ "Stream length 1-too-short (severity=4)"),
+ VIOL(SEQ(h_many1(h_butnot(h_ch_range(0, 255), CHX(KW("endobj"),
+ SEQ(npair, wel, KW("obj")),
+ KW("xref"),
+ LIT("endstream")))), LIT("endstream")),
+ "Stream length >1-too-short (severity=5)"),
+ VIOL(h_many1(h_butnot(h_ch_range(0, 255), CHX(KW("endobj"),
+ SEQ(npair, wel, KW("obj")),
+ KW("xref")))),
+ "Missing endstream token (severity=7)")));
+
H_RULE(stream, h_left(h_bind(stmbeg, kstream, aux), stmend));
// XXX is whitespace allowed between the eol and "endstream"?
@@ -890,12 +942,18 @@ init_parser(struct Env *aux)
*/
/* header */
- H_RULE(version, SEQ(pdigit, IGN(period), pdigit));
+ H_RULE(version, SEQ(pdigit, IGN(period), digit));
H_RULE(header, h_middle(LIT("%PDF-"), version, nl));
/* body */
H_RULE(indobj, CHX(stream, obj));
- H_RULE(objdef, SEQ(ws, npair, wel, KW("obj"), ws, indobj, KW("endobj")));
+ H_RULE(objdef, SEQ(ws, npair, wel, KW("obj"), ws, indobj,
+ CHX(VIOL(SEQ(OPT(ws), OPT(lws), KW("endobj"), h_many(CHX(wel, eol)), h_many1(KW("endobj"))),
+ "More than 1 endobj token (severity=1)"),
+ VIOL(SEQ(OPT(ws), OPT(lws), KW("endobj"), h_many(CHX(wel, eol)), h_many1(SEQ(dclose, h_many1(CHX(wchar, eol)), KW("endobj")))),
+ "More than 1 >> and endobj token (severity=2)"),
+ SEQ(OPT(ws), OPT(lws), KW("endobj")),
+ VIOL(h_optional(KW("endobj")), "Missing endobj token (severity=1)"))));
H_RULE(body, h_many(objdef));
/* for object streams */
@@ -918,26 +976,44 @@ init_parser(struct Env *aux)
// XXX skip however much we consumed and check for "endstream endobj"?
/* trailer */
- H_RULE(startxr, SEQ(nl, KW("startxref"), nl,
+ H_RULE(startxr, SEQ(nl, KW("startxref"), nl,
+ lws, nat, nl,
+ LIT("%%EOF"), OPT(nl)));
+
+ /* used for the backwards search */
+ H_RULE(lasteof, SEQ(nl, KW("startxref"), nl,
lws, nat, nl,
- LIT("%%EOF"), CHX(nl, end)));
// XXX the real world sometimes omits nl after %%EOF inside the file.
// the next 'tail' would be appended right after the 'F',
// presumably because the previous version of the file
// ended without a trailing newline. m)
- // this is invalid per spec, because it creates a run-on
+ // this is invalid per spec, because it creates a run-on
// comment, but we should probably accept-and-warn.
// XXX should lws be allowed before EOF marker?
// NB: lws before xref offset is allowed, cf. p.48 (example 4)
+ LIT("%%EOF"),
+ CHX(VIOL(SEQ(nl, h_many1(nl), end),
+ "(offset FROM END) Multiple newlines after final %%EOF (severity=4)"),
+ SEQ(h_many(nl), end),
+ VIOL(SEQ(h_butnot(h_ch_range(0, 255), LIT("%%EOF"))),
+ "(offset FROM END) Data after final %%EOF (severity=7)"))));
+
H_RULE(xr_td, SEQ(xrefs, KW("trailer"), ws, dict));
- H_RULE(tail, SEQ(body, h_optional(xr_td), startxr));
- // XXX the real world likes to omit 'startxr' from all but the
- // last trailer. we should accept-and-warn in that case.
- H_RULE(pdf, SEQ(header, h_many1(tail), end));
+ H_RULE(hdr_junk, VIOL(h_many1(h_butnot(h_ch_range(0, 255), objdef)),
+ "Uncommented junk after header (severity=1)"));
+ H_RULE(tail, SEQ(body, CHX(SEQ(h_optional(xr_td), startxr),
+ VIOL(SEQ(xr_td, OPT(SEQ(nl, KW("startxref"), nl, lws, nat, nl)),
+ OPT(nl), OPT(LIT("%%EOF")), OPT(nl)),
+ "Improper end of trailer - missing startxref and/or %%EOF (severity=5)"))));
+ H_RULE(final_eof_junk, CHX(VIOL(SEQ(h_many1(nl), end), "Multiple newlines after final %%EOF (severity=4)"),
+ VIOL(h_many1(h_butnot(h_ch_range(0, 255), LIT("%%EOF"))),
+ "Data after final %%EOF (severity=7)"),
+ end));
+ H_RULE(pdf, SEQ(header, OPT(hdr_junk), h_many1(tail), final_eof_junk));
/* debug parser to consume as much as possible */
- H_RULE(pdfdbg, SEQ(header, h_many(tail), body, OPT(xr_td), OPT(startxr)));
+ H_RULE(pdfdbg, SEQ(header, OPT(hdr_junk), h_many(tail), body, OPT(xr_td), OPT(SEQ(startxr, final_eof_junk))));
/*
* filters
@@ -979,7 +1055,7 @@ init_parser(struct Env *aux)
/* global parser variables */
p_pdf = pdf;
p_pdfdbg = pdfdbg;
- p_startxref = startxr;
+ p_startxref = lasteof; //startxr;
p_xref = CHX(xr_td, xrstm);
p_objdef = objdef;
p_a85string = a85string;
@@ -990,6 +1066,12 @@ init_parser(struct Env *aux)
p_return_0 = h_action(epsilon, act_return_uint, (void *)0);
p_return_1 = h_action(epsilon, act_return_uint, (void *)1);
+ /* Parsing of severity messages */
+ H_RULE(viol_preamble, SEQ(h_many(NOT_IN("=")), LIT("=")));
+ H_RULE(severity_num, h_action(h_many1(h_action(h_ch_range('0', '9'), act_digit, NULL)),
+ act_nat, NULL));
+ H_RULE(violsev, SEQ(IGN(viol_preamble), severity_num));
+ p_violsev = violsev;
#if 0
// XXX testing
int r;
@@ -1563,7 +1645,7 @@ kstream(HAllocator *mm__, const HParsedToken *x, void *env)
//fprintf(stderr, "parsing stream object, length %zu.\n", sz); // XXX debug
- dict_p = p_return__m(mm__, dict_t);
+ dict_p = p_return__m(mm__, dict_t);
bytes_p = p_take__m(mm__, sz, aux);
spec = h_alloc(mm__, sizeof(struct streamspec));
@@ -1635,9 +1717,9 @@ p_xrefdata__m(HAllocator *mm__, const Dict *dict)
* for field 3 (generation). in fact, these are the only defaults
* defined by ISO 32000-1:2008 (PDF 1.7).
*
- * entry type field no. default value
- * 1 (type) 1
- * 1 ("n") 3 (gen.) 0
+ * entry type field no. default value
+ * 1 (type) 1
+ * 1 ("n") 3 (gen.) 0
*/
/* Size (required) - total size of xref table */
@@ -1680,7 +1762,7 @@ p_xrefdata__m(HAllocator *mm__, const Dict *dict)
*
* [t x y] with t,x,y > 0 full general form
* [0 x y] with x,y > 0 only type-1 ("in use") entries
- * [0 x 0] with x > 0 only type-1 entries, only offsets
+ * [0 x 0] with x > 0 only type-1 entries, only offsets
*
* however, though nonsensical, [t x 0] with t,x > 0 is not disallowed
* by the spec; as long as all entries are of type 1, the xref data can
@@ -1907,10 +1989,15 @@ main(int argc, char *argv[])
int fd;
/* command line handling */
- if (argc != 2) {
+ if (argc > 3) {
fprintf(stderr, "usage: %s file\n", argv[0]);
return 1;
}
+ if (argc == 3) {
+ H_RULE(nat, h_action(h_many1(h_action(h_ch_range('0', '9'), act_digit, NULL)),
+ act_nat, NULL));
+ strictness = h_parse(nat, (uint8_t *)argv[2], strlen(argv[2]))->ast->uint;
+ }
infile = argv[1];
/* mmap the input file */