diff --git a/pdf.c b/pdf.c
index 44dcb83138f2080480193d43eac6de5ed6e70516..1e55c3fcdc5d5fd8440bfd9084ab6353777e970f 100644
--- a/pdf.c
+++ b/pdf.c
@@ -436,7 +436,7 @@ HParsedToken *
act_a85partial2group(const HParseResult *p, void *u)
{
// only a single character is necessary
- // XXX: use H_MAKE?
+ // XXX: use H_ALLOC?
uint8_t *bytes = h_arena_malloc(p->arena, sizeof(uint8_t));
uint32_t fourbytes = 0;
@@ -450,12 +450,13 @@ act_a85partial2group(const HParseResult *p, void *u)
fprintf(stdout, "act_a85partial2group: i = %d, digit = %2lx, fourbytes = %4x (%d)\n", i, H_CAST_UINT(digits[i]->seq->elements[0]),
fourbytes, fourbytes); // XXX DEBUG
}
+
// fill the other bytes with u (dec 117-33)
+ // XXX I'm confused. I think to restore the padding, '!' needs to be added instead of 'u'
for (int i=2; i<5; i++) {
fourbytes = fourbytes * 85 + 84;
}
-
// truncate and return only the high order byte
bytes[0] = (fourbytes & 0xFF000000) >> 24;
fprintf(stdout, "act_a85partial2group: %4x (%d) ==> %2x \n", fourbytes, fourbytes, bytes[0]); // XXX DEBUG
@@ -482,7 +483,7 @@ act_a85partial3group(const HParseResult *p, void *u)
uint8_t *bytes = h_arena_malloc(p->arena, sizeof(uint8_t) * 2);
for (int i=0; i<4; i++) {
- assert(digits[i]->token_type == TT_SEQUENCE);
+ assert(digits[i]->token_type == TT_SEQUENCE); // XXX assert fails on test input A85(Hello, World!a) -> 87cURD_*#4DfTZ)+^k~>
fourbytes = fourbytes * 85 + H_CAST_UINT(digits[i]->seq->elements[0]);
}
@@ -492,12 +493,13 @@ act_a85partial3group(const HParseResult *p, void *u)
}
// truncate and return only the high order bytes
+ // I think groups of 3 A85 characters decode to 2 bytes
bytes[0] = (fourbytes & 0xFF000000) >> 24;
bytes[1] = (fourbytes & 0x00FF0000) >> 16;
bytes[2] = (fourbytes & 0x0000FF00) >> 8;
- fprintf(stdout, "act_a85partial2group: %4x (%d) ==> %2x %2x %2x \n", fourbytes, fourbytes, bytes[0],
- bytes[1], bytes[2]);
+ fprintf(stdout, "act_a85partial3group: %4x (%d) ==> %2x %2x %2x \n", fourbytes, fourbytes, bytes[0],
+ bytes[1], bytes[2]); // XXX DEBUG
return H_MAKE_BYTES(bytes, 3); // XXX test: should this be 2?
}
@@ -527,8 +529,9 @@ act_a85partial4group(const HParseResult *p, void *u)
// truncate and return only the high order bytes
bytes[0] = (fourbytes & 0xFF000000) >> 24;
bytes[1] = (fourbytes & 0x00FF0000) >> 16;
+ bytes[2] = (fourbytes & 0x0000FF00) >> 8;
- fprintf(stdout, "act_a85partial2group: %4x (%d) ==> %2x %2x \n", fourbytes, fourbytes, bytes[0], bytes[1]);
+ fprintf(stdout, "act_a85partial4group: %4x (%d) ==> %2x %2x \n", fourbytes, fourbytes, bytes[0], bytes[1], bytes[2]); // XXX debug
return H_MAKE_BYTES(bytes, 3);
}
@@ -1253,8 +1256,6 @@ init_parser(struct Env *aux)
*/
/* Whitespace can occur between any digit and has to be ignored, */
- /* Comments are not allowed inside streams, and % character should cause
- * a parse error. */
H_RULE(aws, IGN(h_many(wchar))); // all white space, include CR & LF, but not comments
#define MANY_AWS(X) h_many(CHX(aws, X))
@@ -1265,8 +1266,6 @@ init_parser(struct Env *aux)
H_ARULE(a85digit, h_ch_range('!', 'u'));
/* Line whitespace can occur between any digit and has to be ignored, */
- /* Comments are not allowed inside streams, and % character should cause
- * a parse error. */
#define MANY_LWS(X) h_many(CHX(lws, X))
/* This encoding of zero is not allowed */
// H_RULE(a85fiveexcl, h_repeat_n(SEQ(h_ch('!'), aws), 5)); // seeing this is a violation
@@ -1291,7 +1290,16 @@ init_parser(struct Env *aux)
* Not sure whether comments can be embedded within content streams
* If not, use the rule aws rather than ws
*/
- // XXX TODO: probably not, but check the standard again
+ /*
+ * It seems somewhat unclear. ASCII85Decode definitely can't have
+ * comments, because % can be part of a valid ASCII85Encoded character.
+ * However, it seems that comments are generally allowed:
+ * "Any occurrence of the PERCENT SIGN (25h) outside a string or
+ * inside a content stream (see 7.8.2, "Content streams") introduces
+ * a comment." ISO32000:2-2017 7.2.4
+ */
+ // XXX Ask Peter Wyatt
+
/* AsciiHexDecode */
H_RULE(ahexeod, h_ch('>'));
@@ -2099,7 +2107,7 @@ p_take__m(HAllocator *mm__, size_t n, struct Env *aux)
HParser *p_xrefdata__m(HAllocator *, const Dict *);
HParser *p_objstm__m(HAllocator *, const Dict *);
-
+HParser *p_raw_test__m(HAllocator *, const Dict *);
/*
* Look into the dictionary associated with the stream to see if there is data
@@ -2112,7 +2120,8 @@ p_stream_data__m(HAllocator *mm__, const Dict *dict)
v = dictentry(dict, "Type");
if (v == NULL || v->token_type != TT_BYTES) // XXX -> custom type
- return NULL; /* no /Type field */
+ return p_raw_test__m(mm__, dict); /* no /Type field */
+ // XXX restore return NULL above
/* interpret known stream types */
if (bytes_eq(v->bytes, "XRef"))
@@ -2401,6 +2410,25 @@ p_objstm__m(HAllocator *mm__, const Dict *dict)
// XXX consistency-check against /First, idx, /N
}
+HParser*
+
+/* Debug parser to test encoded text streams */
+// XXX remove
+
+p_raw_test__m(HAllocator *mm__, const Dict *dict)
+{
+ /*const HParsedToken *v;
+ size_t N;
+
+ v = dictentry(dict, "Length");
+ if (v == NULL || v->token_type != TT_SINT || v->sint < 0 ||
+ (uint64_t)v->sint > SIZE_MAX) {
+ fprintf(stderr, "missing /Length on object stream\n");
+ return p_fail;
+ }*/
+ return h_sequence__m(mm__, h_many__m(mm__, h_uint8__m(mm__)), NULL);
+}
+
/*
* This continuation is very similar to kstream, except that it does not
* rely on /Length to consume the right amount of input. If /Length is