2022-08-16 12:59:32 +00:00
|
|
|
#include <qpdf/QPDFParser.hh>
|
|
|
|
|
|
|
|
#include <qpdf/QPDF.hh>
|
|
|
|
#include <qpdf/QPDFObjectHandle.hh>
|
|
|
|
#include <qpdf/QPDF_Array.hh>
|
|
|
|
#include <qpdf/QTC.hh>
|
|
|
|
#include <qpdf/QUtil.hh>
|
|
|
|
#include <qpdf/SparseOHArray.hh>
|
|
|
|
|
|
|
|
QPDFObjectHandle
|
|
|
|
QPDFParser::parse(bool& empty, bool content_stream)
|
|
|
|
{
|
|
|
|
// This method must take care not to resolve any objects. Don't
|
|
|
|
// check the type of any object without first ensuring that it is
|
|
|
|
// a direct object. Otherwise, doing so may have the side effect
|
|
|
|
// of reading the object and changing the file pointer. If you do
|
|
|
|
// this, it will cause a logic error to be thrown from
|
|
|
|
// QPDF::inParse().
|
|
|
|
|
|
|
|
QPDF::ParseGuard pg(context);
|
|
|
|
|
|
|
|
empty = false;
|
|
|
|
|
|
|
|
QPDFObjectHandle object;
|
|
|
|
bool set_offset = false;
|
|
|
|
|
|
|
|
std::vector<SparseOHArray> olist_stack;
|
|
|
|
olist_stack.push_back(SparseOHArray());
|
|
|
|
std::vector<parser_state_e> state_stack;
|
|
|
|
state_stack.push_back(st_top);
|
|
|
|
std::vector<qpdf_offset_t> offset_stack;
|
|
|
|
qpdf_offset_t offset = input->tell();
|
|
|
|
offset_stack.push_back(offset);
|
|
|
|
bool done = false;
|
|
|
|
int bad_count = 0;
|
|
|
|
int good_count = 0;
|
|
|
|
bool b_contents = false;
|
|
|
|
std::vector<std::string> contents_string_stack;
|
|
|
|
contents_string_stack.push_back("");
|
|
|
|
std::vector<qpdf_offset_t> contents_offset_stack;
|
|
|
|
contents_offset_stack.push_back(-1);
|
|
|
|
while (!done) {
|
|
|
|
bool bad = false;
|
|
|
|
SparseOHArray& olist = olist_stack.back();
|
|
|
|
parser_state_e state = state_stack.back();
|
|
|
|
offset = offset_stack.back();
|
|
|
|
std::string& contents_string = contents_string_stack.back();
|
|
|
|
qpdf_offset_t& contents_offset = contents_offset_stack.back();
|
|
|
|
|
|
|
|
object = QPDFObjectHandle();
|
|
|
|
set_offset = false;
|
|
|
|
|
|
|
|
QPDFTokenizer::Token token =
|
|
|
|
tokenizer.readToken(input, object_description, true);
|
|
|
|
std::string const& token_error_message = token.getErrorMessage();
|
|
|
|
if (!token_error_message.empty()) {
|
|
|
|
// Tokens other than tt_bad can still generate warnings.
|
2022-08-16 13:40:14 +00:00
|
|
|
warn(token_error_message);
|
2022-08-16 12:59:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
switch (token.getType()) {
|
|
|
|
case QPDFTokenizer::tt_eof:
|
|
|
|
if (!content_stream) {
|
|
|
|
QTC::TC("qpdf", "QPDFParser eof in parse");
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("unexpected EOF");
|
2022-08-16 12:59:32 +00:00
|
|
|
}
|
|
|
|
bad = true;
|
|
|
|
state = st_eof;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_bad:
|
|
|
|
QTC::TC("qpdf", "QPDFParser bad token in parse");
|
|
|
|
bad = true;
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_brace_open:
|
|
|
|
case QPDFTokenizer::tt_brace_close:
|
|
|
|
QTC::TC("qpdf", "QPDFParser bad brace");
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("treating unexpected brace token as null");
|
2022-08-16 12:59:32 +00:00
|
|
|
bad = true;
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_array_close:
|
|
|
|
if (state == st_array) {
|
|
|
|
state = st_stop;
|
|
|
|
} else {
|
|
|
|
QTC::TC("qpdf", "QPDFParser bad array close");
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("treating unexpected array close token as null");
|
2022-08-16 12:59:32 +00:00
|
|
|
bad = true;
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_dict_close:
|
|
|
|
if (state == st_dictionary) {
|
|
|
|
state = st_stop;
|
|
|
|
} else {
|
|
|
|
QTC::TC("qpdf", "QPDFParser bad dictionary close");
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("unexpected dictionary close token");
|
2022-08-16 12:59:32 +00:00
|
|
|
bad = true;
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_array_open:
|
|
|
|
case QPDFTokenizer::tt_dict_open:
|
|
|
|
if (olist_stack.size() > 500) {
|
|
|
|
QTC::TC("qpdf", "QPDFParser too deep");
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("ignoring excessively deeply nested data structure");
|
2022-08-16 12:59:32 +00:00
|
|
|
bad = true;
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
state = st_top;
|
|
|
|
} else {
|
|
|
|
olist_stack.push_back(SparseOHArray());
|
|
|
|
state = st_start;
|
|
|
|
offset_stack.push_back(input->tell());
|
|
|
|
state_stack.push_back(
|
|
|
|
(token.getType() == QPDFTokenizer::tt_array_open)
|
|
|
|
? st_array
|
|
|
|
: st_dictionary);
|
|
|
|
b_contents = false;
|
|
|
|
contents_string_stack.push_back("");
|
|
|
|
contents_offset_stack.push_back(-1);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_bool:
|
|
|
|
object = QPDFObjectHandle::newBool((token.getValue() == "true"));
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_null:
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_integer:
|
|
|
|
object = QPDFObjectHandle::newInteger(
|
|
|
|
QUtil::string_to_ll(token.getValue().c_str()));
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_real:
|
|
|
|
object = QPDFObjectHandle::newReal(token.getValue());
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_name:
|
|
|
|
{
|
|
|
|
std::string name = token.getValue();
|
|
|
|
object = QPDFObjectHandle::newName(name);
|
|
|
|
|
|
|
|
if (name == "/Contents") {
|
|
|
|
b_contents = true;
|
|
|
|
} else {
|
|
|
|
b_contents = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_word:
|
|
|
|
{
|
|
|
|
std::string const& value = token.getValue();
|
|
|
|
if (content_stream) {
|
|
|
|
object = QPDFObjectHandle::newOperator(value);
|
|
|
|
} else if (
|
|
|
|
(value == "R") && (state != st_top) &&
|
|
|
|
(olist.size() >= 2) &&
|
|
|
|
(!olist.at(olist.size() - 1).isIndirect()) &&
|
|
|
|
(olist.at(olist.size() - 1).isInteger()) &&
|
|
|
|
(!olist.at(olist.size() - 2).isIndirect()) &&
|
|
|
|
(olist.at(olist.size() - 2).isInteger())) {
|
|
|
|
if (context == nullptr) {
|
|
|
|
QTC::TC("qpdf", "QPDFParser indirect without context");
|
|
|
|
throw std::logic_error(
|
|
|
|
"QPDFObjectHandle::parse called without context"
|
|
|
|
" on an object with indirect references");
|
|
|
|
}
|
|
|
|
// Try to resolve indirect objects
|
|
|
|
object = QPDFObjectHandle::newIndirect(
|
|
|
|
context,
|
|
|
|
QPDFObjGen(
|
|
|
|
olist.at(olist.size() - 2).getIntValueAsInt(),
|
|
|
|
olist.at(olist.size() - 1).getIntValueAsInt()));
|
|
|
|
olist.remove_last();
|
|
|
|
olist.remove_last();
|
|
|
|
} else if ((value == "endobj") && (state == st_top)) {
|
|
|
|
// We just saw endobj without having read
|
|
|
|
// anything. Treat this as a null and do not move
|
|
|
|
// the input source's offset.
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
input->seek(input->getLastOffset(), SEEK_SET);
|
|
|
|
empty = true;
|
|
|
|
} else {
|
|
|
|
QTC::TC("qpdf", "QPDFParser treat word as string");
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("unknown token while reading object;"
|
|
|
|
" treating as string");
|
2022-08-16 12:59:32 +00:00
|
|
|
bad = true;
|
|
|
|
object = QPDFObjectHandle::newString(value);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case QPDFTokenizer::tt_string:
|
|
|
|
{
|
|
|
|
std::string val = token.getValue();
|
|
|
|
if (decrypter) {
|
|
|
|
if (b_contents) {
|
|
|
|
contents_string = val;
|
|
|
|
contents_offset = input->getLastOffset();
|
|
|
|
b_contents = false;
|
|
|
|
}
|
|
|
|
decrypter->decryptString(val);
|
|
|
|
}
|
|
|
|
object = QPDFObjectHandle::newString(val);
|
|
|
|
}
|
|
|
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("treating unknown token type as null while "
|
|
|
|
"reading object");
|
2022-08-16 12:59:32 +00:00
|
|
|
bad = true;
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((!object.isInitialized()) &&
|
|
|
|
(!((state == st_start) || (state == st_stop) ||
|
|
|
|
(state == st_eof)))) {
|
|
|
|
throw std::logic_error("QPDFObjectHandle::parseInternal: "
|
|
|
|
"unexpected uninitialized object");
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (bad) {
|
|
|
|
++bad_count;
|
|
|
|
good_count = 0;
|
|
|
|
} else {
|
|
|
|
++good_count;
|
|
|
|
if (good_count > 3) {
|
|
|
|
bad_count = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (bad_count > 5) {
|
|
|
|
// We had too many consecutive errors without enough
|
|
|
|
// intervening successful objects. Give up.
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("too many errors; giving up on reading object");
|
2022-08-16 12:59:32 +00:00
|
|
|
state = st_top;
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (state) {
|
|
|
|
case st_eof:
|
|
|
|
if (state_stack.size() > 1) {
|
2022-08-16 13:40:14 +00:00
|
|
|
warn("parse error while reading object");
|
2022-08-16 12:59:32 +00:00
|
|
|
}
|
|
|
|
done = true;
|
|
|
|
// In content stream mode, leave object uninitialized to
|
|
|
|
// indicate EOF
|
|
|
|
if (!content_stream) {
|
|
|
|
object = QPDFObjectHandle::newNull();
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case st_dictionary:
|
|
|
|
case st_array:
|
2022-08-30 05:42:46 +00:00
|
|
|
setDescriptionFromInput(object, input->getLastOffset());
|
2022-08-16 12:59:32 +00:00
|
|
|
object.setParsedOffset(input->getLastOffset());
|
|
|
|
set_offset = true;
|
|
|
|
olist.append(object);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case st_top:
|
|
|
|
done = true;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case st_start:
|
|
|
|
break;
|
|
|
|
|
|
|
|
case st_stop:
|
|
|
|
if ((state_stack.size() < 2) || (olist_stack.size() < 2)) {
|
|
|
|
throw std::logic_error(
|
|
|
|
"QPDFObjectHandle::parseInternal: st_stop encountered"
|
|
|
|
" with insufficient elements in stack");
|
|
|
|
}
|
|
|
|
parser_state_e old_state = state_stack.back();
|
|
|
|
state_stack.pop_back();
|
|
|
|
if (old_state == st_array) {
|
|
|
|
// There's no newArray(SparseOHArray) since
|
|
|
|
// SparseOHArray is not part of the public API.
|
|
|
|
object = QPDFObjectHandle(QPDF_Array::create(olist));
|
2022-08-30 05:42:46 +00:00
|
|
|
setDescriptionFromInput(object, offset);
|
2022-08-16 12:59:32 +00:00
|
|
|
// The `offset` points to the next of "[". Set the
|
|
|
|
// rewind offset to point to the beginning of "[".
|
|
|
|
// This has been explicitly tested with whitespace
|
|
|
|
// surrounding the array start delimiter.
|
|
|
|
// getLastOffset points to the array end token and
|
|
|
|
// therefore can't be used here.
|
|
|
|
object.setParsedOffset(offset - 1);
|
|
|
|
set_offset = true;
|
|
|
|
} else if (old_state == st_dictionary) {
|
|
|
|
// Convert list to map. Alternating elements are keys.
|
|
|
|
// Attempt to recover more or less gracefully from
|
|
|
|
// invalid dictionaries.
|
|
|
|
std::set<std::string> names;
|
|
|
|
size_t n_elements = olist.size();
|
|
|
|
for (size_t i = 0; i < n_elements; ++i) {
|
|
|
|
QPDFObjectHandle oh = olist.at(i);
|
|
|
|
if ((!oh.isIndirect()) && oh.isName()) {
|
|
|
|
names.insert(oh.getName());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
std::map<std::string, QPDFObjectHandle> dict;
|
|
|
|
int next_fake_key = 1;
|
|
|
|
for (unsigned int i = 0; i < olist.size(); ++i) {
|
|
|
|
QPDFObjectHandle key_obj = olist.at(i);
|
|
|
|
QPDFObjectHandle val;
|
|
|
|
if (key_obj.isIndirect() || (!key_obj.isName())) {
|
|
|
|
bool found_fake = false;
|
|
|
|
std::string candidate;
|
|
|
|
while (!found_fake) {
|
|
|
|
candidate = "/QPDFFake" +
|
|
|
|
QUtil::int_to_string(next_fake_key++);
|
|
|
|
found_fake = (names.count(candidate) == 0);
|
|
|
|
QTC::TC(
|
|
|
|
"qpdf",
|
|
|
|
"QPDFParser found fake",
|
|
|
|
(found_fake ? 0 : 1));
|
|
|
|
}
|
|
|
|
warn(
|
2022-08-16 13:40:14 +00:00
|
|
|
offset,
|
|
|
|
"expected dictionary key but found"
|
|
|
|
" non-name object; inserting key " +
|
|
|
|
candidate);
|
2022-08-16 12:59:32 +00:00
|
|
|
val = key_obj;
|
|
|
|
key_obj = QPDFObjectHandle::newName(candidate);
|
|
|
|
} else if (i + 1 >= olist.size()) {
|
|
|
|
QTC::TC("qpdf", "QPDFParser no val for last key");
|
|
|
|
warn(
|
2022-08-16 13:40:14 +00:00
|
|
|
offset,
|
|
|
|
"dictionary ended prematurely; "
|
|
|
|
"using null as value for last key");
|
2022-08-16 12:59:32 +00:00
|
|
|
val = QPDFObjectHandle::newNull();
|
2022-08-30 05:42:46 +00:00
|
|
|
setDescriptionFromInput(val, offset);
|
2022-08-16 12:59:32 +00:00
|
|
|
} else {
|
|
|
|
val = olist.at(++i);
|
|
|
|
}
|
|
|
|
std::string key = key_obj.getName();
|
|
|
|
if (dict.count(key) > 0) {
|
|
|
|
QTC::TC("qpdf", "QPDFParser duplicate dict key");
|
|
|
|
warn(
|
2022-08-16 13:40:14 +00:00
|
|
|
offset,
|
|
|
|
"dictionary has duplicated key " + key +
|
|
|
|
"; last occurrence overrides earlier "
|
|
|
|
"ones");
|
2022-08-16 12:59:32 +00:00
|
|
|
}
|
|
|
|
dict[key] = val;
|
|
|
|
}
|
|
|
|
if (!contents_string.empty() && dict.count("/Type") &&
|
|
|
|
dict["/Type"].isNameAndEquals("/Sig") &&
|
|
|
|
dict.count("/ByteRange") && dict.count("/Contents") &&
|
|
|
|
dict["/Contents"].isString()) {
|
|
|
|
dict["/Contents"] =
|
|
|
|
QPDFObjectHandle::newString(contents_string);
|
|
|
|
dict["/Contents"].setParsedOffset(contents_offset);
|
|
|
|
}
|
|
|
|
object = QPDFObjectHandle::newDictionary(dict);
|
2022-08-30 05:42:46 +00:00
|
|
|
setDescriptionFromInput(object, offset);
|
2022-08-16 12:59:32 +00:00
|
|
|
// The `offset` points to the next of "<<". Set the
|
|
|
|
// rewind offset to point to the beginning of "<<".
|
|
|
|
// This has been explicitly tested with whitespace
|
|
|
|
// surrounding the dictionary start delimiter.
|
|
|
|
// getLastOffset points to the dictionary end token
|
|
|
|
// and therefore can't be used here.
|
|
|
|
object.setParsedOffset(offset - 2);
|
|
|
|
set_offset = true;
|
|
|
|
}
|
|
|
|
olist_stack.pop_back();
|
|
|
|
offset_stack.pop_back();
|
|
|
|
if (state_stack.back() == st_top) {
|
|
|
|
done = true;
|
|
|
|
} else {
|
|
|
|
olist_stack.back().append(object);
|
|
|
|
}
|
|
|
|
contents_string_stack.pop_back();
|
|
|
|
contents_offset_stack.pop_back();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!set_offset) {
|
2022-08-30 05:42:46 +00:00
|
|
|
setDescriptionFromInput(object, offset);
|
2022-08-16 12:59:32 +00:00
|
|
|
object.setParsedOffset(offset);
|
|
|
|
}
|
|
|
|
return object;
|
|
|
|
}
|
|
|
|
|
2022-08-30 05:42:46 +00:00
|
|
|
void
|
|
|
|
QPDFParser::setDescriptionFromInput(
|
|
|
|
QPDFObjectHandle oh, qpdf_offset_t offset) const
|
|
|
|
{
|
|
|
|
oh.setObjectDescription(
|
|
|
|
context,
|
|
|
|
(input->getName() + ", " + object_description + " at offset " +
|
|
|
|
QUtil::int_to_string(offset)));
|
|
|
|
}
|
|
|
|
|
2022-08-16 12:59:32 +00:00
|
|
|
void
|
|
|
|
QPDFParser::warn(QPDF* qpdf, QPDFExc const& e)
|
|
|
|
{
|
|
|
|
// If parsing on behalf of a QPDF object and want to give a
|
|
|
|
// warning, we can warn through the object. If parsing for some
|
|
|
|
// other reason, such as an explicit creation of an object from a
|
|
|
|
// string, then just throw the exception.
|
|
|
|
if (qpdf) {
|
|
|
|
qpdf->warn(e);
|
|
|
|
} else {
|
|
|
|
throw e;
|
|
|
|
}
|
|
|
|
}
|
2022-08-16 13:40:14 +00:00
|
|
|
|
|
|
|
void
|
|
|
|
QPDFParser::warn(qpdf_offset_t offset, std::string const& msg) const
|
|
|
|
{
|
|
|
|
warn(
|
|
|
|
context,
|
|
|
|
QPDFExc(
|
|
|
|
qpdf_e_damaged_pdf,
|
|
|
|
input->getName(),
|
|
|
|
object_description,
|
|
|
|
offset,
|
|
|
|
msg));
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
QPDFParser::warn(std::string const& msg) const
|
|
|
|
{
|
|
|
|
warn(input->getLastOffset(), msg);
|
|
|
|
}
|