aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--.clang-format8
-rw-r--r--.gitignore8
-rw-r--r--LICENSE2
-rw-r--r--README.adoc35
-rw-r--r--cmd/extfs-pdf/main.go141
-rw-r--r--go.mod6
-rw-r--r--go.sum8
-rw-r--r--meson.build4
-rw-r--r--pdf-simple-sign.adoc2
-rw-r--r--pdf-simple-sign.cpp98
-rw-r--r--pdf/pdf.go554
-rwxr-xr-xtest.sh22
12 files changed, 747 insertions, 141 deletions
diff --git a/.clang-format b/.clang-format
new file mode 100644
index 0000000..13cbee9
--- /dev/null
+++ b/.clang-format
@@ -0,0 +1,8 @@
+BasedOnStyle: Chromium
+ColumnLimit: 100
+IndentCaseLabels: false
+AccessModifierOffset: -2
+ContinuationIndentWidth: 2
+SpaceAfterTemplateKeyword: false
+SpaceAfterCStyleCast: true
+SpacesBeforeTrailingComments: 2
diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..d046c48
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,8 @@
+/builddir
+/pdf-simple-sign.cflags
+/pdf-simple-sign.config
+/pdf-simple-sign.creator
+/pdf-simple-sign.creator.user
+/pdf-simple-sign.cxxflags
+/pdf-simple-sign.files
+/pdf-simple-sign.includes
diff --git a/LICENSE b/LICENSE
index 3bfb256..7511f3e 100644
--- a/LICENSE
+++ b/LICENSE
@@ -1,4 +1,4 @@
-Copyright (c) 2017 - 2020, Přemysl Eric Janouch <p@janouch.name>
+Copyright (c) 2017 - 2024, Přemysl Eric Janouch <p@janouch.name>
Permission to use, copy, modify, and/or distribute this software for any
purpose with or without fee is hereby granted.
diff --git a/README.adoc b/README.adoc
index 56eecb8..10e581f 100644
--- a/README.adoc
+++ b/README.adoc
@@ -2,16 +2,26 @@ pdf-simple-sign
===============
'pdf-simple-sign' is a simple PDF signer intended for documents produced by
-the Cairo library, GNU troff, ImageMagick, or similar.
+the Cairo library (≤ 1.17.4 or using PDF 1.4), GNU troff, ImageMagick,
+or similar.
I don't aim to extend the functionality any further. The project is fairly
self-contained and it should be easy to grasp and change to suit to your needs.
+Packages
+--------
+Regular releases are sporadic. git master should be stable enough.
+You can get a package with the latest development version using Arch Linux's
+https://aur.archlinux.org/packages/pdf-simple-sign-git[AUR],
+or as a https://git.janouch.name/p/nixexprs[Nix derivation].
+
Documentation
-------------
See the link:pdf-simple-sign.adoc[man page] for information about usage.
The rest of this README will concern itself with externalities.
+image:https://pkg.go.dev/badge/janouch.name/pdf-simple-sign@master/pdf["PkgGoDev", link="https://pkg.go.dev/janouch.name/pdf-simple-sign@master/pdf"]
+
Building
--------
Build dependencies: Meson, Asciidoctor, a C++11 compiler, pkg-config +
@@ -23,9 +33,28 @@ Runtime dependencies: libcrypto (OpenSSL 1.1 API)
$ cd builddir
$ ninja
-In addition to the C++ version, also included is a native Go port:
+In addition to the C++ version, also included is a native Go port,
+which has enhanced PDF 1.5 support:
+
+----
+$ go install janouch.name/pdf-simple-sign/cmd/pdf-simple-sign@master
+----
+
+and a crude external VFS for Midnight Commander, that may be used to extract
+all streams from a given PDF file:
+
+----
+$ GOBIN=$HOME/.local/share/mc/extfs.d \
+ go install janouch.name/pdf-simple-sign/cmd/extfs-pdf@master
+----
+
+To enable the VFS, edit your _~/.config/mc/mc.ext.ini_ to contain:
- $ go get janouch.name/pdf-simple-sign/cmd/pdf-simple-sign
+----
+[pdf]
+Type=^PDF
+Open=%cd %p/extfs-pdf://
+----
Contributing and Support
------------------------
diff --git a/cmd/extfs-pdf/main.go b/cmd/extfs-pdf/main.go
new file mode 100644
index 0000000..eab3e2b
--- /dev/null
+++ b/cmd/extfs-pdf/main.go
@@ -0,0 +1,141 @@
+//
+// Copyright (c) 2021 - 2024, Přemysl Eric Janouch <p@janouch.name>
+//
+// Permission to use, copy, modify, and/or distribute this software for any
+// purpose with or without fee is hereby granted.
+//
+// THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+// WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+// MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+// SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+// WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+// OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+// CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+//
+
+// extfs-pdf is an external VFS plugin for Midnight Commander.
+// More serious image extractors should rewrite this to use pdfimages(1).
+package main
+
+import (
+ "flag"
+ "fmt"
+ "os"
+ "time"
+
+ "janouch.name/pdf-simple-sign/pdf"
+)
+
+func die(status int, format string, args ...interface{}) {
+ os.Stderr.WriteString(fmt.Sprintf(format+"\n", args...))
+ os.Exit(status)
+}
+
+func usage() {
+ die(1, "Usage: %s [-h] COMMAND DOCUMENT [ARG...]", os.Args[0])
+}
+
+func streamSuffix(o *pdf.Object) string {
+ if filter, _ := o.Dict["Filter"]; filter.Kind == pdf.Name {
+ switch filter.String {
+ case "JBIG2Decode":
+ // This is the file extension used by pdfimages(1).
+ // This is not a complete JBIG2 standalone file.
+ return "jb2e"
+ case "JPXDecode":
+ return "jp2"
+ case "DCTDecode":
+ return "jpg"
+ case "FlateDecode":
+ return "zz"
+ default:
+ return filter.String
+ }
+ }
+ return "stream"
+}
+
+func list(mtime time.Time, updater *pdf.Updater) {
+ stamp := mtime.Local().Format("01-02-2006 15:04:05")
+ for _, o := range updater.ListIndirect() {
+ object, err := updater.Get(o.N, o.Generation)
+ size := 0
+ if err != nil {
+ fmt.Fprintf(os.Stderr, "%s\n", err)
+ } else {
+ // Accidental transformation, retrieving original data is more work.
+ size = len(object.Serialize())
+ }
+ fmt.Printf("-r--r--r-- 1 0 0 %d %s n%dg%d\n",
+ size, stamp, o.N, o.Generation)
+ if object.Kind == pdf.Stream {
+ fmt.Printf("-r--r--r-- 1 0 0 %d %s n%dg%d.%s\n", len(object.Stream),
+ stamp, o.N, o.Generation, streamSuffix(&object))
+ }
+ }
+}
+
+func copyout(updater *pdf.Updater, storedFilename, extractTo string) {
+ var (
+ n, generation uint
+ suffix string
+ )
+ m, err := fmt.Sscanf(storedFilename, "n%dg%d%s", &n, &generation, &suffix)
+ if m < 2 {
+ die(3, "%s: %s", storedFilename, err)
+ }
+
+ object, err := updater.Get(n, generation)
+ if err != nil {
+ die(3, "%s: %s", storedFilename, err)
+ }
+
+ content := []byte(object.Serialize())
+ if suffix != "" {
+ content = object.Stream
+ }
+ if err = os.WriteFile(extractTo, content, 0666); err != nil {
+ die(3, "%s", err)
+ }
+}
+
+func main() {
+ flag.Usage = usage
+ flag.Parse()
+ if flag.NArg() < 2 {
+ usage()
+ }
+
+ command, documentPath := flag.Arg(0), flag.Arg(1)
+ doc, err := os.ReadFile(documentPath)
+ if err != nil {
+ die(1, "%s", err)
+ }
+
+ mtime := time.UnixMilli(0)
+ if info, err := os.Stat(documentPath); err == nil {
+ mtime = info.ModTime()
+ }
+
+ updater, err := pdf.NewUpdater(doc)
+ if err != nil {
+ die(2, "%s", err)
+ }
+
+ switch command {
+ default:
+ die(1, "unsupported command: %s", command)
+ case "list":
+ if flag.NArg() != 2 {
+ usage()
+ } else {
+ list(mtime, updater)
+ }
+ case "copyout":
+ if flag.NArg() != 4 {
+ usage()
+ } else {
+ copyout(updater, flag.Arg(2), flag.Arg(3))
+ }
+ }
+}
diff --git a/go.mod b/go.mod
index 7ee935f..0e84ffc 100644
--- a/go.mod
+++ b/go.mod
@@ -1,8 +1,8 @@
module janouch.name/pdf-simple-sign
-go 1.14
+go 1.17
require (
- go.mozilla.org/pkcs7 v0.0.0-20200128120323-432b2356ecb1
- golang.org/x/crypto v0.0.0-20200728195943-123391ffb6de
+ go.mozilla.org/pkcs7 v0.0.0-20210826202110-33d05740a352
+ golang.org/x/crypto v0.10.0
)
diff --git a/go.sum b/go.sum
index 03d7a29..4cf11b0 100644
--- a/go.sum
+++ b/go.sum
@@ -1,12 +1,12 @@
-go.mozilla.org/pkcs7 v0.0.0-20181029144607-24857c352dd8 h1:W3oGFPlHBLgXdsbPVixWFMYsuPhm81/Qww3XAgBbn/0=
-go.mozilla.org/pkcs7 v0.0.0-20181029144607-24857c352dd8/go.mod h1:5fWP3IVYEMc04wC+lMJAfkmNmKAl2P1swVv8VS+URZ8=
go.mozilla.org/pkcs7 v0.0.0-20200128120323-432b2356ecb1 h1:A/5uWzF44DlIgdm/PQFwfMkW0JX+cIcQi/SwLAmZP5M=
go.mozilla.org/pkcs7 v0.0.0-20200128120323-432b2356ecb1/go.mod h1:SNgMg+EgDFwmvSmLRTNKC5fegJjB7v23qTQ0XLGUNHk=
-golang.org/x/crypto v0.0.0-20181127143415-eb0de9b17e85 h1:et7+NAX3lLIk5qUCTA9QelBjGE/NkhzYw/mhnr0s7nI=
-golang.org/x/crypto v0.0.0-20181127143415-eb0de9b17e85/go.mod h1:6SG95UA2DQfeDnfUPMdvaQW0Q7yPrPDi9nlGo2tz2b4=
+go.mozilla.org/pkcs7 v0.0.0-20210826202110-33d05740a352 h1:CCriYyAfq1Br1aIYettdHZTy8mBTIPo7We18TuO/bak=
+go.mozilla.org/pkcs7 v0.0.0-20210826202110-33d05740a352/go.mod h1:SNgMg+EgDFwmvSmLRTNKC5fegJjB7v23qTQ0XLGUNHk=
golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
golang.org/x/crypto v0.0.0-20200728195943-123391ffb6de h1:ikNHVSjEfnvz6sxdSPCaPt572qowuyMDMJLLm3Db3ig=
golang.org/x/crypto v0.0.0-20200728195943-123391ffb6de/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
+golang.org/x/crypto v0.10.0 h1:LKqV2xt9+kDzSTfOhx4FrkEBcMrAgHSYgzywV9zcGmM=
+golang.org/x/crypto v0.10.0/go.mod h1:o4eNf7Ede1fv+hwOwZsTHl9EsPFO6q6ZvYR8vYfY45I=
golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20190412213103-97732733099d/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
diff --git a/meson.build b/meson.build
index b9d18ad..d68a99d 100644
--- a/meson.build
+++ b/meson.build
@@ -2,8 +2,8 @@ project('pdf-simple-sign', 'cpp', default_options : ['cpp_std=c++11'],
version : '1.1.1')
conf = configuration_data()
-conf.set('PROJECT_NAME', '"' + meson.project_name() + '"')
-conf.set('PROJECT_VERSION', '"' + meson.project_version() + '"')
+conf.set_quoted('PROJECT_NAME', meson.project_name())
+conf.set_quoted('PROJECT_VERSION', meson.project_version())
configure_file(output : 'config.h', configuration : conf)
cryptodep = dependency('libcrypto')
diff --git a/pdf-simple-sign.adoc b/pdf-simple-sign.adoc
index 491fa64..4ab1bc5 100644
--- a/pdf-simple-sign.adoc
+++ b/pdf-simple-sign.adoc
@@ -14,7 +14,7 @@ Synopsis
Description
-----------
-'pdf-simple-sign' is a simple PDF signer intended for documents produced by
+*pdf-simple-sign* is a simple PDF signer intended for documents produced by
the Cairo library, GNU troff, ImageMagick, or similar. As such, it currently
comes with some restrictions:
diff --git a/pdf-simple-sign.cpp b/pdf-simple-sign.cpp
index 6bbb284..8b9d1fe 100644
--- a/pdf-simple-sign.cpp
+++ b/pdf-simple-sign.cpp
@@ -16,26 +16,26 @@
// CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
//
-#include <cstdio>
#include <cmath>
+#include <cstdio>
#undef NDEBUG
#include <cassert>
-#include <vector>
#include <map>
-#include <regex>
#include <memory>
+#include <regex>
#include <set>
+#include <vector>
#if defined __GLIBCXX__ && __GLIBCXX__ < 20140422
#error Need libstdc++ >= 4.9 for <regex>
#endif
-#include <unistd.h>
#include <getopt.h>
#include <openssl/err.h>
-#include <openssl/x509v3.h>
#include <openssl/pkcs12.h>
+#include <openssl/x509v3.h>
+#include <unistd.h>
#include "config.h"
@@ -55,7 +55,7 @@ static std::string concatenate(const std::vector<std::string>& v, const std::str
template<typename... Args>
std::string ssprintf(const std::string& format, Args... args) {
- size_t size = std::snprintf(nullptr, 0, format.c_str(), args... ) + 1;
+ size_t size = std::snprintf(nullptr, 0, format.c_str(), args...) + 1;
std::unique_ptr<char[]> buf(new char[size]);
std::snprintf(buf.get(), size, format.c_str(), args...);
return std::string(buf.get(), buf.get() + size - 1);
@@ -64,7 +64,7 @@ std::string ssprintf(const std::string& format, Args... args) {
// -------------------------------------------------------------------------------------------------
/// PDF token/object thingy. Objects may be composed either from one or a sequence of tokens.
-/// The PDF Reference doesn't actually speak of tokens.
+/// The PDF Reference doesn't actually speak of tokens, though ISO 32000-1:2008 does.
struct pdf_object {
enum type {
END, NL, COMMENT, NIL, BOOL, NUMERIC, KEYWORD, NAME, STRING,
@@ -262,14 +262,12 @@ static std::string pdf_serialize(const pdf_object& o) {
case pdf_object::NL: return "\n";
case pdf_object::NIL: return "null";
case pdf_object::BOOL: return o.number ? "true" : "false";
- case pdf_object::NUMERIC:
- {
+ case pdf_object::NUMERIC: {
if (o.is_integer()) return std::to_string((long long) o.number);
return std::to_string(o.number);
}
case pdf_object::KEYWORD: return o.string;
- case pdf_object::NAME:
- {
+ case pdf_object::NAME: {
std::string escaped = "/";
for (char c : o.string) {
if (c == '#' || strchr(pdf_lexer::delimiters, c) || strchr(pdf_lexer::whitespace, c))
@@ -279,8 +277,7 @@ static std::string pdf_serialize(const pdf_object& o) {
}
return escaped;
}
- case pdf_object::STRING:
- {
+ case pdf_object::STRING: {
std::string escaped;
for (char c : o.string) {
if (c == '\\' || c == '(' || c == ')')
@@ -293,15 +290,13 @@ static std::string pdf_serialize(const pdf_object& o) {
case pdf_object::E_ARRAY: return "]";
case pdf_object::B_DICT: return "<<";
case pdf_object::E_DICT: return ">>";
- case pdf_object::ARRAY:
- {
+ case pdf_object::ARRAY: {
std::vector<std::string> v;
for (const auto& i : o.array)
v.push_back(pdf_serialize(i));
return "[ " + concatenate(v, " ") + " ]";
}
- case pdf_object::DICT:
- {
+ case pdf_object::DICT: {
std::string s;
for (const auto i : o.dict)
// FIXME the key is also supposed to be escaped by pdf_serialize()
@@ -372,8 +367,8 @@ pdf_object pdf_updater::parse_obj(pdf_lexer& lex, std::vector<pdf_object>& stack
auto g = stack.back(); stack.pop_back();
auto n = stack.back(); stack.pop_back();
- if (!g.is_integer() || g.number < 0 || g.number > UINT_MAX
- || !n.is_integer() || n.number < 0 || n.number > UINT_MAX)
+ if (!g.is_integer() || g.number < 0 || g.number > UINT_MAX ||
+ !n.is_integer() || n.number < 0 || n.number > UINT_MAX)
return {pdf_object::END, "invalid object ID pair"};
pdf_object obj{pdf_object::OBJECT};
@@ -397,8 +392,8 @@ pdf_object pdf_updater::parse_R(std::vector<pdf_object>& stack) const {
auto g = stack.back(); stack.pop_back();
auto n = stack.back(); stack.pop_back();
- if (!g.is_integer() || g.number < 0 || g.number > UINT_MAX
- || !n.is_integer() || n.number < 0 || n.number > UINT_MAX)
+ if (!g.is_integer() || g.number < 0 || g.number > UINT_MAX ||
+ !n.is_integer() || n.number < 0 || n.number > UINT_MAX)
return {pdf_object::END, "invalid reference ID pair"};
pdf_object ref{pdf_object::REFERENCE};
@@ -415,8 +410,7 @@ pdf_object pdf_updater::parse(pdf_lexer& lex, std::vector<pdf_object>& stack) co
case pdf_object::COMMENT:
// These are not important to parsing, not even for this procedure's needs
return parse(lex, stack);
- case pdf_object::B_ARRAY:
- {
+ case pdf_object::B_ARRAY: {
std::vector<pdf_object> array;
while (1) {
auto object = parse(lex, array);
@@ -428,8 +422,7 @@ pdf_object pdf_updater::parse(pdf_lexer& lex, std::vector<pdf_object>& stack) co
}
return array;
}
- case pdf_object::B_DICT:
- {
+ case pdf_object::B_DICT: {
std::vector<pdf_object> array;
while (1) {
auto object = parse(lex, array);
@@ -477,8 +470,8 @@ std::string pdf_updater::load_xref(pdf_lexer& lex, std::set<uint>& loaded_entrie
break;
auto second = parse(lex, throwaway_stack);
- if (!object.is_integer() || object.number < 0 || object.number > UINT_MAX
- || !second.is_integer() || second.number < 0 || second.number > UINT_MAX)
+ if (!object.is_integer() || object.number < 0 || object.number > UINT_MAX ||
+ !second.is_integer() || second.number < 0 || second.number > UINT_MAX)
return "invalid xref section header";
const size_t start = object.number;
@@ -487,9 +480,9 @@ std::string pdf_updater::load_xref(pdf_lexer& lex, std::set<uint>& loaded_entrie
auto off = parse(lex, throwaway_stack);
auto gen = parse(lex, throwaway_stack);
auto key = parse(lex, throwaway_stack);
- if (!off.is_integer() || off.number < 0 || off.number > document.length()
- || !gen.is_integer() || gen.number < 0 || gen.number > 65535
- || key.type != pdf_object::KEYWORD)
+ if (!off.is_integer() || off.number < 0 || off.number > document.length() ||
+ !gen.is_integer() || gen.number < 0 || gen.number > 65535 ||
+ key.type != pdf_object::KEYWORD)
return "invalid xref entry";
bool free = true;
@@ -550,8 +543,8 @@ std::string pdf_updater::initialize() {
const auto prev_offset = trailer.dict.find("Prev");
if (prev_offset == trailer.dict.end())
break;
- // FIXME we don't check for size_t over or underflow
- if (!prev_offset->second.is_integer())
+ // FIXME do not read offsets and sizes as floating point numbers
+ if (!prev_offset->second.is_integer() || prev_offset->second.number < 0)
return "invalid Prev offset";
xref_offset = prev_offset->second.number;
}
@@ -657,8 +650,8 @@ void pdf_updater::flush_updates() {
}
trailer["Size"] = {pdf_object::NUMERIC, double(xref_size)};
- document += "trailer\n" + pdf_serialize(trailer)
- + ssprintf("\nstartxref\n%zu\n%%%%EOF\n", startxref);
+ document +=
+ "trailer\n" + pdf_serialize(trailer) + ssprintf("\nstartxref\n%zu\n%%%%EOF\n", startxref);
}
// -------------------------------------------------------------------------------------------------
@@ -700,9 +693,9 @@ static pdf_object pdf_get_first_page(pdf_updater& pdf, uint node_n, uint node_ge
// XXX technically speaking, this may be an indirect reference. The correct way to solve this
// seems to be having "pdf_updater" include a wrapper around "obj.dict.find"
auto kids = obj.dict.find("Kids");
- if (kids == obj.dict.end() || kids->second.type != pdf_object::ARRAY
- || kids->second.array.empty()
- || kids->second.array.at(0).type != pdf_object::REFERENCE)
+ if (kids == obj.dict.end() || kids->second.type != pdf_object::ARRAY ||
+ kids->second.array.empty() ||
+ kids->second.array.at(0).type != pdf_object::REFERENCE)
return {pdf_object::NIL};
// XXX nothing prevents us from recursing in an evil circular graph
@@ -740,8 +733,8 @@ static std::string pdf_fill_in_signature(std::string& document, size_t sign_off,
// OpenSSL error reasons will usually be of more value than any distinction I can come up with
std::string err = "OpenSSL failure";
- if (!(p12 = d2i_PKCS12_fp(pkcs12_fp, nullptr))
- || !PKCS12_parse(p12, pkcs12_pass.c_str(), &private_key, &certificate, &chain)) {
+ if (!(p12 = d2i_PKCS12_fp(pkcs12_fp, nullptr)) ||
+ !PKCS12_parse(p12, pkcs12_pass.c_str(), &private_key, &certificate, &chain)) {
err = pkcs12_path + ": parse failure";
goto error;
}
@@ -766,8 +759,8 @@ static std::string pdf_fill_in_signature(std::string& document, size_t sign_off,
#endif
// The default digest is SHA1, which is mildly insecure now -- hence using PKCS7_sign_add_signer
- if (!(p7 = PKCS7_sign(nullptr, nullptr, nullptr, nullptr, sign_flags))
- || !PKCS7_sign_add_signer(p7, certificate, private_key, EVP_sha256(), sign_flags))
+ if (!(p7 = PKCS7_sign(nullptr, nullptr, nullptr, nullptr, sign_flags)) ||
+ !PKCS7_sign_add_signer(p7, certificate, private_key, EVP_sha256(), sign_flags))
goto error;
// For RFC 3161, this is roughly how a timestamp token would be attached (see Appendix A):
// PKCS7_add_attribute(signer_info, NID_id_smime_aa_timeStampToken, V_ASN1_SEQUENCE, value)
@@ -777,10 +770,10 @@ static std::string pdf_fill_in_signature(std::string& document, size_t sign_off,
// Adaptation of the innards of the undocumented PKCS7_final() -- I didn't feel like making
// a copy of the whole document. Hopefully this writes directly into a digest BIO.
- if (!(p7bio = PKCS7_dataInit(p7, nullptr))
- || (ssize_t) sign_off != BIO_write(p7bio, document.data(), sign_off)
- || (ssize_t) tail_len != BIO_write(p7bio, document.data() + tail_off, tail_len)
- || BIO_flush(p7bio) != 1 || !PKCS7_dataFinal(p7, p7bio))
+ if (!(p7bio = PKCS7_dataInit(p7, nullptr)) ||
+ (ssize_t) sign_off != BIO_write(p7bio, document.data(), sign_off) ||
+ (ssize_t) tail_len != BIO_write(p7bio, document.data() + tail_off, tail_len) ||
+ BIO_flush(p7bio) != 1 || !PKCS7_dataFinal(p7, p7bio))
goto error;
#if 0
@@ -850,7 +843,7 @@ static std::string pdf_sign(std::string& document, ushort reservation) {
// 8.7 Digital Signatures - /signature dictionary/
auto sigdict_n = pdf.allocate();
size_t byterange_off = 0, byterange_len = 0, sign_off = 0, sign_len = 0;
- pdf.update(sigdict_n, [&]{
+ pdf.update(sigdict_n, [&] {
// The timestamp is important for Adobe Acrobat Reader DC. The ideal would be to use RFC 3161.
pdf.document.append("<< /Type/Sig /Filter/Adobe.PPKLite /SubFilter/adbe.pkcs7.detached\n"
" /M" + pdf_serialize(pdf_date(time(nullptr))) + " /ByteRange ");
@@ -883,7 +876,7 @@ static std::string pdf_sign(std::string& document, ushort reservation) {
}}});
auto sigfield_n = pdf.allocate();
- pdf.update(sigfield_n, [&]{ pdf.document += pdf_serialize(sigfield); });
+ pdf.update(sigfield_n, [&] { pdf.document += pdf_serialize(sigfield); });
auto pages_ref = root.dict.find("Pages");
if (pages_ref == root.dict.end() || pages_ref->second.type != pdf_object::REFERENCE)
@@ -901,7 +894,7 @@ static std::string pdf_sign(std::string& document, ushort reservation) {
annots = {pdf_object::ARRAY};
}
annots.array.emplace_back(pdf_object::REFERENCE, sigfield_n, 0);
- pdf.update(page.n, [&]{ pdf.document += pdf_serialize(page); });
+ pdf.update(page.n, [&] { pdf.document += pdf_serialize(page); });
// 8.6.1 Interactive Form Dictionary
if (root.dict.count("AcroForm"))
@@ -918,7 +911,7 @@ static std::string pdf_sign(std::string& document, ushort reservation) {
if (pdf.version(root) < 16)
root.dict["Version"] = {pdf_object::NAME, "1.6"};
- pdf.update(root_ref->second.n, [&]{ pdf.document += pdf_serialize(root); });
+ pdf.update(root_ref->second.n, [&] { pdf.document += pdf_serialize(root); });
pdf.flush_updates();
// Now that we know the length of everything, store byte ranges of what we're about to sign,
@@ -947,9 +940,9 @@ static void die(int status, const char* format, ...) {
int main(int argc, char* argv[]) {
auto invocation_name = argv[0];
- auto usage = [=]{
+ auto usage = [=] {
die(1, "Usage: %s [-h] [-r RESERVATION] INPUT-FILENAME OUTPUT-FILENAME PKCS12-PATH PKCS12-PASS",
- invocation_name);
+ invocation_name);
};
static struct option opts[] = {
@@ -963,8 +956,7 @@ int main(int argc, char* argv[]) {
long reservation = 4096;
while (1) {
int option_index = 0;
- auto c = getopt_long(argc, const_cast<char* const*>(argv),
- "hVr:", opts, &option_index);
+ auto c = getopt_long(argc, const_cast<char* const*>(argv), "hVr:", opts, &option_index);
if (c == -1)
break;
diff --git a/pdf/pdf.go b/pdf/pdf.go
index 4f1f5e1..1fcdaa4 100644
--- a/pdf/pdf.go
+++ b/pdf/pdf.go
@@ -1,5 +1,5 @@
//
-// Copyright (c) 2018 - 2020, Přemysl Eric Janouch <p@janouch.name>
+// Copyright (c) 2018 - 2024, Přemysl Eric Janouch <p@janouch.name>
//
// Permission to use, copy, modify, and/or distribute this software for any
// purpose with or without fee is hereby granted.
@@ -18,6 +18,8 @@ package pdf
import (
"bytes"
+ "compress/zlib"
+ "encoding/binary"
"encoding/hex"
"errors"
"fmt"
@@ -59,20 +61,22 @@ const (
// higher-level objects
Array
Dict
+ Stream
Indirect
Reference
)
-// Object is a PDF token/object thingy. Objects may be composed either from
+// Object is a PDF token/object thingy. Objects may be composed either from
// one or a sequence of tokens. The PDF Reference doesn't actually speak
-// of tokens.
+// of tokens, though ISO 32000-1:2008 does.
type Object struct {
Kind ObjectKind
String string // Comment/Keyword/Name/String
Number float64 // Bool, Numeric
Array []Object // Array, Indirect
- Dict map[string]Object // Dict, in the future also Stream
+ Dict map[string]Object // Dict, Stream
+ Stream []byte // Stream
N, Generation uint // Indirect, Reference
}
@@ -118,6 +122,13 @@ func NewDict(d map[string]Object) Object {
return Object{Kind: Dict, Dict: d}
}
+func NewStream(d map[string]Object, s []byte) Object {
+ if d == nil {
+ d = make(map[string]Object)
+ }
+ return Object{Kind: Stream, Dict: d, Stream: s}
+}
+
func NewIndirect(o Object, n, generation uint) Object {
return Object{Kind: Indirect, N: n, Generation: generation,
Array: []Object{o}}
@@ -458,6 +469,10 @@ func (o *Object) Serialize() string {
fmt.Fprint(b, " /", k, " ", v.Serialize())
}
return "<<" + b.String() + " >>"
+ case Stream:
+ d := NewDict(o.Dict)
+ d.Dict["Length"] = NewNumeric(float64(len(o.Stream)))
+ return d.Serialize() + "\nstream\n" + string(o.Stream) + "\nendstream"
case Indirect:
return fmt.Sprintf("%d %d obj\n%s\nendobj", o.N, o.Generation,
o.Array[0].Serialize())
@@ -471,8 +486,9 @@ func (o *Object) Serialize() string {
// -----------------------------------------------------------------------------
type ref struct {
- offset int64 // file offset or N of the next free entry
+ offset int64 // file offset, or N of the next free entry, or index
generation uint // object generation
+ compressed *uint // PDF 1.5: N of the containing compressed object
nonfree bool // whether this N is taken (for a good zero value)
}
@@ -497,6 +513,65 @@ type Updater struct {
Trailer map[string]Object
}
+// ListIndirect returns the whole cross-reference table as Reference Objects.
+func (u *Updater) ListIndirect() []Object {
+ result := []Object{}
+ for i := 0; i < len(u.xref); i++ {
+ if u.xref[i].nonfree {
+ result = append(result, NewReference(uint(i), u.xref[i].generation))
+ }
+ }
+ return result
+}
+
+func (u *Updater) parseStream(lex *Lexer, stack *[]Object) (Object, error) {
+ lenStack := len(*stack)
+ if lenStack < 1 {
+ return newError("missing stream dictionary")
+ }
+ dict := (*stack)[lenStack-1]
+ if dict.Kind != Dict {
+ return newError("stream not preceded by a dictionary")
+ }
+
+ *stack = (*stack)[:lenStack-1]
+ length, ok := dict.Dict["Length"]
+ if !ok {
+ return newError("missing stream Length")
+ }
+ length, err := u.Dereference(length)
+ if err != nil {
+ return length, err
+ }
+ if !length.IsUint() || length.Number > math.MaxInt {
+ return newError("stream Length not an unsigned integer")
+ }
+
+ // Expect exactly one newline.
+ if nl, err := lex.Next(); err != nil {
+ return nl, err
+ } else if nl.Kind != NL {
+ return newError("stream does not start with a newline")
+ }
+
+ size := int(length.Number)
+ if len(lex.P) < size {
+ return newError("stream is longer than the document")
+ }
+
+ dict.Kind = Stream
+ dict.Stream = lex.P[:size]
+ lex.P = lex.P[size:]
+
+ // Skip any number of trailing newlines or comments.
+ if end, err := u.parse(lex, stack); err != nil {
+ return end, err
+ } else if end.Kind != Keyword || end.String != "endstream" {
+ return newError("improperly terminated stream")
+ }
+ return dict, nil
+}
+
func (u *Updater) parseIndirect(lex *Lexer, stack *[]Object) (Object, error) {
lenStack := len(*stack)
if lenStack < 2 {
@@ -590,15 +665,11 @@ func (u *Updater) parse(lex *Lexer, stack *[]Object) (Object, error) {
}
return NewDict(dict), nil
case Keyword:
- // Appears in the document body, typically needs
- // to access the cross-reference table.
- //
- // TODO(p): Use the xref to read /Length etc. once we
- // actually need to read such objects; presumably
- // streams can use the Object.String member.
switch token.String {
case "stream":
- return newError("streams are not supported yet")
+ // Appears in the document body,
+ // typically needs to access the cross-reference table.
+ return u.parseStream(lex, stack)
case "obj":
return u.parseIndirect(lex, stack)
case "R":
@@ -610,16 +681,159 @@ func (u *Updater) parse(lex *Lexer, stack *[]Object) (Object, error) {
}
}
-func (u *Updater) loadXref(lex *Lexer, loadedEntries map[uint]struct{}) error {
+func (u *Updater) loadXrefEntry(
+ n uint, r ref, loadedEntries map[uint]struct{}) {
+ if _, ok := loadedEntries[n]; ok {
+ return
+ }
+ if lenXref := uint(len(u.xref)); n >= lenXref {
+ u.xref = append(u.xref, make([]ref, n-lenXref+1)...)
+ }
+ loadedEntries[n] = struct{}{}
+
+ u.xref[n] = r
+}
+
+func (u *Updater) loadXrefStream(
+ lex *Lexer, stack []Object, loadedEntries map[uint]struct{}) (
+ Object, error) {
+ var object Object
+ for {
+ var err error
+ if object, err = u.parse(lex, &stack); err != nil {
+ return New(End), fmt.Errorf("invalid xref table: %s", err)
+ } else if object.Kind == End {
+ return newError("invalid xref table")
+ }
+
+ // For the sake of simplicity, keep stacking until we find an object.
+ if object.Kind == Indirect {
+ break
+ }
+
+ stack = append(stack, object)
+ }
+
+ // ISO 32000-2:2020 7.5.8.2 Cross-reference stream dictionary
+ stream := object.Array[0]
+ if stream.Kind != Stream {
+ return newError("invalid xref table")
+ }
+ if typ, ok := stream.Dict["Type"]; !ok ||
+ typ.Kind != Name || typ.String != "XRef" {
+ return newError("invalid xref stream")
+ }
+
+ data, err := u.GetStreamData(stream)
+ if err != nil {
+ return New(End), fmt.Errorf("invalid xref stream: %s", err)
+ }
+
+ size, ok := stream.Dict["Size"]
+ if !ok || !size.IsUint() || size.Number <= 0 {
+ return newError("invalid or missing cross-reference stream Size")
+ }
+
+ type pair struct{ start, count uint }
+ pairs := []pair{}
+ if index, ok := stream.Dict["Index"]; !ok {
+ pairs = append(pairs, pair{0, uint(size.Number)})
+ } else {
+ if index.Kind != Array || len(index.Array)%2 != 0 {
+ return newError("invalid cross-reference stream Index")
+ }
+
+ a := index.Array
+ for i := 0; i < len(a); i += 2 {
+ if !a[i].IsUint() || !a[i+1].IsUint() {
+ return newError("invalid cross-reference stream Index")
+ }
+ pairs = append(pairs, pair{uint(a[i].Number), uint(a[i+1].Number)})
+ }
+ }
+
+ w, ok := stream.Dict["W"]
+ if !ok || w.Kind != Array || len(w.Array) != 3 ||
+ !w.Array[0].IsUint() || !w.Array[1].IsUint() || !w.Array[2].IsUint() {
+ return newError("invalid or missing cross-reference stream W")
+ }
+
+ w1 := uint(w.Array[0].Number)
+ w2 := uint(w.Array[1].Number)
+ w3 := uint(w.Array[2].Number)
+ if w2 == 0 {
+ return newError("invalid cross-reference stream W")
+ }
+
+ unit := w1 + w2 + w3
+ if uint(len(data))%unit != 0 {
+ return newError("invalid cross-reference stream length")
+ }
+
+ readField := func(data []byte, width uint) (uint, []byte) {
+ var n uint
+ for ; width != 0; width-- {
+ n = n<<8 | uint(data[0])
+ data = data[1:]
+ }
+ return n, data
+ }
+
+ // ISO 32000-2:2020 7.5.8.3 Cross-reference stream data
+ for _, pair := range pairs {
+ for i := uint(0); i < pair.count; i++ {
+ if uint(len(data)) < unit {
+ return newError("premature cross-reference stream EOF")
+ }
+
+ var f1, f2, f3 uint = 1, 0, 0
+ if w1 > 0 {
+ f1, data = readField(data, w1)
+ }
+ f2, data = readField(data, w2)
+ if w3 > 0 {
+ f3, data = readField(data, w3)
+ }
+
+ var r ref
+ switch f1 {
+ case 0:
+ r.offset = int64(f2)
+ r.generation = f3
+ case 1:
+ r.offset = int64(f2)
+ r.generation = f3
+ r.nonfree = true
+ case 2:
+ r.offset = int64(f3)
+ r.compressed = &f2
+ r.nonfree = true
+ default:
+ // TODO(p): It should be treated as a reference to
+ // the null object. We can't currently represent that.
+ return newError("unsupported cross-reference stream contents")
+ }
+
+ u.loadXrefEntry(pair.start+i, r, loadedEntries)
+ }
+ }
+
+ stream.Kind = Dict
+ stream.Stream = nil
+ return stream, nil
+}
+
+func (u *Updater) loadXref(lex *Lexer, loadedEntries map[uint]struct{}) (
+ Object, error) {
var throwawayStack []Object
- if keyword, _ := u.parse(lex,
- &throwawayStack); keyword.Kind != Keyword || keyword.String != "xref" {
- return errors.New("invalid xref table")
+ if object, _ := u.parse(lex,
+ &throwawayStack); object.Kind != Keyword || object.String != "xref" {
+ return u.loadXrefStream(lex, []Object{object}, loadedEntries)
}
for {
object, _ := u.parse(lex, &throwawayStack)
if object.Kind == End {
- return errors.New("unexpected EOF while looking for the trailer")
+ return newError("unexpected EOF while looking for the trailer")
}
if object.Kind == Keyword && object.String == "trailer" {
break
@@ -627,7 +841,7 @@ func (u *Updater) loadXref(lex *Lexer, loadedEntries map[uint]struct{}) error {
second, _ := u.parse(lex, &throwawayStack)
if !object.IsUint() || !second.IsUint() {
- return errors.New("invalid xref section header")
+ return newError("invalid xref section header")
}
start, count := uint(object.Number), uint(second.Number)
@@ -639,33 +853,29 @@ func (u *Updater) loadXref(lex *Lexer, loadedEntries map[uint]struct{}) error {
off.Number > float64(len(u.Document)) ||
!gen.IsInteger() || gen.Number < 0 || gen.Number > 65535 ||
key.Kind != Keyword {
- return errors.New("invalid xref entry")
+ return newError("invalid xref entry")
}
free := true
if key.String == "n" {
free = false
} else if key.String != "f" {
- return errors.New("invalid xref entry")
- }
-
- n := start + i
- if _, ok := loadedEntries[n]; ok {
- continue
+ return newError("invalid xref entry")
}
- if lenXref := uint(len(u.xref)); n >= lenXref {
- u.xref = append(u.xref, make([]ref, n-lenXref+1)...)
- }
- loadedEntries[n] = struct{}{}
- u.xref[n] = ref{
+ u.loadXrefEntry(start+i, ref{
offset: int64(off.Number),
generation: uint(gen.Number),
nonfree: !free,
- }
+ }, loadedEntries)
}
}
- return nil
+
+ trailer, _ := u.parse(lex, &throwawayStack)
+ if trailer.Kind != Dict {
+ return newError("invalid trailer dictionary")
+ }
+ return trailer, nil
}
// -----------------------------------------------------------------------------
@@ -695,7 +905,6 @@ func NewUpdater(document []byte) (*Updater, error) {
loadedXrefs := make(map[int64]struct{})
loadedEntries := make(map[uint]struct{})
- var throwawayStack []Object
for {
if _, ok := loadedXrefs[xrefOffset]; ok {
return nil, errors.New("circular xref offsets")
@@ -705,24 +914,26 @@ func NewUpdater(document []byte) (*Updater, error) {
}
lex := Lexer{u.Document[xrefOffset:]}
- if err := u.loadXref(&lex, loadedEntries); err != nil {
+ trailer, err := u.loadXref(&lex, loadedEntries)
+ if err != nil {
return nil, err
}
- trailer, _ := u.parse(&lex, &throwawayStack)
- if trailer.Kind != Dict {
- return nil, errors.New("invalid trailer dictionary")
- }
if len(loadedXrefs) == 0 {
u.Trailer = trailer.Dict
}
loadedXrefs[xrefOffset] = struct{}{}
+ // TODO(p): Descend into XRefStm here first, if present,
+ // which is also a linked list.
+
+ // We allow for mixed cross-reference tables and streams
+ // within a single Prev list, although this should never occur.
prevOffset, ok := trailer.Dict["Prev"]
if !ok {
break
}
- // FIXME: We don't check for size_t over or underflow.
+ // FIXME: Do not read offsets and sizes as floating point numbers.
if !prevOffset.IsInteger() {
return nil, errors.New("invalid Prev offset")
}
@@ -764,18 +975,115 @@ func (u *Updater) Version(root *Object) int {
return 0
}
+func (u *Updater) getFromObjStm(nObjStm, n uint) (Object, error) {
+ if nObjStm == n {
+ return newError("ObjStm recursion")
+ }
+
+ stream, err := u.Get(nObjStm, 0)
+ if err != nil {
+ return stream, err
+ }
+ if stream.Kind != Stream {
+ return newError("invalid ObjStm")
+ }
+ if typ, ok := stream.Dict["Type"]; !ok ||
+ typ.Kind != Name || typ.String != "ObjStm" {
+ return newError("invalid ObjStm")
+ }
+
+ data, err := u.GetStreamData(stream)
+ if err != nil {
+ return New(End), fmt.Errorf("invalid ObjStm: %s", err)
+ }
+ entryN, ok := stream.Dict["N"]
+ if !ok || !entryN.IsUint() || entryN.Number <= 0 {
+ return newError("invalid ObjStm N")
+ }
+ entryFirst, ok := stream.Dict["First"]
+ if !ok || !entryFirst.IsUint() || entryFirst.Number <= 0 {
+ return newError("invalid ObjStm First")
+ }
+
+ // NOTE: This means descending into that stream if n is not found here.
+ // It is meant to be an object reference.
+ if extends, ok := stream.Dict["Extends"]; ok && extends.Kind != Nil {
+ return newError("ObjStm extensions are unsupported")
+ }
+
+ count := uint(entryN.Number)
+ first := uint(entryFirst.Number)
+ if first > uint(len(data)) {
+ return newError("invalid ObjStm First")
+ }
+
+ lex1 := Lexer{data[:first]}
+ data = data[first:]
+
+ type pair struct{ n, offset uint }
+ pairs := []pair{}
+ for i := uint(0); i < count; i++ {
+ var throwawayStack []Object
+ objN, _ := u.parse(&lex1, &throwawayStack)
+ objOffset, _ := u.parse(&lex1, &throwawayStack)
+ if !objN.IsUint() || !objOffset.IsUint() {
+ return newError("invalid ObjStm pairs")
+ }
+ pairs = append(pairs, pair{uint(objN.Number), uint(objOffset.Number)})
+ }
+ for i, pair := range pairs {
+ if pair.offset > uint(len(data)) ||
+ i > 0 && pairs[i-1].offset >= pair.offset {
+ return newError("invalid ObjStm pairs")
+ }
+ }
+
+ for i, pair := range pairs {
+ if pair.n != n {
+ continue
+ }
+
+ if i+1 < len(pairs) {
+ data = data[pair.offset:pairs[i+1].offset]
+ } else {
+ data = data[pair.offset:]
+ }
+
+ lex2 := Lexer{data}
+ var stack []Object
+ for {
+ object, err := u.parse(&lex2, &stack)
+ if err != nil {
+ return object, err
+ } else if object.Kind == End {
+ break
+ } else {
+ stack = append(stack, object)
+ }
+ }
+ if len(stack) == 0 {
+ return newError("empty ObjStm object")
+ }
+ return stack[0], nil
+ }
+ return newError("object not found in ObjStm")
+}
+
// Get retrieves an object by its number and generation--may return
// Nil or End with an error.
-//
-// TODO(p): We should fix all uses of this not to eat the error.
func (u *Updater) Get(n, generation uint) (Object, error) {
if n >= u.xrefSize {
return New(Nil), nil
}
ref := u.xref[n]
- if !ref.nonfree || ref.generation != generation ||
- ref.offset >= int64(len(u.Document)) {
+ if !ref.nonfree || ref.generation != generation {
+ return New(Nil), nil
+ }
+
+ if ref.compressed != nil {
+ return u.getFromObjStm(*ref.compressed, n)
+ } else if ref.offset >= int64(len(u.Document)) {
return New(Nil), nil
}
@@ -796,6 +1104,14 @@ func (u *Updater) Get(n, generation uint) (Object, error) {
}
}
+// Derefence dereferences Reference objects, and passes the other kinds through.
+func (u *Updater) Dereference(o Object) (Object, error) {
+ if o.Kind != Reference {
+ return o, nil
+ }
+ return u.Get(o.N, o.Generation)
+}
+
// Allocate allocates a new object number.
func (u *Updater) Allocate() uint {
n := u.xrefSize
@@ -822,8 +1138,8 @@ type BytesWriter interface {
WriteString(s string) (n int, err error)
}
-// Update appends an updated object to the end of the document. The fill
-// callback must write exactly one PDF object.
+// Update appends an updated object to the end of the document.
+// The fill callback must write exactly one PDF object.
func (u *Updater) Update(n uint, fill func(buf BytesWriter)) {
oldRef := u.xref[n]
u.updated[n] = struct{}{}
@@ -843,20 +1159,62 @@ func (u *Updater) Update(n uint, fill func(buf BytesWriter)) {
u.Document = buf.Bytes()
}
-// FlushUpdates writes an updated cross-reference table and trailer.
-func (u *Updater) FlushUpdates() {
- updated := make([]uint, 0, len(u.updated))
- for n := range u.updated {
- updated = append(updated, n)
+func (u *Updater) flushXRefStm(updated []uint, buf *bytes.Buffer) {
+ // The cross-reference stream has to point to itself.
+ // XXX: We only duplicate Update code here due to how we currently buffer.
+ n := u.Allocate()
+ updated = append(updated, n)
+
+ u.updated[n] = struct{}{}
+ u.xref[n] = ref{
+ offset: int64(buf.Len() + 1),
+ generation: 0,
+ nonfree: true,
}
- sort.Slice(updated, func(i, j int) bool {
- return updated[i] < updated[j]
+
+ index, b := []Object{}, []byte{}
+ write := func(f1 byte, f2, f3 uint64) {
+ b = append(b, f1)
+ b = binary.BigEndian.AppendUint64(b, f2)
+ b = binary.BigEndian.AppendUint64(b, f3)
+ }
+ for i := 0; i < len(updated); {
+ start, stop := updated[i], updated[i]+1
+ for i++; i < len(updated) && updated[i] == stop; i++ {
+ stop++
+ }
+
+ index = append(index,
+ NewNumeric(float64(start)), NewNumeric(float64(stop-start)))
+ for ; start < stop; start++ {
+ ref := u.xref[start]
+ if ref.compressed != nil {
+ write(2, uint64(*ref.compressed), uint64(ref.offset))
+ } else if ref.nonfree {
+ write(1, uint64(ref.offset), uint64(ref.generation))
+ } else {
+ write(0, uint64(ref.offset), uint64(ref.generation))
+ }
+ }
+ }
+
+ u.Trailer["Size"] = NewNumeric(float64(u.xrefSize))
+ u.Trailer["Index"] = NewArray(index)
+ u.Trailer["W"] = NewArray([]Object{
+ NewNumeric(1), NewNumeric(8), NewNumeric(8),
})
- buf := bytes.NewBuffer(u.Document)
- startXref := buf.Len() + 1
- buf.WriteString("\nxref\n")
+ for _, key := range []string{
+ "Filter", "DecodeParms", "F", "FFilter", "FDecodeParms", "DL"} {
+ delete(u.Trailer, key)
+ }
+ stream := NewStream(u.Trailer, b)
+ fmt.Fprintf(buf, "\n%d 0 obj\n%s\nendobj", n, stream.Serialize())
+}
+
+func (u *Updater) flushXRefTable(updated []uint, buf *bytes.Buffer) {
+ buf.WriteString("\nxref\n")
for i := 0; i < len(updated); {
start, stop := updated[i], updated[i]+1
for i++; i < len(updated) && updated[i] == stop; i++ {
@@ -865,8 +1223,9 @@ func (u *Updater) FlushUpdates() {
fmt.Fprintf(buf, "%d %d\n", start, stop-start)
for ; start < stop; start++ {
+ // XXX: We should warn about any object streams here.
ref := u.xref[start]
- if ref.nonfree {
+ if ref.nonfree && ref.compressed == nil {
fmt.Fprintf(buf, "%010d %05d n \n", ref.offset, ref.generation)
} else {
fmt.Fprintf(buf, "%010d %05d f \n", ref.offset, ref.generation)
@@ -883,10 +1242,38 @@ func (u *Updater) FlushUpdates() {
u.Trailer["Size"] = NewNumeric(float64(u.xrefSize))
trailer := NewDict(u.Trailer)
+ fmt.Fprintf(buf, "trailer\n%s", trailer.Serialize())
+}
+
+// FlushUpdates writes an updated cross-reference table and trailer, or stream.
+func (u *Updater) FlushUpdates() {
+ updated := make([]uint, 0, len(u.updated))
+ for n := range u.updated {
+ updated = append(updated, n)
+ }
+ sort.Slice(updated, func(i, j int) bool {
+ return updated[i] < updated[j]
+ })
+
+ // It does not seem to be possible to upgrade a PDF file
+ // from trailer dictionaries to cross-reference streams,
+ // so keep continuity either way.
+ //
+ // (Downgrading from cross-reference streams using XRefStm would not
+ // create a true hybrid-reference file, although it should work.)
+ buf := bytes.NewBuffer(u.Document)
+ startXref := buf.Len() + 1 /* '\n' */
+ if typ, _ := u.Trailer["Type"]; typ.Kind == Name && typ.String == "XRef" {
+ u.flushXRefStm(updated, buf)
+ } else {
+ u.flushXRefTable(updated, buf)
+ }
- fmt.Fprintf(buf, "trailer\n%s\nstartxref\n%d\n%%%%EOF\n",
- trailer.Serialize(), startXref)
+ fmt.Fprintf(buf, "\nstartxref\n%d\n%%%%EOF\n", startXref)
u.Document = buf.Bytes()
+ u.updated = make(map[uint]struct{})
+
+ u.Trailer["Prev"] = NewNumeric(float64(startXref))
}
// -----------------------------------------------------------------------------
@@ -904,17 +1291,47 @@ func NewDate(ts time.Time) Object {
return NewString(string(buf))
}
+// GetStreamData returns the actual data stored in a stream object,
+// applying any filters.
+func (u *Updater) GetStreamData(stream Object) ([]byte, error) {
+ if f, ok := stream.Dict["F"]; ok && f.Kind != Nil {
+ return nil, errors.New("stream data in other files are unsupported")
+ }
+
+ // Support just enough to decode a common cross-reference stream.
+ if filter, ok := stream.Dict["Filter"]; !ok {
+ return stream.Stream, nil
+ } else if filter.Kind != Name || filter.String != "FlateDecode" {
+ return nil, errors.New("unsupported stream Filter")
+ }
+
+ // TODO(p): Support << /Columns N /Predictor 12 >>
+ // which usually appears in files with cross-reference streams.
+ if parms, ok := stream.Dict["DecodeParms"]; ok && parms.Kind != Nil {
+ return nil, errors.New("DecodeParms are not supported")
+ }
+
+ r, err := zlib.NewReader(bytes.NewReader(stream.Stream))
+ if err != nil {
+ return nil, err
+ }
+
+ var b bytes.Buffer
+ _, err = b.ReadFrom(r)
+ return b.Bytes(), err
+}
+
// GetFirstPage retrieves the first page of the given page (sub)tree reference,
// or returns a Nil object if unsuccessful.
-func (u *Updater) GetFirstPage(nodeN, nodeGeneration uint) Object {
- obj, _ := u.Get(nodeN, nodeGeneration)
- if obj.Kind != Dict {
+func (u *Updater) GetFirstPage(node Object) Object {
+ obj, err := u.Dereference(node)
+ if err != nil || obj.Kind != Dict {
return New(Nil)
}
// Out of convenience; these aren't filled normally.
- obj.N = nodeN
- obj.Generation = nodeGeneration
+ obj.N = node.N
+ obj.Generation = node.Generation
if typ, ok := obj.Dict["Type"]; !ok || typ.Kind != Name {
return New(Nil)
@@ -934,7 +1351,7 @@ func (u *Updater) GetFirstPage(nodeN, nodeGeneration uint) Object {
}
// XXX: Nothing prevents us from recursing in an evil circular graph.
- return u.GetFirstPage(kids.Array[0].N, kids.Array[0].Generation)
+ return u.GetFirstPage(kids.Array[0])
}
// -----------------------------------------------------------------------------
@@ -1128,7 +1545,10 @@ func Sign(document []byte, key crypto.PrivateKey, certs []*x509.Certificate,
if !ok || rootRef.Kind != Reference {
return nil, errors.New("trailer does not contain a reference to Root")
}
- root, _ := pdf.Get(rootRef.N, rootRef.Generation)
+ root, err := pdf.Dereference(rootRef)
+ if err != nil {
+ return nil, fmt.Errorf("Root dictionary retrieval failed: %s", err)
+ }
if root.Kind != Dict {
return nil, errors.New("invalid Root dictionary reference")
}
@@ -1182,7 +1602,7 @@ func Sign(document []byte, key crypto.PrivateKey, certs []*x509.Certificate,
if !ok || pagesRef.Kind != Reference {
return nil, errors.New("invalid Pages reference")
}
- page := pdf.GetFirstPage(pagesRef.N, pagesRef.Generation)
+ page := pdf.GetFirstPage(pagesRef)
if page.Kind != Dict {
return nil, errors.New("invalid or unsupported page tree")
}
@@ -1204,7 +1624,7 @@ func Sign(document []byte, key crypto.PrivateKey, certs []*x509.Certificate,
})
// 8.6.1 Interactive Form Dictionary
- if _, ok := root.Dict["AcroForm"]; ok {
+ if acroform, ok := root.Dict["AcroForm"]; ok && acroform.Kind != Nil {
return nil, errors.New("the document already contains forms, " +
"they would be overwritten")
}
diff --git a/test.sh b/test.sh
index f87d525..831a15e 100755
--- a/test.sh
+++ b/test.sh
@@ -11,11 +11,15 @@ mkdir tmp
# Create documents in various tools
log "Creating source documents"
-inkscape --pipe --export-filename=tmp/cairo.pdf <<'EOF' 2>/dev/null || :
+inkscape --pipe --export-filename=tmp/cairo.pdf --export-pdf-version=1.4 \
+<<'EOF' 2>/dev/null || :
<svg xmlns="http://www.w3.org/2000/svg"><text x="5" y="10">Hello</text></svg>
EOF
-date | tee tmp/lowriter.txt | groff -T pdf > tmp/groff.pdf || :
+date > tmp/lowriter.txt
+if command -v gropdf >/dev/null
+then groff -T pdf < tmp/lowriter.txt > tmp/groff.pdf
+fi
lowriter --convert-to pdf tmp/lowriter.txt --outdir tmp >/dev/null || :
convert rose: tmp/imagemagick.pdf || :
@@ -45,7 +49,11 @@ openssl x509 -req -in tmp/cert.csr -out tmp/cert.pem \
-CA tmp/ca.cert.pem -CAkey tmp/ca.key.pem -set_serial 1 \
-extensions smime -extfile tmp/cert.cfg 2>/dev/null
openssl verify -CAfile tmp/ca.cert.pem tmp/cert.pem >/dev/null
+
+# The second line accomodates the Go signer,
+# which doesn't support SHA-256 within pkcs12 handling
openssl pkcs12 -inkey tmp/key.pem -in tmp/cert.pem \
+ -certpbe PBE-SHA1-3DES -keypbe PBE-SHA1-3DES -macalg sha1 \
-export -passout pass: -out tmp/key-pair.p12
for tool in "$@"; do
@@ -55,6 +63,11 @@ for tool in "$@"; do
result=${source%.pdf}.signed.pdf
$tool "$source" "$result" tmp/key-pair.p12 ""
pdfsig -nssdir sql:tmp/nssdir "$result" | grep Validation
+
+ # Only some of our generators use PDF versions higher than 1.5
+ log "Testing $tool for version detection"
+ grep -q "/Version /1.6" "$result" || grep -q "^%PDF-1.6" "$result" \
+ || die "Version detection seems to misbehave (no upgrade)"
done
log "Testing $tool for expected failures"
@@ -63,11 +76,6 @@ for tool in "$@"; do
$tool -r 1 "$source" "$source.fail.pdf" tmp/key-pair.p12 "" \
&& die "Too low reservations shouldn't succeed"
- # Our generators do not use PDF versions higher than 1.5
- log "Testing $tool for version detection"
- grep -q "/Version /1.6" "$result" \
- || die "Version detection seems to misbehave (no upgrade)"
-
sed '1s/%PDF-1../%PDF-1.7/' "$source" > "$source.alt"
$tool "$source.alt" "$result.alt" tmp/key-pair.p12 ""
grep -q "/Version /1.6" "$result.alt" \