blob: cbd4bb48d1e5e1ea6f40e0860202919507a72b54 [file] [log] [blame] [edit]
// Copyright 2013 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// Functions for canonicalizing "mailto:" URLs.
#include "url/url_canon.h"
#include "url/url_canon_internal.h"
#include "url/url_file.h"
#include "url/url_parse_internal.h"
namespace url {
namespace {
// Certain characters should be percent-encoded when they appear in the path
// component of a mailto URL, to improve compatibility and mitigate against
// command-injection attacks on mailto handlers. See https://crbug.com/711020.
template <typename UCHAR>
bool ShouldEncodeMailboxCharacter(UCHAR uch) {
if (uch < 0x21 || // space & control characters.
uch > 0x7e || // high-ascii characters.
uch == 0x22 || // quote.
uch == 0x3c || uch == 0x3e || // angle brackets.
uch == 0x60 || // backtick.
uch == 0x7b || uch == 0x7c || uch == 0x7d // braces and pipe.
) {
return true;
}
return false;
}
template <typename CHAR, typename UCHAR>
bool DoCanonicalizeMailtoURL(const URLComponentSource<CHAR>& source,
const Parsed& parsed,
CanonOutput* output,
Parsed* new_parsed) {
// mailto: only uses {scheme, path, query} -- clear the rest.
new_parsed->username = Component();
new_parsed->password = Component();
new_parsed->host = Component();
new_parsed->port = Component();
new_parsed->ref = Component();
// Scheme (known, so we don't bother running it through the more
// complicated scheme canonicalizer).
new_parsed->scheme.begin = output->length();
output->Append("mailto:");
new_parsed->scheme.len = 6;
bool success = true;
// Path
if (parsed.path.is_valid()) {
new_parsed->path.begin = output->length();
// Copy the path using path URL's more lax escaping rules.
// We convert to UTF-8 and escape non-ASCII, but leave most
// ASCII characters alone.
size_t end = static_cast<size_t>(parsed.path.end());
for (size_t i = static_cast<size_t>(parsed.path.begin); i < end; ++i) {
UCHAR uch = static_cast<UCHAR>(source.path[i]);
if (ShouldEncodeMailboxCharacter<UCHAR>(uch))
success &= AppendUTF8EscapedChar(source.path, &i, end, output);
else
output->push_back(static_cast<char>(uch));
}
new_parsed->path.len = output->length() - new_parsed->path.begin;
} else {
// No path at all
new_parsed->path.reset();
}
// Query -- always use the default UTF8 charset converter.
CanonicalizeQuery(source.query, parsed.query, NULL,
output, &new_parsed->query);
return success;
}
} // namespace
bool CanonicalizeMailtoURL(const char* spec,
int spec_len,
const Parsed& parsed,
CanonOutput* output,
Parsed* new_parsed) {
return DoCanonicalizeMailtoURL<char, unsigned char>(
URLComponentSource<char>(spec), parsed, output, new_parsed);
}
bool CanonicalizeMailtoURL(const char16_t* spec,
int spec_len,
const Parsed& parsed,
CanonOutput* output,
Parsed* new_parsed) {
return DoCanonicalizeMailtoURL<char16_t, char16_t>(
URLComponentSource<char16_t>(spec), parsed, output, new_parsed);
}
bool ReplaceMailtoURL(const char* base,
const Parsed& base_parsed,
const Replacements<char>& replacements,
CanonOutput* output,
Parsed* new_parsed) {
URLComponentSource<char> source(base);
Parsed parsed(base_parsed);
SetupOverrideComponents(base, replacements, &source, &parsed);
return DoCanonicalizeMailtoURL<char, unsigned char>(
source, parsed, output, new_parsed);
}
bool ReplaceMailtoURL(const char* base,
const Parsed& base_parsed,
const Replacements<char16_t>& replacements,
CanonOutput* output,
Parsed* new_parsed) {
RawCanonOutput<1024> utf8;
URLComponentSource<char> source(base);
Parsed parsed(base_parsed);
SetupUTF16OverrideComponents(base, replacements, &utf8, &source, &parsed);
return DoCanonicalizeMailtoURL<char, unsigned char>(
source, parsed, output, new_parsed);
}
} // namespace url