unicode: inject invalid character on encode failure

...rather than throw. This mimics Windows behaviour.

Signed-off-by: Simon Rozman <simon@rozman.si>
This commit is contained in:
Simon Rozman 2025-06-05 12:06:07 +02:00
parent f454ae9cab
commit 2bfe310d47
2 changed files with 37 additions and 22 deletions

View File

@ -64,26 +64,33 @@ namespace UnitTests
void unicode::charset_encoder() void unicode::charset_encoder()
{ {
stdex::charset_encoder<char, char> win1250_to_utf8(stdex::charset_id::windows1250, stdex::charset_id::utf8); {
stdex::charset_encoder<char, char> win1250_to_utf8(stdex::charset_id::windows1250, stdex::charset_id::utf8);
Assert::AreEqual( Assert::AreEqual(
"This is a test.", "This is a test.",
win1250_to_utf8.convert("This is a test.").c_str()); win1250_to_utf8.convert("This is a test.").c_str());
Assert::AreEqual( Assert::AreEqual(
"Thíš i· a teşt.", "Thíš i· a teşt.",
win1250_to_utf8.convert("Th\xed\x9a i\xb7 a te\xbat.").c_str()); win1250_to_utf8.convert("Th\xed\x9a i\xb7 a te\xbat.").c_str());
std::string src, dst; std::string src, dst;
for (size_t i = 0; i < 1000; i++) { for (size_t i = 0; i < 1000; i++) {
src += "V ko\x9eu\x9a\xe8ku zlobnega mizarja stopiclja fant in kli\xe8" "e 0123456789.\r\n"; src += "V ko\x9eu\x9a\xe8ku zlobnega mizarja stopiclja fant in kli\xe8" "e 0123456789.\r\n";
dst += "V kožuščku zlobnega mizarja stopiclja fant in kliče 0123456789.\r\n"; dst += "V kožuščku zlobnega mizarja stopiclja fant in kliče 0123456789.\r\n";
}
Assert::AreEqual(dst.c_str(), win1250_to_utf8.convert(src).c_str());
Assert::AreEqual(
"",
win1250_to_utf8.convert("test", 0).c_str());
Assert::AreEqual(
"",
win1250_to_utf8.convert(nullptr, 0).c_str());
}
{
stdex::charset_encoder<char16_t, char> encode(stdex::charset_id::utf16, stdex::charset_id::ascii, '?');
Assert::AreEqual("Te?t.", encode.convert(u"Tešt.").c_str());
} }
Assert::AreEqual(dst.c_str(), win1250_to_utf8.convert(src).c_str());
Assert::AreEqual(
"",
win1250_to_utf8.convert("test", 0).c_str());
Assert::AreEqual(
"",
win1250_to_utf8.convert(nullptr, 0).c_str());
} }
void unicode::normalize() void unicode::normalize()

View File

@ -1,4 +1,4 @@
/* /*
SPDX-License-Identifier: MIT SPDX-License-Identifier: MIT
Copyright © 2023-2025 Amebis Copyright © 2023-2025 Amebis
*/ */
@ -153,11 +153,13 @@ namespace stdex
{ {
protected: protected:
charset_id m_from, m_to; charset_id m_from, m_to;
T_to m_invalid;
public: public:
charset_encoder(_In_ charset_id from, _In_ charset_id to) : charset_encoder(_In_ charset_id from, _In_ charset_id to, _In_ T_to invalid = '?') :
m_from(from), m_from(from),
m_to(to) m_to(to),
m_invalid(invalid)
{ {
#ifdef _WIN32 #ifdef _WIN32
m_from_wincp = to_encoding(from); m_from_wincp = to_encoding(from);
@ -198,7 +200,6 @@ namespace stdex
#ifdef _WIN32 #ifdef _WIN32
constexpr DWORD dwFlagsWCMB = 0; constexpr DWORD dwFlagsWCMB = 0;
constexpr LPCCH lpDefaultChar = NULL;
stdex_assert(src); stdex_assert(src);
if (m_from_wincp == m_to_wincp) _Unlikely_{ if (m_from_wincp == m_to_wincp) _Unlikely_{
@ -235,6 +236,7 @@ namespace stdex
#pragma warning(suppress: 4127) // Can't use precompiler #if on template arguments, using "if" makes MSVC warnings. #pragma warning(suppress: 4127) // Can't use precompiler #if on template arguments, using "if" makes MSVC warnings.
if constexpr (sizeof(T_from) == sizeof(wchar_t) && sizeof(T_to) == sizeof(char)) { if constexpr (sizeof(T_from) == sizeof(wchar_t) && sizeof(T_to) == sizeof(char)) {
stdex_assert(count_src < INT_MAX || count_src == SIZE_MAX); stdex_assert(count_src < INT_MAX || count_src == SIZE_MAX);
LPCCH lpDefaultChar = m_to_wincp == charset_id::utf8 || m_to_wincp == charset_id::utf7 ? NULL : &m_invalid;
// Try to convert to stack buffer first. // Try to convert to stack buffer first.
CHAR szStackBuffer[1024 / sizeof(CHAR)]; CHAR szStackBuffer[1024 / sizeof(CHAR)];
@ -261,6 +263,7 @@ namespace stdex
#pragma warning(suppress: 4127) // Can't use precompiler #if on template arguments, using "if" makes MSVC warnings. #pragma warning(suppress: 4127) // Can't use precompiler #if on template arguments, using "if" makes MSVC warnings.
if constexpr (sizeof(T_from) == sizeof(char) && sizeof(T_to) == sizeof(char)) { if constexpr (sizeof(T_from) == sizeof(char) && sizeof(T_to) == sizeof(char)) {
stdex_assert(count_src < INT_MAX || count_src == SIZE_MAX); stdex_assert(count_src < INT_MAX || count_src == SIZE_MAX);
LPCCH lpDefaultChar = m_to_wincp == charset_id::utf8 || m_to_wincp == charset_id::utf7 ? NULL : &m_invalid;
// Try to convert to stack buffer first. // Try to convert to stack buffer first.
DWORD dwResult; DWORD dwResult;
@ -325,6 +328,11 @@ namespace stdex
break; break;
if (errno == E2BIG) if (errno == E2BIG)
continue; continue;
if (errno == EILSEQ) {
dst.append(1, m_invalid);
++src; src_size -= sizeof(T_from);
continue;
}
throw std::system_error(errno, std::system_category(), "iconv failed"); throw std::system_error(errno, std::system_category(), "iconv failed");
} }
#endif #endif