X-Git-Url: https://git.tcpdump.org/libpcap/blobdiff_plain/8a2938e3f128ebcece251889c3e8da09fe139ddb..HEAD:/fmtutils.c diff --git a/fmtutils.c b/fmtutils.c index f1a89073..a5a4fe62 100644 --- a/fmtutils.c +++ b/fmtutils.c @@ -35,9 +35,7 @@ * Utilities for message formatting used both by libpcap and rpcapd. */ -#ifdef HAVE_CONFIG_H #include -#endif #include "ftmacros.h" @@ -47,33 +45,244 @@ #include #include -#include +#include "pcap-int.h" #include "portability.h" #include "fmtutils.h" +#ifdef _WIN32 +#include "charconv.h" +#endif + +/* + * Set the encoding. + */ +#ifdef _WIN32 +/* + * True if we should use UTF-8. + */ +static int use_utf_8; + +void +pcapint_fmt_set_encoding(unsigned int opts) +{ + if (opts == PCAP_CHAR_ENC_UTF_8) + use_utf_8 = 1; +} +#else +void +pcapint_fmt_set_encoding(unsigned int opts _U_) +{ + /* + * Nothing to do here. + */ +} +#endif + +#ifdef _WIN32 +/* + * Convert a null-terminated UTF-16LE string to UTF-8, putting it into + * a buffer starting at the specified location and stopping if we go + * past the specified size. This will only put out complete UTF-8 + * sequences. + * + * We do this ourselves because Microsoft doesn't offer a "convert and + * stop at a UTF-8 character boundary if we run out of space" routine. + */ +#define IS_LEADING_SURROGATE(c) \ + ((c) >= 0xd800 && (c) < 0xdc00) +#define IS_TRAILING_SURROGATE(c) \ + ((c) >= 0xdc00 && (c) < 0xe000) +#define SURROGATE_VALUE(leading, trailing) \ + (((((leading) - 0xd800) << 10) | ((trailing) - 0xdc00)) + 0x10000) +#define REPLACEMENT_CHARACTER 0x0FFFD + +static char * +utf_16le_to_utf_8_truncated(const wchar_t *utf_16, char *utf_8, + size_t utf_8_len) +{ + wchar_t c, c2; + uint32_t uc; + + if (utf_8_len == 0) { + /* + * Not even enough room for a trailing '\0'. + * Don't put anything into the buffer. + */ + return (utf_8); + } + + while ((c = *utf_16++) != '\0') { + if (IS_LEADING_SURROGATE(c)) { + /* + * Leading surrogate. Must be followed by + * a trailing surrogate. + */ + c2 = *utf_16; + if (c2 == '\0') { + /* + * Oops, string ends with a lead + * surrogate. Try to drop in + * a REPLACEMENT CHARACTER, and + * don't move the string pointer, + * so on the next trip through + * the loop we grab the terminating + * '\0' and quit. + */ + uc = REPLACEMENT_CHARACTER; + } else { + /* + * OK, we can consume this 2-octet + * value. + */ + utf_16++; + if (IS_TRAILING_SURROGATE(c2)) { + /* + * Trailing surrogate. + * This calculation will, + * for c being a leading + * surrogate and c2 being + * a trailing surrogate, + * produce a value between + * 0x100000 and 0x10ffff, + * so it's always going to be + * a valid Unicode code point. + */ + uc = SURROGATE_VALUE(c, c2); + } else { + /* + * Not a trailing surrogate; + * try to drop in a + * REPLACEMENT CHARACTER. + */ + uc = REPLACEMENT_CHARACTER; + } + } + } else { + /* + * Not a leading surrogate. + */ + if (IS_TRAILING_SURROGATE(c)) { + /* + * Trailing surrogate without + * a preceding leading surrogate. + * Try to drop in a REPLACEMENT + * CHARACTER. + */ + uc = REPLACEMENT_CHARACTER; + } else { + /* + * This is a valid BMP character; + * drop it in. + */ + uc = c; + } + } + + /* + * OK, uc is a valid Unicode character; how + * many bytes worth of UTF-8 does it require? + */ + if (uc < 0x0080) { + /* 1 byte. */ + if (utf_8_len < 2) { + /* + * Not enough room for that byte + * plus a trailing '\0'. + */ + break; + } + *utf_8++ = (char)uc; + utf_8_len--; + } else if (uc < 0x0800) { + /* 2 bytes. */ + if (utf_8_len < 3) { + /* + * Not enough room for those bytes + * plus a trailing '\0'. + */ + break; + } + *utf_8++ = ((uc >> 6) & 0x3F) | 0xC0; + *utf_8++ = ((uc >> 0) & 0x3F) | 0x80; + utf_8_len -= 2; + } else if (uc < 0x010000) { + /* 3 bytes. */ + if (utf_8_len < 4) { + /* + * Not enough room for those bytes + * plus a trailing '\0'. + */ + break; + } + *utf_8++ = ((uc >> 12) & 0x0F) | 0xE0; + *utf_8++ = ((uc >> 6) & 0x3F) | 0x80; + *utf_8++ = ((uc >> 0) & 0x3F) | 0x80; + utf_8_len -= 3; + } else { + /* 4 bytes. */ + if (utf_8_len < 5) { + /* + * Not enough room for those bytes + * plus a trailing '\0'. + */ + break; + } + *utf_8++ = ((uc >> 18) & 0x03) | 0xF0; + *utf_8++ = ((uc >> 12) & 0x3F) | 0x80; + *utf_8++ = ((uc >> 6) & 0x3F) | 0x80; + *utf_8++ = ((uc >> 0) & 0x3F) | 0x80; + utf_8_len -= 3; + } + } + + /* + * OK, we have enough room for (at least) a trailing '\0'. + * (We started out with enough room, thanks to the test + * for a zero-length buffer at the beginning, and if + * there wasn't enough room for any character we wanted + * to put into the buffer *plus* a trailing '\0', + * we'd have quit before putting it into the buffer, + * and thus would have left enough room for the trailing + * '\0'.) + * + * Drop it in. + */ + *utf_8 = '\0'; + + /* + * Return a pointer to the terminating '\0', in case we + * want to drop something in after that. + */ + return (utf_8); +} +#endif /* _WIN32 */ + /* * Generate an error message based on a format, arguments, and an * errno, with a message for the errno after the formatted output. */ void -pcap_fmt_errmsg_for_errno(char *errbuf, size_t errbuflen, int errnum, +pcapint_fmt_errmsg_for_errno(char *errbuf, size_t errbuflen, int errnum, const char *fmt, ...) { va_list ap; + + va_start(ap, fmt); + pcapint_vfmt_errmsg_for_errno(errbuf, errbuflen, errnum, fmt, ap); + va_end(ap); +} + +void +pcapint_vfmt_errmsg_for_errno(char *errbuf, size_t errbuflen, int errnum, + const char *fmt, va_list ap) +{ size_t msglen; char *p; size_t errbuflen_remaining; -#if defined(HAVE_STRERROR_S) - errno_t err; -#elif defined(HAVE_STRERROR_R) - int err; -#endif - va_start(ap, fmt); - pcap_vsnprintf(errbuf, errbuflen, fmt, ap); - va_end(ap); + (void)vsnprintf(errbuf, errbuflen, fmt, ap); msglen = strlen(errbuf); /* @@ -89,43 +298,144 @@ pcap_fmt_errmsg_for_errno(char *errbuf, size_t errbuflen, int errnum, *p++ = ':'; *p++ = ' '; *p = '\0'; - msglen += 2; errbuflen_remaining -= 2; /* * Now append the string for the error code. */ -#if defined(HAVE_STRERROR_S) - err = strerror_s(p, errbuflen_remaining, errnum); +#if defined(HAVE__WCSERROR_S) + /* + * We have a Windows-style _wcserror_s(). + * Generate a UTF-16LE error message. + */ + wchar_t utf_16_errbuf[PCAP_ERRBUF_SIZE]; + errno_t err = _wcserror_s(utf_16_errbuf, PCAP_ERRBUF_SIZE, errnum); if (err != 0) { /* * It doesn't appear to be documented anywhere obvious - * what the error returns from strerror_s(). + * what the error returns from _wcserror_s(). */ - pcap_snprintf(p, errbuflen_remaining, "Error %d", errnum); + snprintf(p, errbuflen_remaining, "Error %d", errnum); + return; } -#elif defined(HAVE_STRERROR_R) - err = strerror_r(errnum, p, errbuflen_remaining); - if (err == EINVAL) { - /* - * UNIX 03 says this isn't guaranteed to produce a - * fallback error message. - */ - pcap_snprintf(p, errbuflen_remaining, "Unknown error: %d", - errnum); - } else if (err == ERANGE) { + + /* + * Now convert it from UTF-16LE to UTF-8, dropping it in the + * remaining space in the buffer, and truncating it - cleanly, + * on a UTF-8 character boundary - if it doesn't fit. + */ + utf_16le_to_utf_8_truncated(utf_16_errbuf, p, errbuflen_remaining); + + /* + * Now, if we're not in UTF-8 mode, convert errbuf to the + * local code page. + */ + if (!use_utf_8) + utf_8_to_acp_truncated(errbuf); +#else + /* + * Either Windows without _wcserror_s() or not Windows. Let pcap_strerror() + * solve the non-UTF-16 part of this problem space. + */ + snprintf(p, errbuflen_remaining, "%s", pcap_strerror(errnum)); +#endif +} + +#ifdef _WIN32 +/* + * Generate an error message based on a format, arguments, and a + * Win32 error, with a message for the Win32 error after the formatted output. + */ +void +pcapint_fmt_errmsg_for_win32_err(char *errbuf, size_t errbuflen, DWORD errnum, + const char *fmt, ...) +{ + va_list ap; + + va_start(ap, fmt); + pcapint_vfmt_errmsg_for_win32_err(errbuf, errbuflen, errnum, fmt, ap); + va_end(ap); +} + +void +pcapint_vfmt_errmsg_for_win32_err(char *errbuf, size_t errbuflen, DWORD errnum, + const char *fmt, va_list ap) +{ + size_t msglen; + char *p; + size_t errbuflen_remaining; + DWORD retval; + wchar_t utf_16_errbuf[PCAP_ERRBUF_SIZE]; + size_t utf_8_len; + + vsnprintf(errbuf, errbuflen, fmt, ap); + msglen = strlen(errbuf); + + /* + * Do we have enough space to append ": "? + * Including the terminating '\0', that's 3 bytes. + */ + if (msglen + 3 > errbuflen) { + /* No - just give them what we've produced. */ + return; + } + p = errbuf + msglen; + errbuflen_remaining = errbuflen - msglen; + *p++ = ':'; + *p++ = ' '; + *p = '\0'; + msglen += 2; + errbuflen_remaining -= 2; + + /* + * Now append the string for the error code. + * + * XXX - what language ID to use? + * + * For UN*Xes, pcap_strerror() may or may not return localized + * strings. + * + * We currently don't have localized messages for libpcap, but + * we might want to do so. On the other hand, if most of these + * messages are going to be read by libpcap developers and + * perhaps by developers of libpcap-based applications, English + * might be a better choice, so the developer doesn't have to + * get the message translated if it's in a language they don't + * happen to understand. + */ + retval = FormatMessageW(FORMAT_MESSAGE_FROM_SYSTEM|FORMAT_MESSAGE_IGNORE_INSERTS|FORMAT_MESSAGE_MAX_WIDTH_MASK, + NULL, errnum, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), + utf_16_errbuf, PCAP_ERRBUF_SIZE, NULL); + if (retval == 0) { /* - * UNIX 03 says this isn't guaranteed to produce a - * fallback error message. + * Failed. */ - pcap_snprintf(p, errbuflen_remaining, - "Message for error %d is too long", errnum); + snprintf(p, errbuflen_remaining, + "Couldn't get error message for error (%lu)", errnum); + return; } -#else + /* - * We have neither strerror_s() nor strerror_r(), so we're - * stuck with using pcap_strerror(). + * Now convert it from UTF-16LE to UTF-8. */ - pcap_snprintf(p, errbuflen_remaining, "%s", pcap_strerror(errnum)); -#endif + p = utf_16le_to_utf_8_truncated(utf_16_errbuf, p, errbuflen_remaining); + + /* + * Now append the error number, if it fits. + */ + utf_8_len = p - errbuf; + errbuflen_remaining -= utf_8_len; + if (utf_8_len == 0) { + /* The message was empty. */ + snprintf(p, errbuflen_remaining, "(%lu)", errnum); + } else + snprintf(p, errbuflen_remaining, " (%lu)", errnum); + + /* + * Now, if we're not in UTF-8 mode, convert errbuf to the + * local code page. + */ + if (!use_utf_8) + utf_8_to_acp_truncated(errbuf); } +#endif