static void proto_tree_add_item_ucs2string (proto_tree *tree, int hfindex, tvbuff_t *tvb, gint start )
{
guint16 ucs2_len;
gboolean dump_bytes = FALSE;
ucs2_len = tvb_get_ntohs( tvb, start );
if ( ucs2_iconv != (GIConv)-1 && ucs2_len <= MC_MAX_UCS2LEN )
{
gchar *in, *out, *out_const;
gsize in_len, out_len, rv_iconv;
in_len = ucs2_len * 2;
in = (gchar*)tvb_get_ptr(tvb, start + MC_TYPELEN_UCS2LEN, in_len);
if ( in )
{
out_len = ucs2_len*2;
out_const = ep_alloc0( (size_t)out_len + 1);
out = out_const;
}
if ( in != NULL && out != NULL )
{
rv_iconv = g_iconv( ucs2_iconv, &in, &in_len, &out, &out_len );
if ( -1 != rv_iconv )
{
proto_tree_add_bytes_format_value( tree, hfindex, tvb, start, ucs2_len*2 + MC_TYPELEN_UCS2LEN,
tvb_get_ptr( tvb, start, ucs2_len*2 + MC_TYPELEN_UCS2LEN),
"%s", out_const );
}
else
{
g_iconv( ucs2_iconv, NULL, 0, NULL, 0 );
dump_bytes = TRUE;
}
}
else
{
dump_bytes = TRUE;
}
}
else
{
dump_bytes = TRUE;
}
if ( dump_bytes )
{
proto_tree_add_item(tree, hfindex, tvb, start, MC_TYPELEN_UCS2LEN + ucs2_len * 2, FALSE);
}
}
estr_t
str_translate_char (GIConv conv, const char *keys, size_t ch_size, char *output, size_t out_size)
{
size_t left;
size_t cnv;
g_iconv (conv, NULL, NULL, NULL, NULL);
left = (ch_size == (size_t) (-1)) ? strlen (keys) : ch_size;
cnv = g_iconv (conv, (gchar **) & keys, &left, &output, &out_size);
if (cnv == (size_t) (-1))
return (errno == EINVAL) ? ESTR_PROBLEM : ESTR_FAILURE;
output[0] = '\0';
return ESTR_SUCCESS;
}
/*
* returns the number of bytes that represent the UTF8 encoding buffer
* in the original encoding that the user specified.
*
* NOTE: this is slow, but we only call this for the remainder of our
* buffer (e.g. the partial line at the end of our last chunk of read
* data). Also, this is only invoked if the file uses an encoding.
*/
static gsize
log_proto_text_server_get_raw_size_of_buffer(LogProtoTextServer *self, const guchar *buffer, gsize buffer_len)
{
gchar *out;
const guchar *in;
gsize avail_out, avail_in;
gint ret;
if (self->reverse_convert == ((GIConv) -1) && !self->convert_scale)
{
/* try to speed up raw size calculation by recognizing the most
* prominent character encodings and in the case the encoding
* uses fixed size characters set that in self->convert_scale,
* which in turn will speed up the reversal of the UTF8 buffer
* size to raw buffer sizes.
*/
self->convert_scale = log_proto_get_char_size_for_fixed_encoding(self->super.super.options->encoding);
if (self->convert_scale == 0)
{
/* this encoding is not known, do the conversion for real :( */
self->reverse_convert = g_iconv_open(self->super.super.options->encoding, "utf-8");
}
}
if (self->convert_scale)
return g_utf8_strlen((gchar *) buffer, buffer_len) * self->convert_scale;
if (self->reverse_buffer_len < buffer_len * 6)
{
/* we free and malloc, since we never need the data still in reverse buffer */
g_free(self->reverse_buffer);
self->reverse_buffer_len = buffer_len * 6;
self->reverse_buffer = g_malloc(buffer_len * 6);
}
avail_out = self->reverse_buffer_len;
out = self->reverse_buffer;
avail_in = buffer_len;
in = buffer;
ret = g_iconv(self->reverse_convert, (gchar **) &in, &avail_in, &out, &avail_out);
if (ret == (gsize) -1)
{
/* oops, we cannot reverse that we ourselves converted to UTF-8,
* this is simply impossible, but never say never */
msg_error("Internal error, couldn't reverse the internal UTF8 string to the original encoding",
evt_tag_printf("buffer", "%.*s", (gint) buffer_len, buffer),
NULL);
return 0;
}
else
{
return self->reverse_buffer_len - avail_out;
}
}
/* Returns a newly allocated gchar, converted according to the given
handler */
gchar *gnc_call_iconv(GIConv handler, const gchar* input)
{
gchar *inbuffer = (gchar*)input;
gchar *outbuffer, *outbufferstart;
gsize inbytes, outbytes;
inbytes = strlen(inbuffer);
outbytes = inbytes + 2;
outbufferstart = g_strndup(inbuffer, outbytes);
outbuffer = outbufferstart;
g_iconv(handler, &inbuffer, &inbytes, &outbuffer, &outbytes);
if (outbytes > 0)
*outbuffer = '\0';
return outbufferstart;
}
bool
ISpellChecker::checkWord(const char * const utf8Word, size_t length)
{
ichar_t iWord[INPUTWORDLEN + MAXAFFIXLEN];
char szWord[INPUTWORDLEN + MAXAFFIXLEN];
if (!m_bSuccessfulInit)
return false;
if (!utf8Word || length >= (INPUTWORDLEN + MAXAFFIXLEN) || length == 0)
return false;
bool retVal = false;
if (!g_iconv_is_valid(m_translate_in))
return false;
else
{
/* convert to 8bit string and null terminate */
size_t len_in, len_out, result;
// the 8bit encodings use precomposed forms
char *normalizedWord = g_utf8_normalize (utf8Word, length, G_NORMALIZE_NFC);
char *In = normalizedWord;
char *Out = szWord;
len_in = strlen(In);
len_out = sizeof( szWord ) - 1;
result = g_iconv(m_translate_in, &In, &len_in, &Out, &len_out);
g_free(normalizedWord);
if ((size_t)-1 == result)
return false;
*Out = '\0';
}
if (!strtoichar(iWord, szWord, sizeof(iWord), 0))
{
if (good(iWord, 0, 0, 1, 0) == 1 ||
compoundgood(iWord, 1) == 1)
{
retVal = true;
}
}
return retVal;
}
bool
MySpellChecker::checkWord(const char *utf8Word, size_t len)
{
if (len > MAXWORDLEN || !g_iconv_is_valid(m_translate_in))
return false;
// the 8bit encodings use precomposed forms
char *normalizedWord = g_utf8_normalize (utf8Word, len, G_NORMALIZE_NFC);
char *in = normalizedWord;
char word8[MAXWORDLEN + 1];
char *out = word8;
size_t len_in = strlen(in);
size_t len_out = sizeof( word8 ) - 1;
size_t result = g_iconv(m_translate_in, &in, &len_in, &out, &len_out);
g_free(normalizedWord);
if ((size_t)-1 == result)
return false;
*out = '\0';
if (myspell->spell(word8))
return true;
else
return false;
}
char* iofunctions_decode_text (gchar* text) {
GError* err = NULL;
gchar* result = 0;
gsize read = 0, written = 0;
if (! (result = g_locale_to_utf8 (text, -1, &read, &written, &err))) {
g_error_free (err);
slog (L_ERROR, "failed to convert text from default locale, trying "
"ISO-8859-1\n");
gsize in_size = strlen (text), out_size = in_size * 2;
gchar* out = (gchar*)g_malloc (out_size);
gchar* process = out;
/* TODO: replace these calls to the non-raw glib functions */
GIConv cd = g_iconv_open ("UTF-8//IGNORE", "ISO−8859-1");
if (-1 == g_iconv (cd, &text, &in_size, &process, &out_size)) {
slog (L_G_ERROR, _("Can not convert text to UTF-8!\n"));
g_free (out);
out = NULL;
}
result = out;
}
return result;
}
static gboolean
convert_text (GeditDocumentOutputStream *stream,
const gchar *inbuf,
gsize inbuf_len,
gchar **outbuf,
gsize *outbuf_len,
GError **error)
{
gchar *out, *dest;
gsize in_left, out_left, outbuf_size, res;
gint errsv;
gboolean done, have_error;
in_left = inbuf_len;
/* set an arbitrary length if inbuf_len is 0, this is needed to flush
the iconv data */
outbuf_size = (inbuf_len > 0) ? inbuf_len : 100;
out_left = outbuf_size;
out = dest = g_malloc (outbuf_size);
done = FALSE;
have_error = FALSE;
while (!done && !have_error)
{
/* If we reached here is because we need to convert the text,
so we convert it using iconv.
See that if inbuf is NULL the data will be flushed */
res = g_iconv (stream->priv->iconv,
(gchar **)&inbuf, &in_left,
&out, &out_left);
/* something went wrong */
if (res == (gsize)-1)
{
errsv = errno;
switch (errsv)
{
case EINVAL:
/* Incomplete text, do not report an error */
stream->priv->iconv_buffer = g_strndup (inbuf, in_left);
stream->priv->iconv_buflen = in_left;
done = TRUE;
break;
case E2BIG:
{
/* allocate more space */
gsize used = out - dest;
outbuf_size *= 2;
dest = g_realloc (dest, outbuf_size);
out = dest + used;
out_left = outbuf_size - used;
}
break;
case EILSEQ:
/* TODO: we should escape this text.*/
g_set_error_literal (error, G_CONVERT_ERROR,
G_CONVERT_ERROR_ILLEGAL_SEQUENCE,
_("Invalid byte sequence in conversion input"));
have_error = TRUE;
break;
default:
g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
_("Error during conversion: %s"),
g_strerror (errsv));
have_error = TRUE;
break;
}
}
else
{
done = TRUE;
}
}
if (have_error)
{
g_free (dest);
*outbuf = NULL;
*outbuf_len = 0;
return FALSE;
}
*outbuf = dest;
*outbuf_len = out - dest;
return TRUE;
}
bool Gobby::OperationOpen::on_idle()
{
static const unsigned int CONVERT_BUFFER_SIZE = 1024;
const char* inbuffer = &m_raw_content[m_raw_pos];
char* inbuf = const_cast<char*>(inbuffer);
gsize inbytes = m_raw_content.size() - m_raw_pos;
char outbuffer[CONVERT_BUFFER_SIZE];
gchar* outbuf = outbuffer;
gsize outbytes = CONVERT_BUFFER_SIZE;
/* iconv is defined as libiconv on Windows, or at least when using the
* binary packages from ftp.gnome.org. Therefore we can't propely
* call Glib::IConv::iconv. Therefore, we use the C API here. */
const std::size_t result = g_iconv(m_iconv->gobj(),
&inbuf, &inbytes, &outbuf, &outbytes);
bool more_to_process = (inbytes != 0);
if(result == static_cast<std::size_t>(-1))
{
if(errno == EILSEQ)
{
// Invalid text for the current encoding
encoding_error();
return false;
}
if(errno == EINVAL)
{
// If EINVAL is set, this means that an incomplete
// multibyte sequence was at the end of the input.
// We might have some more bytes, but those do not
// make up a whole character, so we need to wait for
// more input.
if(!m_stream)
{
// However, if we already read all input, then
// there is no more input to come. We
// consider this an error since the file
// should not end with an incomplete multibyte
// sequence.
encoding_error();
return false;
}
else
{
// Otherwise, we need to wait for more data
// to process.
more_to_process = false;
}
}
}
m_raw_pos += (inbuf - inbuffer);
// We now have outbuf - outbuffer bytes of valid UTF-8 in outbuffer.
char* prev = outbuffer;
char* pos;
const char to_find[] = { '\r', '\n', '\0' };
/* TODO: Write directly into the buffer here,
* instead of memmoving stuff. */
while( (pos = std::find_first_of<char*>(prev, outbuf,
to_find, to_find + sizeof(to_find))) != outbuf)
{
if(*pos == '\0')
{
// There is a nullbyte in the conversion. As normal
// text files don't contain nullbytes, this only
// occurs when converting for example a UTF-16 from
// ISO-8859-1 to UTF-8 (note that the UTF-16 file is
// valid ISO-8859-1, it just contains lots of
// nullbytes). We therefore produce an error here.
encoding_error();
return false;
}
else
{
// We convert everything to '\n' as line separator,
// but remember the current eol-style to correctly
// save the document back to disk.
prev = pos + 1;
if(*pos == '\r' && prev != outbuf && *prev == '\n')
{
// CRLF style line break
std::memmove(prev, prev + 1,
outbuf - prev - 1);
m_eol_style = DocumentInfoStorage::EOL_CRLF;
--outbuf;
}
else if(*pos == '\r')
{
*pos = '\n';
m_eol_style = DocumentInfoStorage::EOL_CR;
}
else
{
m_eol_style = DocumentInfoStorage::EOL_LF;
}
}
}
GtkTextIter insert_iter;
gtk_text_buffer_get_end_iter(m_content, &insert_iter);
gtk_text_buffer_insert(m_content, &insert_iter, outbuffer,
outbuf - outbuffer);
// Done reading and converting the whole file
if(!more_to_process && !m_stream)
read_finish();
return more_to_process;
}
/**
* Converts a given string using the given iconv converter. This is similar to g_convert_with_fallback, except that it is tolerant of sequences in
* the original input that are invalid even in from_encoding. g_convert_with_fallback fails for such text, whereas this function replaces such a
* sequence with the fallback string.
*
* If len is -1, strlen(text) is used to calculate the length. Do not pass -1 if text is supposed to contain \0 bytes, such as if from_encoding is a
* multi-byte encoding like UTF-16.
*/
gchar *
text_convert_invalid (const gchar* text, gssize len, GIConv converter, const gchar *fallback, gsize *len_out)
{
gchar *result_part;
gsize result_part_len;
const gchar *end;
gsize invalid_start_pos;
GString *result;
const gchar *current_start;
if (len == -1)
{
len = strlen (text);
}
end = text + len;
/* Find the first position of an invalid sequence. */
result_part = g_convert_with_iconv (text, len, converter, &invalid_start_pos, &result_part_len, NULL);
g_iconv (converter, NULL, NULL, NULL, NULL);
if (result_part != NULL)
{
/* All text converted successfully on the first try. Return it. */
if (len_out != NULL)
{
*len_out = result_part_len;
}
return result_part;
}
/* One or more invalid sequences exist that need to be replaced with the fallback. */
result = g_string_sized_new (len);
current_start = text;
for (;;)
{
g_assert (current_start + invalid_start_pos < end);
/* Convert everything before the position of the invalid sequence. It should be successful.
* But iconv may not convert everything till invalid_start_pos since the last few bytes may be part of a shift sequence.
* So get the new bytes_read and use it as the actual invalid_start_pos to handle this.
*
* See https://github.com/hexchat/hexchat/issues/1758
*/
result_part = g_convert_with_iconv (current_start, invalid_start_pos, converter, &invalid_start_pos, &result_part_len, NULL);
g_iconv (converter, NULL, NULL, NULL, NULL);
g_assert (result_part != NULL);
g_string_append_len (result, result_part, result_part_len);
g_free (result_part);
/* Append the fallback */
g_string_append (result, fallback);
/* Now try converting everything after the invalid sequence. */
current_start += invalid_start_pos + 1;
result_part = g_convert_with_iconv (current_start, end - current_start, converter, &invalid_start_pos, &result_part_len, NULL);
g_iconv (converter, NULL, NULL, NULL, NULL);
if (result_part != NULL)
{
/* The rest of the text converted successfully. Append it and return the whole converted text. */
g_string_append_len (result, result_part, result_part_len);
g_free (result_part);
if (len_out != NULL)
{
*len_out = result->len;
}
return g_string_free (result, FALSE);
}
/* The rest of the text didn't convert successfully. invalid_start_pos has the position of the next invalid sequence. */
}
}
static gboolean
log_proto_buffered_server_convert_from_raw(LogProtoBufferedServer *self, const guchar *raw_buffer, gsize raw_buffer_len)
{
/* some data was read */
gsize avail_in = raw_buffer_len;
gsize avail_out;
gchar *out;
gint ret = -1;
gboolean success = FALSE;
LogProtoBufferedServerState *state = log_proto_buffered_server_get_state(self);
do
{
avail_out = state->buffer_size - state->pending_buffer_end;
out = (gchar *) self->buffer + state->pending_buffer_end;
ret = g_iconv(self->convert, (gchar **) &raw_buffer, &avail_in, (gchar **) &out, &avail_out);
if (ret == (gsize) -1)
{
switch (errno)
{
case EINVAL:
if (self->stream_based)
{
/* Incomplete text, do not report an error, rather try to read again */
state->pending_buffer_end = state->buffer_size - avail_out;
if (avail_in > 0)
{
if (avail_in > sizeof(state->raw_buffer_leftover))
{
msg_error("Invalid byte sequence, the remaining raw buffer is larger than the supported leftover size",
evt_tag_str("encoding", self->super.options->encoding),
evt_tag_int("avail_in", avail_in),
evt_tag_int("leftover_size", sizeof(state->raw_buffer_leftover)));
goto error;
}
memcpy(state->raw_buffer_leftover, raw_buffer, avail_in);
state->raw_buffer_leftover_size = avail_in;
state->raw_buffer_size -= avail_in;
msg_trace("Leftover characters remained after conversion, delaying message until another chunk arrives",
evt_tag_str("encoding", self->super.options->encoding),
evt_tag_int("avail_in", avail_in));
goto success;
}
}
else
{
msg_error("Byte sequence too short, cannot convert an individual frame in its entirety",
evt_tag_str("encoding", self->super.options->encoding),
evt_tag_int("avail_in", avail_in));
goto error;
}
break;
case E2BIG:
state->pending_buffer_end = state->buffer_size - avail_out;
/* extend the buffer */
if (state->buffer_size < self->super.options->max_buffer_size)
{
state->buffer_size *= 2;
if (state->buffer_size > self->super.options->max_buffer_size)
state->buffer_size = self->super.options->max_buffer_size;
self->buffer = g_realloc(self->buffer, state->buffer_size);
/* recalculate the out pointer, and add what we have now */
ret = -1;
}
else
{
msg_error("Incoming byte stream requires a too large conversion buffer, probably invalid character sequence",
evt_tag_str("encoding", self->super.options->encoding),
evt_tag_printf("buffer", "%.*s", (gint) state->pending_buffer_end, self->buffer));
goto error;
}
break;
case EILSEQ:
default:
msg_notice("Invalid byte sequence or other error while converting input, skipping character",
evt_tag_str("encoding", self->super.options->encoding),
evt_tag_printf("char", "0x%02x", *(guchar *) raw_buffer));
goto error;
}
}
else
{
state->pending_buffer_end = state->buffer_size - avail_out;
}
}
while (avail_in > 0);
success:
success = TRUE;
error:
log_proto_buffered_server_put_state(self);
return success;
}
char *str_convert_encoding(int from, int to, const char *str)
{
/*
* And here it should just be a matter of calling glib's g_convert().
* Or so I thought. Alas, they chickened out of the hard part: how to
* figure out the size of a zero-terminaded string in any arbitrary
* encoding.
* Since there is no advantage in using g_convert(), might as well
* keep my old implementation. It's worth using their iconv wrapppers
* though, because they provide libiconv in systems that don't have it
* natively.
*/
GIConv conv;
char *result;
char *inbuf, *outbuf;
size_t inbpc, outbpc;
size_t inbytes, outbytes;
size_t inbytesleft, outbytesleft;
size_t res;
inbytes = strsize(from, str);
if (strcasecmp(encoding_name(to), encoding_name(from)) == 0) {
result = malloc(inbytes);
memcpy(result, str, inbytes);
return result;
}
conv = g_iconv_open(encoding_name(to), encoding_name(from));
if (conv == (GIConv)-1) {
fprintf(stderr, "convert_encoding: cannot convert from %s to %s\n",
encoding_name(from), encoding_name(to));
return NULL;
}
inbpc = bytes_per_char(from);
outbpc = bytes_per_char(to);
/* estimate the converted size */
outbytes = ((double)outbpc / (double)inbpc) * inbytes;
/* optimize common cases (tuned for western european languages) */
if (to == UTF_8 && inbpc == 1)
outbytes = ceil(1.25 * inbytes);
else if (to == UTF_16)
outbytes += 2; /* for the BOM */
//printf("inbytes : %i\noutbytes: %i\n", inbytes, outbytes);
result = malloc(outbytes);
inbuf = (char*)str;
inbytesleft = inbytes;
outbuf = result;
outbytesleft = outbytes;
while(1) {
res = g_iconv(conv, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
if (res == (size_t)-1) {
if (errno == E2BIG) {
/* Ran out of space, alloc more
This code tries hard to avoid the need for a second realloc,
while still keeping over-allocation to a minimum */
double done = 1.0 - (double)inbytesleft / (double)inbytes;
size_t bytes_written = outbuf - result;
size_t newsize = ceil((bytes_written / done) * 1.1);
//printf("growing: done=%g%%, old size=%i, new size=%i\n",
// 100.0*done, outbytes, newsize);
outbytesleft += newsize - outbytes;
outbytes = newsize;
result = realloc(result, outbytes);
outbuf = result + bytes_written;
continue;
}
else {
/* Invalid or inconvertible char, skip it
Seems better than aborting the conversion... */
fprintf(stderr, "convert_encoding: conversion error at offset %i\n",
inbytes-inbytesleft);
inbuf += inbpc;
inbytesleft = max(inbytesleft - inbpc, 0);
outbuf += outbpc;
outbytesleft = max(outbytesleft - outbpc, 0);
continue;
}
}
break;
}
//printf("%i of %i bytes unused (wasted %g%%)\n",
// outbytesleft, outbytes, 100.0*(double)outbytesleft/(double)outbytes);
g_iconv_close(conv);
return result;
}
//!
//! @brief Copies a file and creates a new one using the new encoding
//! @param SOURCE_PATH The source file to change the encoding on.
//! @param TARGET_PATH The place to save the new file with the new encoding.
//! @param SOURCE_ENCODING The encoding of the source file.
//! @param TARGET_ENCODING THe wanted encoding in the new file to be created.
//! @param cb A LwIoProgressCallback to use to give progress feedback or NULL
//! @param data A gpointer to data to pass to the LwIoProgressCallback
//! @param error pointer to a GError to write errors to
//! @return The status of the conversion opertaion
//!
gboolean
lw_io_copy_with_encoding (const gchar *SOURCE_PATH,
const gchar *TARGET_PATH,
const gchar *SOURCE_ENCODING,
const gchar *TARGET_ENCODING,
LwIoProgressCallback cb,
gpointer data,
GCancellable *cancellable,
GError **error)
{
if (*error != NULL) return FALSE;
//Declarations
FILE* readfd = NULL;
FILE* writefd = NULL;
const gint MAX = 1024 * 2;
gchar source_buffer[MAX];
gchar target_buffer[MAX];
gchar *sptr, *tptr;
size_t read, source_bytes_left, target_bytes_left;
gdouble fraction;
size_t position, filesize;
GIConv conv;
filesize = lw_io_get_filesize (SOURCE_PATH);
position = 0;
//Initializations
readfd = fopen (SOURCE_PATH, "rb");
writefd = fopen (TARGET_PATH, "wb");
conv = g_iconv_open (TARGET_ENCODING, SOURCE_ENCODING);
//Read a chunk
while (ferror(readfd) == 0 && feof(readfd) == 0)
{
read = fread(source_buffer, sizeof(gchar), MAX, readfd);
source_bytes_left = read;
sptr = source_buffer;
//Try to convert and write the chunk
while (source_bytes_left > 0 && ferror(writefd) == 0 && feof(writefd) == 0)
{
target_bytes_left = MAX;
tptr = target_buffer;
g_iconv (conv, &sptr, &source_bytes_left, &tptr, &target_bytes_left);
if (MAX != target_bytes_left) //Bytes were converted!
{
fwrite(target_buffer, sizeof(gchar), MAX - target_bytes_left, writefd);
}
else if (source_bytes_left == MAX && target_bytes_left == MAX)
{
fprintf(stderr, "The file you are converting may be corrupt! Trying to skip a character...\n");
fseek(readfd, 1L - source_bytes_left, SEEK_CUR);
}
else if (source_bytes_left > 0) //Bytes failed to convert!
{
fseek(readfd, -source_bytes_left, SEEK_CUR);
source_bytes_left = 0;
}
}
position = ftell(readfd);
fraction = (gdouble) position / (gdouble) filesize;
if (cb != NULL) cb (fraction, data);
}
//Cleanup
g_iconv_close (conv);
fclose(readfd);
fclose(writefd);
return TRUE;
}
gchar*
g_convert_with_iconv (const gchar *str,
gssize len,
GIConv converter,
gsize *bytes_read,
gsize *bytes_written,
GError **error)
{
gchar *dest;
gchar *outp;
const gchar *p;
gsize inbytes_remaining;
gsize outbytes_remaining;
gsize err;
gsize outbuf_size;
gboolean have_error = FALSE;
g_return_val_if_fail (str != NULL, NULL);
g_return_val_if_fail (converter != (GIConv) -1, NULL);
if (len < 0)
len = strlen (str);
p = str;
inbytes_remaining = len;
outbuf_size = len + 1; /* + 1 for nul in case len == 1 */
outbytes_remaining = outbuf_size - 1; /* -1 for nul */
outp = dest = g_malloc (outbuf_size);
again:
err = g_iconv (converter, (char **)&p, &inbytes_remaining, &outp, &outbytes_remaining);
if (err == (size_t) -1)
{
switch (errno)
{
case EINVAL:
/* Incomplete text, do not report an error */
break;
case E2BIG:
{
size_t used = outp - dest;
outbuf_size *= 2;
dest = g_realloc (dest, outbuf_size);
outp = dest + used;
outbytes_remaining = outbuf_size - used - 1; /* -1 for nul */
goto again;
}
case EILSEQ:
g_warning("Invalid byte sequence in conversion input");
have_error = TRUE;
break;
default:
g_warning("Error during conversion: %s", strerror (errno));
have_error = TRUE;
break;
}
}
*outp = '\0';
if (bytes_read)
*bytes_read = p - str;
else
{
if ((p - str) != len)
{
if (!have_error)
{
g_warning("Partial character sequence at end of input");
have_error = TRUE;
}
}
}
if (bytes_written)
*bytes_written = outp - dest; /* Doesn't include '\0' */
if (have_error)
{
g_free (dest);
return NULL;
}
else
return dest;
}
static estr_t
_str_convert (GIConv coder, const char *string, int size, GString * buffer)
{
estr_t state = ESTR_SUCCESS;
gchar *tmp_buff = NULL;
gssize left;
gsize bytes_read = 0;
gsize bytes_written = 0;
GError *error = NULL;
errno = 0;
if (coder == INVALID_CONV)
return ESTR_FAILURE;
if (string == NULL || buffer == NULL)
return ESTR_FAILURE;
/*
if (! used_class.is_valid_string (string))
{
return ESTR_FAILURE;
}
*/
if (size < 0)
{
size = strlen (string);
}
else
{
left = strlen (string);
if (left < size)
size = left;
}
left = size;
g_iconv (coder, NULL, NULL, NULL, NULL);
while (left)
{
tmp_buff = g_convert_with_iconv ((const gchar *) string,
left, coder, &bytes_read, &bytes_written, &error);
if (error)
{
int code = error->code;
g_error_free (error);
error = NULL;
switch (code)
{
case G_CONVERT_ERROR_NO_CONVERSION:
/* Conversion between the requested character sets is not supported. */
tmp_buff = g_strnfill (strlen (string), '?');
g_string_append (buffer, tmp_buff);
g_free (tmp_buff);
return ESTR_FAILURE;
case G_CONVERT_ERROR_ILLEGAL_SEQUENCE:
/* Invalid byte sequence in conversion input. */
if ((tmp_buff == NULL) && (bytes_read != 0))
/* recode valid byte sequence */
tmp_buff = g_convert_with_iconv ((const gchar *) string,
bytes_read, coder, NULL, NULL, NULL);
if (tmp_buff != NULL)
{
g_string_append (buffer, tmp_buff);
g_free (tmp_buff);
}
if ((int) bytes_read < left)
{
string += bytes_read + 1;
size -= (bytes_read + 1);
left -= (bytes_read + 1);
g_string_append_c (buffer, *(string - 1));
}
else
{
return ESTR_PROBLEM;
}
state = ESTR_PROBLEM;
break;
case G_CONVERT_ERROR_PARTIAL_INPUT:
/* Partial character sequence at end of input. */
g_string_append (buffer, tmp_buff);
g_free (tmp_buff);
if ((int) bytes_read < left)
{
left = left - bytes_read;
tmp_buff = g_strnfill (left, '?');
g_string_append (buffer, tmp_buff);
g_free (tmp_buff);
}
return ESTR_PROBLEM;
case G_CONVERT_ERROR_BAD_URI: /* Don't know how handle this error :( */
case G_CONVERT_ERROR_NOT_ABSOLUTE_PATH: /* Don't know how handle this error :( */
case G_CONVERT_ERROR_FAILED: /* Conversion failed for some reason. */
default:
g_free (tmp_buff);
return ESTR_FAILURE;
}
}
else
{
if (tmp_buff != NULL)
{
if (*tmp_buff)
{
g_string_append (buffer, tmp_buff);
g_free (tmp_buff);
string += bytes_read;
left -= bytes_read;
}
else
{
g_free (tmp_buff);
g_string_append (buffer, string);
return state;
}
}
else
{
g_string_append (buffer, string);
return ESTR_PROBLEM;
}
}
}
return state;
}
ssize_t
TextStream::Read (char *buf, size_t n)
{
size_t inleft = buflen;
char *inbuf = bufptr;
char *outbuf = buf;
size_t outleft = n;
ssize_t nread;
size_t r;
do {
if (cd != (GIConv) -1) {
if (g_iconv (cd, &inbuf, &inleft, &outbuf, &outleft) == (size_t) -1) {
switch (errno) {
case E2BIG:
// not enough space available in the output buffer
goto out;
case EINVAL:
// incomplete multibyte character sequence
goto out;
case EILSEQ:
// illegal multibyte sequence
return -1;
default:
// unknown error, fail
return -1;
}
}
} else {
r = MIN (inleft, outleft);
memcpy (outbuf, inbuf, r);
outleft -= r;
outbuf += r;
inleft -= r;
inbuf += r;
}
if (outleft == 0 || eof)
break;
// buffer more data
if (inleft > 0)
memmove (buffer, inbuf, inleft);
inbuf = buffer + inleft;
if ((nread = ReadInternal (inbuf, sizeof (buffer) - inleft)) <= 0) {
eof = true;
break;
}
inleft += nread;
inbuf = buffer;
} while (true);
if (eof && cd != (GIConv) -1)
g_iconv (cd, NULL, NULL, &outbuf, &outleft);
out:
buflen = inleft;
bufptr = inbuf;
return (outbuf - buf);
}
void Gobby::OperationSave::write_next()
{
gchar* inbuf;
gsize inlen;
char newlinebuf[2] = { '\r', '\n' };
if(m_current_line_index < m_current_line->second)
{
inbuf = m_current_line->first + m_current_line_index;
inlen = m_current_line->second - m_current_line_index;
}
else
{
// Write newline
switch(m_eol_style)
{
case DocumentInfoStorage::EOL_CR:
inbuf = newlinebuf + 0;
inlen = 1;
break;
case DocumentInfoStorage::EOL_LF:
inbuf = newlinebuf + 1;
inlen = 1;
break;
case DocumentInfoStorage::EOL_CRLF:
inbuf = newlinebuf + 0;
inlen = 2;
break;
default:
g_assert_not_reached();
break;
}
}
gchar* outbuf = m_buffer;
gsize outlen = BUFFER_SIZE;
gchar* preserve_inbuf = inbuf;
/* iconv is defined as libiconv on Windows, or at least when using the
* binary packages from ftp.gnome.org. Therefore we can't properly
* call Glib::IConv::iconv. Therefore, we use the C API here. */
std::size_t retval = g_iconv(
m_iconv.gobj(), &inbuf, &inlen, &outbuf, &outlen);
if(retval == static_cast<std::size_t>(-1))
{
g_assert(errno != EILSEQ);
// E2BIG and EINVAL are fully OK here.
}
else if(retval > 0)
{
error(_("The document contains one or more characters that "
"cannot be encoded in the specified character "
"coding."));
return;
}
// Advance bytes read.
m_current_line_index += inbuf - preserve_inbuf;
m_buffer_size = BUFFER_SIZE - outlen;
m_buffer_index = 0;
g_assert(m_buffer_size > 0);
if(m_current_line_index > m_current_line->second)
{
// Converted whole line:
g_free(m_current_line->first);
m_current_line = m_lines.erase(m_current_line);
m_current_line_index = 0;
}
m_stream->write_async(m_buffer, m_buffer_size,
sigc::mem_fun(*this,
&OperationSave::on_stream_write));
}
