* Boston, MA 02111-1307, USA.
*/
+#include <config.h>
+
#include <iconv.h>
#include <errno.h>
+#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include "glib.h"
-#include "config.h"
#ifdef G_PLATFORM_WIN32
#define STRICT
#include "glibintl.h"
+#if defined(USE_LIBICONV_GNU) && !defined (_LIBICONV_H)
+#error GNU libiconv in use but included iconv.h not from libiconv
+#endif
+#if !defined(USE_LIBICONV_GNU) && defined (_LIBICONV_H)
+#error GNU libiconv not in use but included iconv.h is from libiconv
+#endif
+
GQuark
-g_convert_error_quark()
+g_convert_error_quark (void)
{
static GQuark quark;
if (!quark)
return quark;
}
-#if defined(USE_LIBICONV) && !defined (_LIBICONV_H)
-#error libiconv in use but included iconv.h not from libiconv
-#endif
-#if !defined(USE_LIBICONV) && defined (_LIBICONV_H)
-#error libiconv not in use but included iconv.h is from libiconv
-#endif
-
static gboolean
try_conversion (const char *to_codeset,
const char *from_codeset,
* @to_codeset: destination codeset
* @from_codeset: source codeset
*
- * Same as the standard Unix routine <function>iconv_open()</function>, but
- * may be implemented via libiconv on Unix flavors that lack
+ * Same as the standard UNIX routine <function>iconv_open()</function>, but
+ * may be implemented via libiconv on UNIX flavors that lack
* a native implementation.
*
- * &GLib; provides g_convert() and g_locale_to_utf8() which are likely
+ * GLib provides g_convert() and g_locale_to_utf8() which are likely
* more convenient than the raw iconv wrappers.
*
* Return value: a "conversion descriptor"
* @outbuf: converted output bytes
* @outbytes_left: inout parameter, bytes available to fill in @outbuf
*
- * Same as the standard Unix routine <function>iconv()</function>, but
- * may be implemented via libiconv on Unix flavors that lack
+ * Same as the standard UNIX routine <function>iconv()</function>, but
+ * may be implemented via libiconv on UNIX flavors that lack
* a native implementation.
*
- * &GLib; provides g_convert() and g_locale_to_utf8() which are likely
+ * GLib provides g_convert() and g_locale_to_utf8() which are likely
* more convenient than the raw iconv wrappers.
*
* Return value: count of non-reversible conversions, or -1 on error
* g_iconv_close:
* @converter: a conversion descriptor from g_iconv_open()
*
- * Same as the standard Unix routine <function>iconv_close()</function>, but
- * may be implemented via libiconv on Unix flavors that lack
+ * Same as the standard UNIX routine <function>iconv_close()</function>, but
+ * may be implemented via libiconv on UNIX flavors that lack
* a native implementation. Should be called to clean up
* the conversion descriptor from g_iconv_open() when
* you are done converting things.
*
- * &GLib; provides g_convert() and g_locale_to_utf8() which are likely
+ * GLib provides g_convert() and g_locale_to_utf8() which are likely
* more convenient than the raw iconv wrappers.
*
* Return value: -1 on error, 0 on success
return iconv_close (cd);
}
+
+#define ICONV_CACHE_SIZE (16)
+
+struct _iconv_cache_bucket {
+ gchar *key;
+ guint32 refcount;
+ gboolean used;
+ iconv_t cd;
+};
+
+static GList *iconv_cache_list;
+static GHashTable *iconv_cache;
+static GHashTable *iconv_open_hash;
+static guint iconv_cache_size = 0;
+G_LOCK_DEFINE_STATIC (iconv_cache_lock);
+
+/* caller *must* hold the iconv_cache_lock */
+static void
+iconv_cache_init (void)
+{
+ static gboolean initialized = FALSE;
+
+ if (initialized)
+ return;
+
+ iconv_cache_list = NULL;
+ iconv_cache = g_hash_table_new (g_str_hash, g_str_equal);
+ iconv_open_hash = g_hash_table_new (g_direct_hash, g_direct_equal);
+
+ initialized = TRUE;
+}
+
+
+/**
+ * iconv_cache_bucket_new:
+ * @key: cache key
+ * @cd: iconv descriptor
+ *
+ * Creates a new cache bucket, inserts it into the cache and
+ * increments the cache size.
+ *
+ * Returns a pointer to the newly allocated cache bucket.
+ **/
+struct _iconv_cache_bucket *
+iconv_cache_bucket_new (const gchar *key, iconv_t cd)
+{
+ struct _iconv_cache_bucket *bucket;
+
+ bucket = g_new (struct _iconv_cache_bucket, 1);
+ bucket->key = g_strdup (key);
+ bucket->refcount = 1;
+ bucket->used = TRUE;
+ bucket->cd = cd;
+
+ g_hash_table_insert (iconv_cache, bucket->key, bucket);
+
+ /* FIXME: if we sorted the list so items with few refcounts were
+ first, then we could expire them faster in iconv_cache_expire_unused () */
+ iconv_cache_list = g_list_prepend (iconv_cache_list, bucket);
+
+ iconv_cache_size++;
+
+ return bucket;
+}
+
+
+/**
+ * iconv_cache_bucket_expire:
+ * @node: cache bucket's node
+ * @bucket: cache bucket
+ *
+ * Expires a single cache bucket @bucket. This should only ever be
+ * called on a bucket that currently has no used iconv descriptors
+ * open.
+ *
+ * @node is not a required argument. If @node is not supplied, we
+ * search for it ourselves.
+ **/
+static void
+iconv_cache_bucket_expire (GList *node, struct _iconv_cache_bucket *bucket)
+{
+ g_hash_table_remove (iconv_cache, bucket->key);
+
+ if (node == NULL)
+ node = g_list_find (iconv_cache_list, bucket);
+
+ g_assert (node != NULL);
+
+ if (node->prev)
+ {
+ node->prev->next = node->next;
+ if (node->next)
+ node->next->prev = node->prev;
+ }
+ else
+ {
+ iconv_cache_list = node->next;
+ if (node->next)
+ node->next->prev = NULL;
+ }
+
+ g_list_free_1 (node);
+
+ g_free (bucket->key);
+ g_iconv_close (bucket->cd);
+ g_free (bucket);
+
+ iconv_cache_size--;
+}
+
+
+/**
+ * iconv_cache_expire_unused:
+ *
+ * Expires as many unused cache buckets as it needs to in order to get
+ * the total number of buckets < ICONV_CACHE_SIZE.
+ **/
+static void
+iconv_cache_expire_unused (void)
+{
+ struct _iconv_cache_bucket *bucket;
+ GList *node, *next;
+
+ node = iconv_cache_list;
+ while (node && iconv_cache_size >= ICONV_CACHE_SIZE)
+ {
+ next = node->next;
+
+ bucket = node->data;
+ if (bucket->refcount == 0)
+ iconv_cache_bucket_expire (node, bucket);
+
+ node = next;
+ }
+}
+
static GIConv
open_converter (const gchar *to_codeset,
- const gchar *from_codeset,
+ const gchar *from_codeset,
GError **error)
{
- GIConv cd = g_iconv_open (to_codeset, from_codeset);
-
- if (cd == (iconv_t) -1)
+ struct _iconv_cache_bucket *bucket;
+ gchar *key;
+ GIConv cd;
+
+ /* create our key */
+ key = g_alloca (strlen (from_codeset) + strlen (to_codeset) + 2);
+ sprintf (key, "%s:%s", from_codeset, to_codeset);
+
+ G_LOCK (iconv_cache_lock);
+
+ /* make sure the cache has been initialized */
+ iconv_cache_init ();
+
+ bucket = g_hash_table_lookup (iconv_cache, key);
+ if (bucket)
{
- /* Something went wrong. */
- if (errno == EINVAL)
- g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_NO_CONVERSION,
- _("Conversion from character set '%s' to '%s' is not supported"),
- from_codeset, to_codeset);
+ if (bucket->used)
+ {
+ cd = g_iconv_open (to_codeset, from_codeset);
+ if (cd == (iconv_t) -1)
+ goto error;
+ }
else
- g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
- _("Could not open converter from '%s' to '%s': %s"),
- from_codeset, to_codeset, strerror (errno));
+ {
+ cd = bucket->cd;
+ bucket->used = TRUE;
+
+ /* reset the descriptor */
+ g_iconv (cd, NULL, NULL, NULL, NULL);
+ }
+
+ bucket->refcount++;
}
-
+ else
+ {
+ cd = g_iconv_open (to_codeset, from_codeset);
+ if (cd == (iconv_t) -1)
+ goto error;
+
+ iconv_cache_expire_unused ();
+
+ bucket = iconv_cache_bucket_new (key, cd);
+ }
+
+ g_hash_table_insert (iconv_open_hash, cd, bucket->key);
+
+ G_UNLOCK (iconv_cache_lock);
+
+ return cd;
+
+ error:
+
+ G_UNLOCK (iconv_cache_lock);
+
+ /* Something went wrong. */
+ if (errno == EINVAL)
+ g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_NO_CONVERSION,
+ _("Conversion from character set '%s' to '%s' is not supported"),
+ from_codeset, to_codeset);
+ else
+ g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
+ _("Could not open converter from '%s' to '%s': %s"),
+ from_codeset, to_codeset, g_strerror (errno));
+
return cd;
+}
+static int
+close_converter (GIConv converter)
+{
+ struct _iconv_cache_bucket *bucket;
+ const gchar *key;
+ iconv_t cd;
+
+ cd = (iconv_t) converter;
+
+ if (cd == (iconv_t) -1)
+ return 0;
+
+ G_LOCK (iconv_cache_lock);
+
+ key = g_hash_table_lookup (iconv_open_hash, cd);
+ if (key)
+ {
+ g_hash_table_remove (iconv_open_hash, cd);
+
+ bucket = g_hash_table_lookup (iconv_cache, key);
+ g_assert (bucket);
+
+ bucket->refcount--;
+
+ if (cd == bucket->cd)
+ bucket->used = FALSE;
+ else
+ g_iconv_close (cd);
+
+ if (!bucket->refcount && iconv_cache_size > ICONV_CACHE_SIZE)
+ {
+ /* expire this cache bucket */
+ iconv_cache_bucket_expire (NULL, bucket);
+ }
+ }
+ else
+ {
+ G_UNLOCK (iconv_cache_lock);
+
+ g_warning ("This iconv context wasn't opened using open_converter");
+
+ return g_iconv_close (converter);
+ }
+
+ G_UNLOCK (iconv_cache_lock);
+
+ return 0;
}
+
/**
* g_convert:
* @str: the string to convert
g_return_val_if_fail (str != NULL, NULL);
g_return_val_if_fail (to_codeset != NULL, NULL);
g_return_val_if_fail (from_codeset != NULL, NULL);
-
+
cd = open_converter (to_codeset, from_codeset, error);
if (cd == (GIConv) -1)
bytes_read, bytes_written,
error);
- g_iconv_close (cd);
+ close_converter (cd);
return res;
}
default:
g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
_("Error during conversion: %s"),
- strerror (errno));
+ g_strerror (errno));
have_error = TRUE;
break;
}
* for the fallback sequences in @fallback will be honored. Some
* systems may do a approximate conversion from @from_codeset
* to @to_codeset in their <function>iconv()</function> functions,
- * in which case &GLib; will simply return that approximate conversion.
+ * in which case GLib will simply return that approximate conversion.
*
* Return value: If the conversion was successful, a newly allocated
* nul-terminated string, which must be freed with
utf8 = g_convert (str, len, "UTF-8", from_codeset,
bytes_read, &inbytes_remaining, error);
if (!utf8)
- return NULL;
+ {
+ close_converter (cd);
+ if (bytes_written)
+ *bytes_written = 0;
+ return NULL;
+ }
/* Now the heart of the code. We loop through the UTF-8 string, and
* whenever we hit an offending character, we form fallback, convert
default:
g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_FAILED,
_("Error during conversion: %s"),
- strerror (errno));
+ g_strerror (errno));
have_error = TRUE;
break;
}
*/
*outp = '\0';
- g_iconv_close (cd);
+ close_converter (cd);
if (bytes_written)
- *bytes_written = outp - str; /* Doesn't include '\0' */
+ *bytes_written = outp - dest; /* Doesn't include '\0' */
g_free (utf8);
*
*/
+#ifndef G_PLATFORM_WIN32
+
static gchar *
strdup_len (const gchar *string,
gssize len,
return g_strndup (string, real_len);
}
+#endif
+
/**
* g_locale_to_utf8:
* @opsysstring: a string in the encoding of the current locale
} UnsafeCharacterSet;
static const guchar acceptable[96] = {
- /* X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 XA XB XC XD XE XF */
- 0x00,0x3F,0x20,0x20,0x20,0x00,0x2C,0x3F,0x3F,0x3F,0x3F,0x22,0x20,0x3F,0x3F,0x1C, /* 2X !"#$%&'()*+,-./ */
- 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x38,0x20,0x20,0x2C,0x20,0x2C, /* 3X 0123456789:;<=>? */
- 0x30,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F, /* 4X @ABCDEFGHIJKLMNO */
- 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x20,0x20,0x20,0x20,0x3F, /* 5X PQRSTUVWXYZ[\]^_ */
- 0x20,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F, /* 6X `abcdefghijklmno */
- 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x20,0x20,0x20,0x3F,0x20 /* 7X pqrstuvwxyz{|}~DEL */
+ /* A table of the ASCII chars from space (32) to DEL (127) */
+ /* ! " # $ % & ' ( ) * + , - . / */
+ 0x00,0x3F,0x20,0x20,0x20,0x00,0x2C,0x3F,0x3F,0x3F,0x3F,0x22,0x20,0x3F,0x3F,0x1C,
+ /* 0 1 2 3 4 5 6 7 8 9 : ; < = > ? */
+ 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x38,0x20,0x20,0x2C,0x20,0x2C,
+ /* @ A B C D E F G H I J K L M N O */
+ 0x30,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,
+ /* P Q R S T U V W X Y Z [ \ ] ^ _ */
+ 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x20,0x20,0x20,0x20,0x3F,
+ /* ` a b c d e f g h i j k l m n o */
+ 0x20,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,
+ /* p q r s t u v w x y z { | } ~ DEL */
+ 0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x20,0x20,0x20,0x3F,0x20
};
static const gchar hex[16] = "0123456789ABCDEF";
use_mask = mask;
for (p = string; *p != '\0'; p++)
{
- c = *p;
+ c = (guchar) *p;
if (!ACCEPTABLE (c))
unacceptable++;
}
use_mask = mask;
for (q = result, p = string; *p != '\0'; p++)
{
- c = (unsigned char)*p;
+ c = (guchar) *p;
if (!ACCEPTABLE (c))
{
char *escaped_path;
char *res;
+#ifdef G_OS_WIN32
+ char *p, *backslash;
+
+ /* Turn backslashes into forward slashes. That's what Netscape
+ * does, and they are actually more or less equivalent in Windows.
+ */
+
+ pathname = g_strdup (pathname);
+ p = (char *) pathname;
+
+ while ((backslash = strchr (p, '\\')) != NULL)
+ {
+ *backslash = '/';
+ p = backslash + 1;
+ }
+#endif
+
if (hostname && *hostname != '\0')
{
escaped_hostname = g_escape_uri_string (hostname, UNSAFE_HOST);
escaped_path,
NULL);
+#ifdef G_OS_WIN32
+ g_free ((char *) pathname);
+#endif
+
g_free (escaped_hostname);
g_free (escaped_path);
int first_digit;
int second_digit;
- first_digit = g_ascii_xdigit_value (*scanner++);
-
+ first_digit = g_ascii_xdigit_value (scanner[0]);
if (first_digit < 0)
return -1;
- second_digit = g_ascii_xdigit_value (*scanner++);
+ second_digit = g_ascii_xdigit_value (scanner[1]);
if (second_digit < 0)
return -1;
}
static gchar *
-g_unescape_uri_string (const gchar *escaped,
- const gchar *illegal_characters,
- int len)
+g_unescape_uri_string (const char *escaped,
+ int len,
+ const char *illegal_escaped_characters,
+ gboolean ascii_must_not_be_escaped)
{
const gchar *in, *in_end;
gchar *out, *result;
- int character;
+ int c;
if (escaped == NULL)
return NULL;
if (len < 0)
len = strlen (escaped);
- result = g_malloc (len + 1);
+ result = g_malloc (len + 1);
out = result;
- for (in = escaped, in_end = escaped + len; in < in_end && *in != '\0'; in++)
+ for (in = escaped, in_end = escaped + len; in < in_end; in++)
{
- character = *in;
- if (character == '%')
+ c = *in;
+
+ if (c == '%')
{
- character = unescape_character (in + 1);
-
- /* Check for an illegal character. We consider '\0' illegal here. */
- if (character == 0
- || (illegal_characters != NULL
- && strchr (illegal_characters, (char)character) != NULL))
- {
- g_free (result);
- return NULL;
- }
+ /* catch partial escape sequences past the end of the substring */
+ if (in + 3 > in_end)
+ break;
+
+ c = unescape_character (in + 1);
+
+ /* catch bad escape sequences and NUL characters */
+ if (c <= 0)
+ break;
+
+ /* catch escaped ASCII */
+ if (ascii_must_not_be_escaped && c <= 0x7F)
+ break;
+
+ /* catch other illegal escaped characters */
+ if (strchr (illegal_escaped_characters, c) != NULL)
+ break;
+
in += 2;
}
- *out++ = character;
+
+ *out++ = c;
}
+ g_assert (out - result <= len);
*out = '\0';
-
- g_assert (out - result <= strlen (escaped));
- if (!g_utf8_validate (result, -1, NULL))
+ if (in != in_end || !g_utf8_validate (result, -1, NULL))
{
g_free (result);
return NULL;
}
-
+
return result;
}
+static gboolean
+is_escalphanum (gunichar c)
+{
+ return c > 0x7F || g_ascii_isalnum (c);
+}
+
+static gboolean
+is_escalpha (gunichar c)
+{
+ return c > 0x7F || g_ascii_isalpha (c);
+}
+
+/* allows an empty string */
+static gboolean
+hostname_validate (const char *hostname)
+{
+ const char *p;
+ gunichar c, first_char, last_char;
+
+ p = hostname;
+ if (*p == '\0')
+ return TRUE;
+ do
+ {
+ /* read in a label */
+ c = g_utf8_get_char (p);
+ p = g_utf8_next_char (p);
+ if (!is_escalphanum (c))
+ return FALSE;
+ first_char = c;
+ do
+ {
+ last_char = c;
+ c = g_utf8_get_char (p);
+ p = g_utf8_next_char (p);
+ }
+ while (is_escalphanum (c) || c == '-');
+ if (last_char == '-')
+ return FALSE;
+
+ /* if that was the last label, check that it was a toplabel */
+ if (c == '\0' || (c == '.' && *p == '\0'))
+ return is_escalpha (first_char);
+ }
+ while (c == '.');
+ return FALSE;
+}
+
/**
* g_filename_from_uri:
* @uri: a uri describing a filename (escaped, encoded in UTF-8).
char *result;
char *filename;
int offs;
+#ifdef G_OS_WIN32
+ char *p, *slash;
+#endif
if (hostname)
*hostname = NULL;
return NULL;
}
- unescaped_hostname = g_unescape_uri_string (host_part, "", path_part - host_part);
- if (unescaped_hostname == NULL)
+ unescaped_hostname = g_unescape_uri_string (host_part, path_part - host_part, "", TRUE);
+
+ if (unescaped_hostname == NULL ||
+ !hostname_validate (unescaped_hostname))
{
+ g_free (unescaped_hostname);
g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_BAD_URI,
- _("The hostname of the URI '%s' contains invalidly escaped characters"),
+ _("The hostname of the URI '%s' is invalid"),
uri);
return NULL;
}
g_free (unescaped_hostname);
}
- filename = g_unescape_uri_string (path_part, "/", -1);
+ filename = g_unescape_uri_string (path_part, -1, "/", FALSE);
if (filename == NULL)
{
return NULL;
}
- /* DOS uri's are like "file://host/c:\foo", so we need to check if we need to
- * drop the initial slash */
offs = 0;
- if (g_path_is_absolute (filename+1))
- offs = 1;
+#ifdef G_OS_WIN32
+ /* Drop localhost */
+ if (hostname && *hostname != NULL &&
+ g_ascii_strcasecmp (*hostname, "localhost") == 0)
+ {
+ g_free (*hostname);
+ *hostname = NULL;
+ }
+
+ /* Turn slashes into backslashes, because that's the canonical spelling */
+ p = filename;
+ while ((slash = strchr (p, '/')) != NULL)
+ {
+ *slash = '\\';
+ p = slash + 1;
+ }
+
+ /* Windows URIs with a drive letter can be like "file://host/c:/foo"
+ * or "file://host/c|/foo" (some Netscape versions). In those cases, start
+ * the filename from the drive letter.
+ */
+ if (g_ascii_isalpha (filename[1]))
+ {
+ if (filename[2] == ':')
+ offs = 1;
+ else if (filename[2] == '|')
+ {
+ filename[2] = ':';
+ offs = 1;
+ }
+ }
+#endif
result = g_filename_from_utf8 (filename + offs, -1, NULL, NULL, error);
g_free (filename);
**/
gchar *
g_filename_to_uri (const char *filename,
- char *hostname,
+ const char *hostname,
GError **error)
{
char *escaped_uri;
return NULL;
}
- utf8_filename = g_filename_to_utf8 (filename, -1, NULL, NULL, error);
- if (utf8_filename == NULL)
- return NULL;
-
if (hostname &&
- !g_utf8_validate (hostname, -1, NULL))
+ !(g_utf8_validate (hostname, -1, NULL)
+ && hostname_validate (hostname)))
{
- g_free (utf8_filename);
g_set_error (error, G_CONVERT_ERROR, G_CONVERT_ERROR_ILLEGAL_SEQUENCE,
- _("Invalid byte sequence in hostname"));
+ _("Invalid hostname"));
return NULL;
}
+ utf8_filename = g_filename_to_utf8 (filename, -1, NULL, NULL, error);
+ if (utf8_filename == NULL)
+ return NULL;
+
+#ifdef G_OS_WIN32
+ /* Don't use localhost unnecessarily */
+ if (hostname && g_ascii_strcasecmp (hostname, "localhost") == 0)
+ hostname = NULL;
+#endif
+
escaped_uri = g_escape_file_uri (hostname,
utf8_filename);
g_free (utf8_filename);