mirror of
https://gitlab.gnome.org/GNOME/glib.git
synced 2024-11-05 00:46:16 +01:00
01f4ccc859
2004-10-06 Matthias Clasen <mclasen@redhat.com> * glib/gfileutils.c: Fix some C99isms. (#154676, Kjartan Maraas)
1221 lines
29 KiB
C
1221 lines
29 KiB
C
/* gfileutils.c - File utility functions
|
|
*
|
|
* Copyright 2000 Red Hat, Inc.
|
|
*
|
|
* GLib is free software; you can redistribute it and/or modify it
|
|
* under the terms of the GNU Lesser General Public License as
|
|
* published by the Free Software Foundation; either version 2 of the
|
|
* License, or (at your option) any later version.
|
|
*
|
|
* GLib is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* Lesser General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
* License along with GLib; see the file COPYING.LIB. If not,
|
|
* write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
|
|
* Boston, MA 02111-1307, USA.
|
|
*/
|
|
|
|
#include "config.h"
|
|
|
|
#include "galias.h"
|
|
#include "glib.h"
|
|
|
|
#include <sys/stat.h>
|
|
#ifdef HAVE_UNISTD_H
|
|
#include <unistd.h>
|
|
#endif
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <stdarg.h>
|
|
#include <string.h>
|
|
#include <errno.h>
|
|
#include <sys/types.h>
|
|
#include <sys/stat.h>
|
|
#include <fcntl.h>
|
|
#include <stdlib.h>
|
|
|
|
#ifdef G_OS_WIN32
|
|
#include <io.h>
|
|
#ifndef F_OK
|
|
#define F_OK 0
|
|
#define W_OK 2
|
|
#define R_OK 4
|
|
#endif /* !F_OK */
|
|
|
|
#ifndef S_ISREG
|
|
#define S_ISREG(mode) ((mode)&_S_IFREG)
|
|
#endif
|
|
|
|
#ifndef S_ISDIR
|
|
#define S_ISDIR(mode) ((mode)&_S_IFDIR)
|
|
#endif
|
|
|
|
#endif /* G_OS_WIN32 */
|
|
|
|
#ifndef S_ISLNK
|
|
#define S_ISLNK(x) 0
|
|
#endif
|
|
|
|
#ifndef O_BINARY
|
|
#define O_BINARY 0
|
|
#endif
|
|
|
|
#include "glibintl.h"
|
|
|
|
#ifdef G_OS_WIN32
|
|
#define G_IS_DIR_SEPARATOR(c) (c == G_DIR_SEPARATOR || c == '/')
|
|
#else
|
|
#define G_IS_DIR_SEPARATOR(c) (c == G_DIR_SEPARATOR)
|
|
#endif
|
|
|
|
/**
|
|
* g_file_test:
|
|
* @filename: a filename to test
|
|
* @test: bitfield of #GFileTest flags
|
|
*
|
|
* Returns %TRUE if any of the tests in the bitfield @test are
|
|
* %TRUE. For example, <literal>(G_FILE_TEST_EXISTS |
|
|
* G_FILE_TEST_IS_DIR)</literal> will return %TRUE if the file exists;
|
|
* the check whether it's a directory doesn't matter since the existence
|
|
* test is %TRUE. With the current set of available tests, there's no point
|
|
* passing in more than one test at a time.
|
|
*
|
|
* Apart from %G_FILE_TEST_IS_SYMLINK all tests follow symbolic links,
|
|
* so for a symbolic link to a regular file g_file_test() will return
|
|
* %TRUE for both %G_FILE_TEST_IS_SYMLINK and %G_FILE_TEST_IS_REGULAR.
|
|
*
|
|
* Note, that for a dangling symbolic link g_file_test() will return
|
|
* %TRUE for %G_FILE_TEST_IS_SYMLINK and %FALSE for all other flags.
|
|
*
|
|
* You should never use g_file_test() to test whether it is safe
|
|
* to perform an operaton, because there is always the possibility
|
|
* of the condition changing before you actually perform the operation.
|
|
* For example, you might think you could use %G_FILE_TEST_IS_SYMLINK
|
|
* to know whether it is is safe to write to a file without being
|
|
* tricked into writing into a different location. It doesn't work!
|
|
*
|
|
* <informalexample><programlisting>
|
|
* /* DON'T DO THIS */
|
|
* if (!g_file_test (filename, G_FILE_TEST_IS_SYMLINK)) {
|
|
* fd = open (filename, O_WRONLY);
|
|
* /* write to fd */
|
|
* }
|
|
* </programlisting></informalexample>
|
|
*
|
|
* Another thing to note is that %G_FILE_TEST_EXISTS and
|
|
* %G_FILE_TEST_IS_EXECUTABLE are implemented using the access()
|
|
* system call. This usually doesn't matter, but if your program
|
|
* is setuid or setgid it means that these tests will give you
|
|
* the answer for the real user ID and group ID , rather than the
|
|
* effective user ID and group ID.
|
|
*
|
|
* Return value: whether a test was %TRUE
|
|
**/
|
|
gboolean
|
|
g_file_test (const gchar *filename,
|
|
GFileTest test)
|
|
{
|
|
if ((test & G_FILE_TEST_EXISTS) && (access (filename, F_OK) == 0))
|
|
return TRUE;
|
|
|
|
#ifndef G_OS_WIN32
|
|
if ((test & G_FILE_TEST_IS_EXECUTABLE) && (access (filename, X_OK) == 0))
|
|
{
|
|
if (getuid () != 0)
|
|
return TRUE;
|
|
|
|
/* For root, on some POSIX systems, access (filename, X_OK)
|
|
* will succeed even if no executable bits are set on the
|
|
* file. We fall through to a stat test to avoid that.
|
|
*/
|
|
}
|
|
else
|
|
test &= ~G_FILE_TEST_IS_EXECUTABLE;
|
|
#endif
|
|
|
|
if (test & G_FILE_TEST_IS_SYMLINK)
|
|
{
|
|
#ifdef G_OS_WIN32
|
|
/* no sym links on win32, no lstat in msvcrt */
|
|
#else
|
|
struct stat s;
|
|
|
|
if ((lstat (filename, &s) == 0) && S_ISLNK (s.st_mode))
|
|
return TRUE;
|
|
#endif
|
|
}
|
|
|
|
if (test & (G_FILE_TEST_IS_REGULAR |
|
|
G_FILE_TEST_IS_DIR |
|
|
G_FILE_TEST_IS_EXECUTABLE))
|
|
{
|
|
struct stat s;
|
|
|
|
if (stat (filename, &s) == 0)
|
|
{
|
|
if ((test & G_FILE_TEST_IS_REGULAR) && S_ISREG (s.st_mode))
|
|
return TRUE;
|
|
|
|
if ((test & G_FILE_TEST_IS_DIR) && S_ISDIR (s.st_mode))
|
|
return TRUE;
|
|
|
|
#ifndef G_OS_WIN32
|
|
/* The extra test for root when access (file, X_OK) succeeds.
|
|
* Probably only makes sense on Unix.
|
|
*/
|
|
if ((test & G_FILE_TEST_IS_EXECUTABLE) &&
|
|
((s.st_mode & S_IXOTH) ||
|
|
(s.st_mode & S_IXUSR) ||
|
|
(s.st_mode & S_IXGRP)))
|
|
return TRUE;
|
|
#else
|
|
if ((test & G_FILE_TEST_IS_EXECUTABLE) &&
|
|
(s.st_mode & _S_IEXEC))
|
|
return TRUE;
|
|
#endif
|
|
}
|
|
}
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
GQuark
|
|
g_file_error_quark (void)
|
|
{
|
|
static GQuark q = 0;
|
|
if (q == 0)
|
|
q = g_quark_from_static_string ("g-file-error-quark");
|
|
|
|
return q;
|
|
}
|
|
|
|
/**
|
|
* g_file_error_from_errno:
|
|
* @err_no: an "errno" value
|
|
*
|
|
* Gets a #GFileError constant based on the passed-in @errno.
|
|
* For example, if you pass in %EEXIST this function returns
|
|
* #G_FILE_ERROR_EXIST. Unlike @errno values, you can portably
|
|
* assume that all #GFileError values will exist.
|
|
*
|
|
* Normally a #GFileError value goes into a #GError returned
|
|
* from a function that manipulates files. So you would use
|
|
* g_file_error_from_errno() when constructing a #GError.
|
|
*
|
|
* Return value: #GFileError corresponding to the given @errno
|
|
**/
|
|
GFileError
|
|
g_file_error_from_errno (gint err_no)
|
|
{
|
|
switch (err_no)
|
|
{
|
|
#ifdef EEXIST
|
|
case EEXIST:
|
|
return G_FILE_ERROR_EXIST;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EISDIR
|
|
case EISDIR:
|
|
return G_FILE_ERROR_ISDIR;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EACCES
|
|
case EACCES:
|
|
return G_FILE_ERROR_ACCES;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENAMETOOLONG
|
|
case ENAMETOOLONG:
|
|
return G_FILE_ERROR_NAMETOOLONG;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENOENT
|
|
case ENOENT:
|
|
return G_FILE_ERROR_NOENT;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENOTDIR
|
|
case ENOTDIR:
|
|
return G_FILE_ERROR_NOTDIR;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENXIO
|
|
case ENXIO:
|
|
return G_FILE_ERROR_NXIO;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENODEV
|
|
case ENODEV:
|
|
return G_FILE_ERROR_NODEV;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EROFS
|
|
case EROFS:
|
|
return G_FILE_ERROR_ROFS;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ETXTBSY
|
|
case ETXTBSY:
|
|
return G_FILE_ERROR_TXTBSY;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EFAULT
|
|
case EFAULT:
|
|
return G_FILE_ERROR_FAULT;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ELOOP
|
|
case ELOOP:
|
|
return G_FILE_ERROR_LOOP;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENOSPC
|
|
case ENOSPC:
|
|
return G_FILE_ERROR_NOSPC;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENOMEM
|
|
case ENOMEM:
|
|
return G_FILE_ERROR_NOMEM;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EMFILE
|
|
case EMFILE:
|
|
return G_FILE_ERROR_MFILE;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENFILE
|
|
case ENFILE:
|
|
return G_FILE_ERROR_NFILE;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EBADF
|
|
case EBADF:
|
|
return G_FILE_ERROR_BADF;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EINVAL
|
|
case EINVAL:
|
|
return G_FILE_ERROR_INVAL;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EPIPE
|
|
case EPIPE:
|
|
return G_FILE_ERROR_PIPE;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EAGAIN
|
|
case EAGAIN:
|
|
return G_FILE_ERROR_AGAIN;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EINTR
|
|
case EINTR:
|
|
return G_FILE_ERROR_INTR;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EIO
|
|
case EIO:
|
|
return G_FILE_ERROR_IO;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef EPERM
|
|
case EPERM:
|
|
return G_FILE_ERROR_PERM;
|
|
break;
|
|
#endif
|
|
|
|
#ifdef ENOSYS
|
|
case ENOSYS:
|
|
return G_FILE_ERROR_NOSYS;
|
|
break;
|
|
#endif
|
|
|
|
default:
|
|
return G_FILE_ERROR_FAILED;
|
|
break;
|
|
}
|
|
}
|
|
|
|
static gboolean
|
|
get_contents_stdio (const gchar *filename,
|
|
FILE *f,
|
|
gchar **contents,
|
|
gsize *length,
|
|
GError **error)
|
|
{
|
|
gchar buf[2048];
|
|
size_t bytes;
|
|
char *str;
|
|
size_t total_bytes;
|
|
size_t total_allocated;
|
|
|
|
g_assert (f != NULL);
|
|
|
|
#define STARTING_ALLOC 64
|
|
|
|
total_bytes = 0;
|
|
total_allocated = STARTING_ALLOC;
|
|
str = g_malloc (STARTING_ALLOC);
|
|
|
|
while (!feof (f))
|
|
{
|
|
bytes = fread (buf, 1, 2048, f);
|
|
|
|
while ((total_bytes + bytes + 1) > total_allocated)
|
|
{
|
|
total_allocated *= 2;
|
|
str = g_try_realloc (str, total_allocated);
|
|
|
|
if (str == NULL)
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
G_FILE_ERROR_NOMEM,
|
|
_("Could not allocate %lu bytes to read file \"%s\""),
|
|
(gulong) total_allocated,
|
|
utf8_filename ? utf8_filename : "???");
|
|
g_free (utf8_filename);
|
|
|
|
goto error;
|
|
}
|
|
}
|
|
|
|
if (ferror (f))
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Error reading file '%s': %s"),
|
|
utf8_filename ? utf8_filename : "???",
|
|
g_strerror (errno));
|
|
g_free (utf8_filename);
|
|
|
|
goto error;
|
|
}
|
|
|
|
memcpy (str + total_bytes, buf, bytes);
|
|
total_bytes += bytes;
|
|
}
|
|
|
|
fclose (f);
|
|
|
|
str[total_bytes] = '\0';
|
|
|
|
if (length)
|
|
*length = total_bytes;
|
|
|
|
*contents = str;
|
|
|
|
return TRUE;
|
|
|
|
error:
|
|
|
|
g_free (str);
|
|
fclose (f);
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
#ifndef G_OS_WIN32
|
|
|
|
static gboolean
|
|
get_contents_regfile (const gchar *filename,
|
|
struct stat *stat_buf,
|
|
gint fd,
|
|
gchar **contents,
|
|
gsize *length,
|
|
GError **error)
|
|
{
|
|
gchar *buf;
|
|
size_t bytes_read;
|
|
size_t size;
|
|
size_t alloc_size;
|
|
|
|
size = stat_buf->st_size;
|
|
|
|
alloc_size = size + 1;
|
|
buf = g_try_malloc (alloc_size);
|
|
|
|
if (buf == NULL)
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
G_FILE_ERROR_NOMEM,
|
|
_("Could not allocate %lu bytes to read file \"%s\""),
|
|
(gulong) alloc_size,
|
|
utf8_filename ? utf8_filename : "???");
|
|
g_free (utf8_filename);
|
|
|
|
goto error;
|
|
}
|
|
|
|
bytes_read = 0;
|
|
while (bytes_read < size)
|
|
{
|
|
gssize rc;
|
|
|
|
rc = read (fd, buf + bytes_read, size - bytes_read);
|
|
|
|
if (rc < 0)
|
|
{
|
|
if (errno != EINTR)
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
g_free (buf);
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Failed to read from file '%s': %s"),
|
|
utf8_filename ? utf8_filename : "???",
|
|
g_strerror (errno));
|
|
g_free (utf8_filename);
|
|
|
|
goto error;
|
|
}
|
|
}
|
|
else if (rc == 0)
|
|
break;
|
|
else
|
|
bytes_read += rc;
|
|
}
|
|
|
|
buf[bytes_read] = '\0';
|
|
|
|
if (length)
|
|
*length = bytes_read;
|
|
|
|
*contents = buf;
|
|
|
|
close (fd);
|
|
|
|
return TRUE;
|
|
|
|
error:
|
|
|
|
close (fd);
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
static gboolean
|
|
get_contents_posix (const gchar *filename,
|
|
gchar **contents,
|
|
gsize *length,
|
|
GError **error)
|
|
{
|
|
struct stat stat_buf;
|
|
gint fd;
|
|
|
|
/* O_BINARY useful on Cygwin */
|
|
fd = open (filename, O_RDONLY|O_BINARY);
|
|
|
|
if (fd < 0)
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Failed to open file '%s': %s"),
|
|
utf8_filename ? utf8_filename : "???",
|
|
g_strerror (errno));
|
|
g_free (utf8_filename);
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
/* I don't think this will ever fail, aside from ENOMEM, but. */
|
|
if (fstat (fd, &stat_buf) < 0)
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
close (fd);
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Failed to get attributes of file '%s': fstat() failed: %s"),
|
|
utf8_filename ? utf8_filename : "???",
|
|
g_strerror (errno));
|
|
g_free (utf8_filename);
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
if (stat_buf.st_size > 0 && S_ISREG (stat_buf.st_mode))
|
|
{
|
|
return get_contents_regfile (filename,
|
|
&stat_buf,
|
|
fd,
|
|
contents,
|
|
length,
|
|
error);
|
|
}
|
|
else
|
|
{
|
|
FILE *f;
|
|
|
|
f = fdopen (fd, "r");
|
|
|
|
if (f == NULL)
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Failed to open file '%s': fdopen() failed: %s"),
|
|
utf8_filename ? utf8_filename : "???",
|
|
g_strerror (errno));
|
|
g_free (utf8_filename);
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
return get_contents_stdio (filename, f, contents, length, error);
|
|
}
|
|
}
|
|
|
|
#else /* G_OS_WIN32 */
|
|
|
|
static gboolean
|
|
get_contents_win32 (const gchar *filename,
|
|
gchar **contents,
|
|
gsize *length,
|
|
GError **error)
|
|
{
|
|
FILE *f;
|
|
|
|
/* I guess you want binary mode; maybe you want text sometimes? */
|
|
f = fopen (filename, "rb");
|
|
|
|
if (f == NULL)
|
|
{
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Failed to open file '%s': %s"),
|
|
utf8_filename ? utf8_filename : "???",
|
|
g_strerror (errno));
|
|
g_free (utf8_filename);
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
return get_contents_stdio (filename, f, contents, length, error);
|
|
}
|
|
|
|
#endif
|
|
|
|
/**
|
|
* g_file_get_contents:
|
|
* @filename: a file to read contents from
|
|
* @contents: location to store an allocated string
|
|
* @length: location to store length in bytes of the contents
|
|
* @error: return location for a #GError
|
|
*
|
|
* Reads an entire file into allocated memory, with good error
|
|
* checking. If @error is set, %FALSE is returned, and @contents is set
|
|
* to %NULL. If %TRUE is returned, @error will not be set, and @contents
|
|
* will be set to the file contents. The string stored in @contents
|
|
* will be nul-terminated, so for text files you can pass %NULL for the
|
|
* @length argument. The error domain is #G_FILE_ERROR. Possible
|
|
* error codes are those in the #GFileError enumeration.
|
|
*
|
|
* Return value: %TRUE on success, %FALSE if error is set
|
|
**/
|
|
gboolean
|
|
g_file_get_contents (const gchar *filename,
|
|
gchar **contents,
|
|
gsize *length,
|
|
GError **error)
|
|
{
|
|
g_return_val_if_fail (filename != NULL, FALSE);
|
|
g_return_val_if_fail (contents != NULL, FALSE);
|
|
|
|
*contents = NULL;
|
|
if (length)
|
|
*length = 0;
|
|
|
|
#ifdef G_OS_WIN32
|
|
return get_contents_win32 (filename, contents, length, error);
|
|
#else
|
|
return get_contents_posix (filename, contents, length, error);
|
|
#endif
|
|
}
|
|
|
|
/*
|
|
* mkstemp() implementation is from the GNU C library.
|
|
* Copyright (C) 1991,92,93,94,95,96,97,98,99 Free Software Foundation, Inc.
|
|
*/
|
|
/**
|
|
* g_mkstemp:
|
|
* @tmpl: template filename
|
|
*
|
|
* Opens a temporary file. See the mkstemp() documentation
|
|
* on most UNIX-like systems. This is a portability wrapper, which simply calls
|
|
* mkstemp() on systems that have it, and implements
|
|
* it in GLib otherwise.
|
|
*
|
|
* The parameter is a string that should match the rules for
|
|
* mkstemp(), i.e. end in "XXXXXX". The X string will
|
|
* be modified to form the name of a file that didn't exist.
|
|
*
|
|
* Return value: A file handle (as from open()) to the file
|
|
* opened for reading and writing. The file is opened in binary mode
|
|
* on platforms where there is a difference. The file handle should be
|
|
* closed with close(). In case of errors, -1 is returned.
|
|
*/
|
|
gint
|
|
g_mkstemp (gchar *tmpl)
|
|
{
|
|
#ifdef HAVE_MKSTEMP
|
|
return mkstemp (tmpl);
|
|
#else
|
|
int len;
|
|
char *XXXXXX;
|
|
int count, fd;
|
|
static const char letters[] =
|
|
"ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789";
|
|
static const int NLETTERS = sizeof (letters) - 1;
|
|
glong value;
|
|
GTimeVal tv;
|
|
static int counter = 0;
|
|
|
|
len = strlen (tmpl);
|
|
if (len < 6 || strcmp (&tmpl[len - 6], "XXXXXX"))
|
|
return -1;
|
|
|
|
/* This is where the Xs start. */
|
|
XXXXXX = &tmpl[len - 6];
|
|
|
|
/* Get some more or less random data. */
|
|
g_get_current_time (&tv);
|
|
value = (tv.tv_usec ^ tv.tv_sec) + counter++;
|
|
|
|
for (count = 0; count < 100; value += 7777, ++count)
|
|
{
|
|
glong v = value;
|
|
|
|
/* Fill in the random bits. */
|
|
XXXXXX[0] = letters[v % NLETTERS];
|
|
v /= NLETTERS;
|
|
XXXXXX[1] = letters[v % NLETTERS];
|
|
v /= NLETTERS;
|
|
XXXXXX[2] = letters[v % NLETTERS];
|
|
v /= NLETTERS;
|
|
XXXXXX[3] = letters[v % NLETTERS];
|
|
v /= NLETTERS;
|
|
XXXXXX[4] = letters[v % NLETTERS];
|
|
v /= NLETTERS;
|
|
XXXXXX[5] = letters[v % NLETTERS];
|
|
|
|
fd = open (tmpl, O_RDWR | O_CREAT | O_EXCL | O_BINARY, 0600);
|
|
|
|
if (fd >= 0)
|
|
return fd;
|
|
else if (errno != EEXIST)
|
|
/* Any other error will apply also to other names we might
|
|
* try, and there are 2^32 or so of them, so give up now.
|
|
*/
|
|
return -1;
|
|
}
|
|
|
|
/* We got out of the loop because we ran out of combinations to try. */
|
|
return -1;
|
|
#endif
|
|
}
|
|
|
|
/**
|
|
* g_file_open_tmp:
|
|
* @tmpl: Template for file name, as in g_mkstemp(), basename only
|
|
* @name_used: location to store actual name used
|
|
* @error: return location for a #GError
|
|
*
|
|
* Opens a file for writing in the preferred directory for temporary
|
|
* files (as returned by g_get_tmp_dir()).
|
|
*
|
|
* @tmpl should be a string ending with six 'X' characters, as the
|
|
* parameter to g_mkstemp() (or mkstemp()).
|
|
* However, unlike these functions, the template should only be a
|
|
* basename, no directory components are allowed. If template is %NULL,
|
|
* a default template is used.
|
|
*
|
|
* Note that in contrast to g_mkstemp() (and mkstemp())
|
|
* @tmpl is not modified, and might thus be a read-only literal string.
|
|
*
|
|
* The actual name used is returned in @name_used if non-%NULL. This
|
|
* string should be freed with g_free() when not needed any longer.
|
|
*
|
|
* Return value: A file handle (as from open()) to
|
|
* the file opened for reading and writing. The file is opened in binary
|
|
* mode on platforms where there is a difference. The file handle should be
|
|
* closed with close(). In case of errors, -1 is returned
|
|
* and @error will be set.
|
|
**/
|
|
gint
|
|
g_file_open_tmp (const gchar *tmpl,
|
|
gchar **name_used,
|
|
GError **error)
|
|
{
|
|
int retval;
|
|
const char *tmpdir;
|
|
char *sep;
|
|
char *fulltemplate;
|
|
const char *slash;
|
|
|
|
if (tmpl == NULL)
|
|
tmpl = ".XXXXXX";
|
|
|
|
if ((slash = strchr (tmpl, G_DIR_SEPARATOR)) != NULL
|
|
#ifdef G_OS_WIN32
|
|
|| (strchr (tmpl, '/') != NULL && (slash = "/"))
|
|
#endif
|
|
)
|
|
{
|
|
char c[2];
|
|
c[0] = *slash;
|
|
c[1] = '\0';
|
|
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
G_FILE_ERROR_FAILED,
|
|
_("Template '%s' invalid, should not contain a '%s'"),
|
|
tmpl, c);
|
|
|
|
return -1;
|
|
}
|
|
|
|
if (strlen (tmpl) < 6 ||
|
|
strcmp (tmpl + strlen (tmpl) - 6, "XXXXXX") != 0)
|
|
{
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
G_FILE_ERROR_FAILED,
|
|
_("Template '%s' doesn't end with XXXXXX"),
|
|
tmpl);
|
|
return -1;
|
|
}
|
|
|
|
tmpdir = g_get_tmp_dir ();
|
|
|
|
if (G_IS_DIR_SEPARATOR (tmpdir [strlen (tmpdir) - 1]))
|
|
sep = "";
|
|
else
|
|
sep = G_DIR_SEPARATOR_S;
|
|
|
|
fulltemplate = g_strconcat (tmpdir, sep, tmpl, NULL);
|
|
|
|
retval = g_mkstemp (fulltemplate);
|
|
|
|
if (retval == -1)
|
|
{
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Failed to create file '%s': %s"),
|
|
fulltemplate, g_strerror (errno));
|
|
g_free (fulltemplate);
|
|
return -1;
|
|
}
|
|
|
|
if (name_used)
|
|
*name_used = fulltemplate;
|
|
else
|
|
g_free (fulltemplate);
|
|
|
|
return retval;
|
|
}
|
|
|
|
static gchar *
|
|
g_build_pathv (const gchar *separator,
|
|
const gchar *first_element,
|
|
va_list args)
|
|
{
|
|
GString *result;
|
|
gint separator_len = strlen (separator);
|
|
gboolean is_first = TRUE;
|
|
gboolean have_leading = FALSE;
|
|
const gchar *single_element = NULL;
|
|
const gchar *next_element;
|
|
const gchar *last_trailing = NULL;
|
|
|
|
result = g_string_new (NULL);
|
|
|
|
next_element = first_element;
|
|
|
|
while (TRUE)
|
|
{
|
|
const gchar *element;
|
|
const gchar *start;
|
|
const gchar *end;
|
|
|
|
if (next_element)
|
|
{
|
|
element = next_element;
|
|
next_element = va_arg (args, gchar *);
|
|
}
|
|
else
|
|
break;
|
|
|
|
/* Ignore empty elements */
|
|
if (!*element)
|
|
continue;
|
|
|
|
start = element;
|
|
|
|
if (separator_len)
|
|
{
|
|
while (start &&
|
|
strncmp (start, separator, separator_len) == 0)
|
|
start += separator_len;
|
|
}
|
|
|
|
end = start + strlen (start);
|
|
|
|
if (separator_len)
|
|
{
|
|
while (end >= start + separator_len &&
|
|
strncmp (end - separator_len, separator, separator_len) == 0)
|
|
end -= separator_len;
|
|
|
|
last_trailing = end;
|
|
while (last_trailing >= element + separator_len &&
|
|
strncmp (last_trailing - separator_len, separator, separator_len) == 0)
|
|
last_trailing -= separator_len;
|
|
|
|
if (!have_leading)
|
|
{
|
|
/* If the leading and trailing separator strings are in the
|
|
* same element and overlap, the result is exactly that element
|
|
*/
|
|
if (last_trailing <= start)
|
|
single_element = element;
|
|
|
|
g_string_append_len (result, element, start - element);
|
|
have_leading = TRUE;
|
|
}
|
|
else
|
|
single_element = NULL;
|
|
}
|
|
|
|
if (end == start)
|
|
continue;
|
|
|
|
if (!is_first)
|
|
g_string_append (result, separator);
|
|
|
|
g_string_append_len (result, start, end - start);
|
|
is_first = FALSE;
|
|
}
|
|
|
|
if (single_element)
|
|
{
|
|
g_string_free (result, TRUE);
|
|
return g_strdup (single_element);
|
|
}
|
|
else
|
|
{
|
|
if (last_trailing)
|
|
g_string_append (result, last_trailing);
|
|
|
|
return g_string_free (result, FALSE);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* g_build_path:
|
|
* @separator: a string used to separator the elements of the path.
|
|
* @first_element: the first element in the path
|
|
* @Varargs: remaining elements in path, terminated by %NULL
|
|
*
|
|
* Creates a path from a series of elements using @separator as the
|
|
* separator between elements. At the boundary between two elements,
|
|
* any trailing occurrences of separator in the first element, or
|
|
* leading occurrences of separator in the second element are removed
|
|
* and exactly one copy of the separator is inserted.
|
|
*
|
|
* Empty elements are ignored.
|
|
*
|
|
* The number of leading copies of the separator on the result is
|
|
* the same as the number of leading copies of the separator on
|
|
* the first non-empty element.
|
|
*
|
|
* The number of trailing copies of the separator on the result is
|
|
* the same as the number of trailing copies of the separator on
|
|
* the last non-empty element. (Determination of the number of
|
|
* trailing copies is done without stripping leading copies, so
|
|
* if the separator is <literal>ABA</literal>, <literal>ABABA</literal>
|
|
* has 1 trailing copy.)
|
|
*
|
|
* However, if there is only a single non-empty element, and there
|
|
* are no characters in that element not part of the leading or
|
|
* trailing separators, then the result is exactly the original value
|
|
* of that element.
|
|
*
|
|
* Other than for determination of the number of leading and trailing
|
|
* copies of the separator, elements consisting only of copies
|
|
* of the separator are ignored.
|
|
*
|
|
* Return value: a newly-allocated string that must be freed with g_free().
|
|
**/
|
|
gchar *
|
|
g_build_path (const gchar *separator,
|
|
const gchar *first_element,
|
|
...)
|
|
{
|
|
gchar *str;
|
|
va_list args;
|
|
|
|
g_return_val_if_fail (separator != NULL, NULL);
|
|
|
|
va_start (args, first_element);
|
|
str = g_build_pathv (separator, first_element, args);
|
|
va_end (args);
|
|
|
|
return str;
|
|
}
|
|
|
|
/**
|
|
* g_build_filename:
|
|
* @first_element: the first element in the path
|
|
* @Varargs: remaining elements in path, terminated by %NULL
|
|
*
|
|
* Creates a filename from a series of elements using the correct
|
|
* separator for filenames.
|
|
*
|
|
* On Unix, this function behaves identically to <literal>g_build_path
|
|
* (G_DIR_SEPARATOR_S, first_element, ....)</literal>.
|
|
*
|
|
* On Windows, it takes into account that either the backslash
|
|
* (<literal>\</literal> or slash (<literal>/</literal>) can be used
|
|
* as separator in filenames, but otherwise behaves as on Unix. When
|
|
* file pathname separators need to be inserted, the one that last
|
|
* previously occurred in the parameters (reading from left to right)
|
|
* is used.
|
|
*
|
|
* No attempt is made to force the resulting filename to be an absolute
|
|
* path. If the first element is a relative path, the result will
|
|
* be a relative path.
|
|
*
|
|
* Return value: a newly-allocated string that must be freed with g_free().
|
|
**/
|
|
gchar *
|
|
g_build_filename (const gchar *first_element,
|
|
...)
|
|
{
|
|
#ifndef G_OS_WIN32
|
|
gchar *str;
|
|
va_list args;
|
|
|
|
va_start (args, first_element);
|
|
str = g_build_pathv (G_DIR_SEPARATOR_S, first_element, args);
|
|
va_end (args);
|
|
|
|
return str;
|
|
#else
|
|
/* Code copied from g_build_pathv(), and modifed to use two
|
|
* alternative single-character separators.
|
|
*/
|
|
va_list args;
|
|
GString *result;
|
|
gboolean is_first = TRUE;
|
|
gboolean have_leading = FALSE;
|
|
const gchar *single_element = NULL;
|
|
const gchar *next_element;
|
|
const gchar *last_trailing = NULL;
|
|
gchar current_separator = '\\';
|
|
|
|
va_start (args, first_element);
|
|
|
|
result = g_string_new (NULL);
|
|
|
|
next_element = first_element;
|
|
|
|
while (TRUE)
|
|
{
|
|
const gchar *element;
|
|
const gchar *start;
|
|
const gchar *end;
|
|
|
|
if (next_element)
|
|
{
|
|
element = next_element;
|
|
next_element = va_arg (args, gchar *);
|
|
}
|
|
else
|
|
break;
|
|
|
|
/* Ignore empty elements */
|
|
if (!*element)
|
|
continue;
|
|
|
|
start = element;
|
|
|
|
if (TRUE)
|
|
{
|
|
while (start &&
|
|
(*start == '\\' || *start == '/'))
|
|
{
|
|
current_separator = *start;
|
|
start++;
|
|
}
|
|
}
|
|
|
|
end = start + strlen (start);
|
|
|
|
if (TRUE)
|
|
{
|
|
while (end >= start + 1 &&
|
|
(end[-1] == '\\' || end[-1] == '/'))
|
|
{
|
|
current_separator = end[-1];
|
|
end--;
|
|
}
|
|
|
|
last_trailing = end;
|
|
while (last_trailing >= element + 1 &&
|
|
(last_trailing[-1] == '\\' || last_trailing[-1] == '/'))
|
|
last_trailing--;
|
|
|
|
if (!have_leading)
|
|
{
|
|
/* If the leading and trailing separator strings are in the
|
|
* same element and overlap, the result is exactly that element
|
|
*/
|
|
if (last_trailing <= start)
|
|
single_element = element;
|
|
|
|
g_string_append_len (result, element, start - element);
|
|
have_leading = TRUE;
|
|
}
|
|
else
|
|
single_element = NULL;
|
|
}
|
|
|
|
if (end == start)
|
|
continue;
|
|
|
|
if (!is_first)
|
|
g_string_append_len (result, ¤t_separator, 1);
|
|
|
|
g_string_append_len (result, start, end - start);
|
|
is_first = FALSE;
|
|
}
|
|
|
|
va_end (args);
|
|
|
|
if (single_element)
|
|
{
|
|
g_string_free (result, TRUE);
|
|
return g_strdup (single_element);
|
|
}
|
|
else
|
|
{
|
|
if (last_trailing)
|
|
g_string_append (result, last_trailing);
|
|
|
|
return g_string_free (result, FALSE);
|
|
}
|
|
#endif
|
|
}
|
|
|
|
/**
|
|
* g_file_read_link:
|
|
* @filename: the symbolic link
|
|
* @error: return location for a #GError
|
|
*
|
|
* Reads the contents of the symbolic link @filename like the POSIX readlink() function.
|
|
* The returned string is in the encoding used for filenames. Use g_filename_to_utf8() to
|
|
* convert it to UTF-8.
|
|
*
|
|
* Returns: A newly allocated string with the contents of the symbolic link,
|
|
* or %NULL if an error occurred.
|
|
*
|
|
* Since: 2.4
|
|
*/
|
|
gchar *
|
|
g_file_read_link (const gchar *filename,
|
|
GError **error)
|
|
{
|
|
#ifdef HAVE_READLINK
|
|
gchar *buffer;
|
|
guint size;
|
|
gint read_size;
|
|
|
|
size = 256;
|
|
buffer = g_malloc (size);
|
|
|
|
while (TRUE)
|
|
{
|
|
read_size = readlink (filename, buffer, size);
|
|
if (read_size < 0) {
|
|
gchar *utf8_filename = g_filename_to_utf8 (filename, -1,
|
|
NULL, NULL, NULL);
|
|
g_free (buffer);
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
g_file_error_from_errno (errno),
|
|
_("Failed to read the symbolic link '%s': %s"),
|
|
utf8_filename ? utf8_filename : "???",
|
|
g_strerror (errno));
|
|
g_free (utf8_filename);
|
|
|
|
return NULL;
|
|
}
|
|
|
|
if (read_size < size)
|
|
{
|
|
buffer[read_size] = 0;
|
|
return buffer;
|
|
}
|
|
|
|
size *= 2;
|
|
buffer = g_realloc (buffer, size);
|
|
}
|
|
#else
|
|
g_set_error (error,
|
|
G_FILE_ERROR,
|
|
G_FILE_ERROR_INVAL,
|
|
_("Symbolic links not supported"));
|
|
|
|
return NULL;
|
|
#endif
|
|
}
|