[Pidgin] #1359: purple_mkstemp method change, now raises assertion failure ...

[Pidgin]

#1359: purple_mkstemp method change, now raises assertion failure ...
------------------------+---------------------------------------------------
  Reporter:  msfbrasil  |       Owner:       
      Type:  defect     |      Status:  new  
  Priority:  minor      |   Milestone:       
 Component:  libpurple  |     Version:  2.0.0
Resolution:             |    Keywords:       
   Pending:  1          |  
------------------------+---------------------------------------------------
Changes (by msfbrasil):

  * pending:  1 => 0

Comment:

 Hello, and thank's for the post.

 The failed assertion is "_ASSERTE(_osfile(filedes) & FOPEN);" on file
 "fdopen.c".

 One thing that I ain't told before is that our customer application and
 libpurple (with associated plugins and protocols) are being build with
 MSVC .net 2003.
 For this reason, those binaries are linked agains msvcr71.dll, despite of
 glib binaries that are linked agains msvcrt.dll.

 When I change back this method implementation to the one used on gaim
 2.0.0 beta 6 below, things work fine.

 <-- begin code -->

 FILE *
 purple_mkstemp(char **fpath, gboolean binary)
 {
         const gchar *tmpdir;
 #ifndef _WIN32
         int fd;
 #endif
         FILE *fp = NULL;

         g_return_val_if_fail(fpath != NULL, NULL);

         if((tmpdir = (gchar*)g_get_tmp_dir()) != NULL) {
                 if((*fpath = g_strdup_printf("%s" G_DIR_SEPARATOR_S "%s",
 tmpdir, purple_mkstemp_templ)) != NULL) {
 #ifdef _WIN32
                         char* result = _mktemp( *fpath );
                         if( result == NULL )
                                 purple_debug(PURPLE_DEBUG_ERROR,
 "purple_mkstemp",
                                                    "Problem creating the
 template\n");
                         else
                         {
                                 if( (fp = g_fopen( result,
 binary?"wb+":"w+")) == NULL ) {
                                         purple_debug(PURPLE_DEBUG_ERROR,
 "purple_mkstemp",
                                                            "Couldn't
 fopen() %s\n", result);
                                 }
                         }
 #else
                         if((fd = mkstemp(*fpath)) == -1) {
                                 purple_debug(PURPLE_DEBUG_ERROR,
 "purple_mkstemp",
                                                    "Couldn't make \"%s\",
 error: %d\n",
                                                    *fpath, errno);
                         } else {
                                 if((fp = fdopen(fd, "r+")) == NULL) {
                                         close(fd);
                                         purple_debug(PURPLE_DEBUG_ERROR,
 "purple_mkstemp",
                                                            "Couldn't
 fdopen(), error: %d\n", errno);
                                 }
                         }
 #endif
                         if(!fp) {
                                 g_free(*fpath);
                                 *fpath = NULL;
                         }
                 }
         } else {
                 purple_debug(PURPLE_DEBUG_ERROR, "purple_mkstemp",
                                    "g_get_tmp_dir() failed!\n");
         }

         return fp;
 }

 <-- end code -->

 I'll wait for some position of you about that, because is much better for
 me to maintain as much libpurple code unchanged as possible.

 Thank's again, and best regards!

-- 
Ticket URL: <http://developer.pidgin.im/ticket/1359#comment:2>
Pidgin <http://pidgin.im>
Pidgin