Corrected memory leak introduced by revision 1005

This commit is contained in:
Thomas Schmitt 2007-09-07 16:40:38 +00:00
parent 624e05055c
commit 74eff6f013
3 changed files with 32 additions and 14 deletions

View File

@ -1 +1 @@
#define Cdrskin_timestamP "2007.09.07.154951"
#define Cdrskin_timestamP "2007.09.07.163949"

View File

@ -132,7 +132,7 @@ static void *scan_worker_func(struct w_list *w)
{
int ret;
ret = burn_drive_scan_sync(w->u.scan.drives, w->u.scan.n_drives);
ret = burn_drive_scan_sync(w->u.scan.drives, w->u.scan.n_drives, 1);
if (ret <= 0)
w->u.scan.done = -1;
else
@ -195,7 +195,7 @@ drive_is_active:;
if (workers != NULL) {
libdax_msgs_submit(libdax_messenger, -1, 0x00020101,
LIBDAX_MSGS_SEV_WARNING, LIBDAX_MSGS_PRIO_HIGH,
"After scan a drive operation is still going on",
"After scan a drive operation is still going on",
0, 0);
return -1;
}

View File

@ -775,9 +775,25 @@ static int drive_getcaps(struct burn_drive *d, struct burn_drive_info *out)
return 1;
}
/* ts A70907 : added parameter flag */
/* @param flag bit0= reset global drive list */
int burn_drive_scan_sync(struct burn_drive_info *drives[],
unsigned int *n_drives)
unsigned int *n_drives, int flag)
{
/* ts A70907 :
There seems to be a misunderstanding about the role of
burn_drive_scan_sync(). It needs no static state because it
is only started once during an asynchronous scan operation.
Its starter, burn_drive_scan(), is the one which ends immediately
and gets called repeatedly. It acts on start of scanning by
calling burn_drive_scan_sync(), returns idle while scanning is
not done and finally removes the worker object which represented
burn_drive_scan_sync().
The scanning itself is not parallel but enumerates sequentially
drive by drive (within scsi_enumerate_drives()).
*/
/* state vars for the scan process */
/* ts A60904 : did set some default values to feel comfortable */
static int scanning = 0, scanned = 0, found = 0;
@ -800,7 +816,8 @@ int burn_drive_scan_sync(struct burn_drive_info *drives[],
#endif /* 0 */
/* ts A70907 : moved here from burn_drive_info_free() */
burn_drive_free_all();
if (flag & 1)
burn_drive_free_all();
/* refresh the lib's drives */
@ -820,6 +837,7 @@ int burn_drive_scan_sync(struct burn_drive_info *drives[],
LIBDAX_MSGS_SEV_FATAL,
LIBDAX_MSGS_PRIO_HIGH,
"Out of virtual memory", 0, 0);
scanning = 0;
return -1;
} else
(*drives)[count].drive = NULL; /* end mark */
@ -896,7 +914,7 @@ void burn_drive_info_free(struct burn_drive_info drive_infos[])
#ifndef Libburn_free_all_drives_on_infO
/* ts A70907 : Solution for wrong behavior below */
for (i = 0; drive_infos[i].drive != NULL; i++)
return burn_drive_free(drive_infos[i].drive);
burn_drive_free(drive_infos[i].drive);
#endif
/* ts A60904 : This looks a bit weird. [ts A70907 : not any more]
@ -1124,6 +1142,8 @@ int burn_drive_scan_and_grab(struct burn_drive_info *drive_infos[], char* adr,
unsigned int n_drives;
int ret;
/* >>> check wether drive adress is already registered */
if(strncmp(adr, "stdio:", 6) == 0) {
ret = burn_drive_grab_dummy(drive_infos, adr + 6);
return ret;
@ -1135,14 +1155,12 @@ int burn_drive_scan_and_grab(struct burn_drive_info *drive_infos[], char* adr,
fprintf(stderr,"libburn: experimental: burn_drive_scan_and_grab(%s)\n",
adr);
*/
while (1) {
ret = burn_drive_scan(drive_infos, &n_drives);
if (ret < 0)
return -1;
if (ret > 0)
break;
usleep(1002);
}
/* ts A70907 : now calling synchronously rather than looping */
ret = burn_drive_scan_sync(drive_infos, &n_drives, 0);
if (ret < 0)
return -1;
if (n_drives <= 0)
return 0;
/*