Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
| Comment: | Closing a file from the wrong thread is harmless on most systems. (See ticket #1611) But on systems like RedHat9 with broken fcntl() locks, it leaks file descriptors. That is better than the alternative of prematurely breaking locks and causing database corruption. Nevertheless, it would be good if we could figure out a way to report errors when closing a file from the wrong thread. (CVS 2946) |
|---|---|
| Downloads: | Tarball | ZIP archive | SQL archive |
| Timelines: | family | ancestors | descendants | both | trunk |
| Files: | files | file ages | folders |
| SHA1: |
ad8f12cad1fcd9c54478abe20050f9a9 |
| User & Date: | drh 2006-01-15 02:43:16 |
Context
|
2006-01-15
| ||
| 11:39 | Rollback any open write-transaction when a shared-cache connection is closed. (CVS 2947) (check-in: 1944d92b user: danielk1977 tags: trunk) | |
| 02:43 | Closing a file from the wrong thread is harmless on most systems. (See ticket #1611) But on systems like RedHat9 with broken fcntl() locks, it leaks file descriptors. That is better than the alternative of prematurely breaking locks and causing database corruption. Nevertheless, it would be good if we could figure out a way to report errors when closing a file from the wrong thread. (CVS 2946) (check-in: ad8f12ca user: drh tags: trunk) | |
| 02:30 | Add tests and fix bugs in the new cross-thread lock resolution code. When an unlock fails, do not leak file descriptors (ticket #1611). But we really ought to report SQLITE_MISUSE or some other error instead of just returning SQLITE_OK. (CVS 2945) (check-in: f68e05cb user: drh tags: trunk) | |
Changes
Changes to src/os_unix.c.
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
....
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
|
** Close a file.
*/
static int unixClose(OsFile **pId){
unixFile *id = (unixFile*)*pId;
int rc;
if( !id ) return SQLITE_OK;
rc = unixUnlock(*pId, NO_LOCK);
if( id->dirfd>=0 ) close(id->dirfd);
id->dirfd = -1;
sqlite3OsEnterMutex();
if( id->pOpen->nLock && rc==SQLITE_OK ){
/* If there are outstanding locks, do not actually close the file just
** yet because that would clear those locks. Instead, add the file
** descriptor to pOpen->aPending. It will be automatically closed when
** the last lock is cleared.
*/
int *aNew;
struct openCnt *pOpen = id->pOpen;
................................................................................
sqlite3OsLeaveMutex();
id->isOpen = 0;
TRACE2("CLOSE %-3d\n", id->h);
OpenCounter(-1);
sqliteFree(id);
*pId = 0;
return SQLITE_OK;
}
/*
** Turn a relative pathname into a full pathname. Return a pointer
** to the full pathname stored in space obtained from sqliteMalloc().
** The calling function is responsible for freeing this space once it
** is no longer needed.
|
|
|
|
|
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
....
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
|
** Close a file.
*/
static int unixClose(OsFile **pId){
unixFile *id = (unixFile*)*pId;
int rc;
if( !id ) return SQLITE_OK;
unixUnlock(*pId, NO_LOCK);
if( id->dirfd>=0 ) close(id->dirfd);
id->dirfd = -1;
sqlite3OsEnterMutex();
if( id->pOpen->nLock ){
/* If there are outstanding locks, do not actually close the file just
** yet because that would clear those locks. Instead, add the file
** descriptor to pOpen->aPending. It will be automatically closed when
** the last lock is cleared.
*/
int *aNew;
struct openCnt *pOpen = id->pOpen;
................................................................................
sqlite3OsLeaveMutex();
id->isOpen = 0;
TRACE2("CLOSE %-3d\n", id->h);
OpenCounter(-1);
sqliteFree(id);
*pId = 0;
return rc;
}
/*
** Turn a relative pathname into a full pathname. Return a pointer
** to the full pathname stored in space obtained from sqliteMalloc().
** The calling function is responsible for freeing this space once it
** is no longer needed.
|