Jan-Henrik,
I've started using threads in my application and now the mysql driver starts to segfault:
Starting program: /home/paul/git/dbmail/src/dbmail-imapd -D [Thread debugging using libthread_db enabled] [New process 27692] [New Thread -1217631440 (LWP 27692)] [New Thread -1218696304 (LWP 27694)] [Thread -1218696304 (LWP 27694) exited]
Program received signal SIGSEGV, Segmentation fault. [Switching to Thread -1217631440 (LWP 27692)] 0xb7d80175 in ResultSet_free (R=0x81526e4) at src/db/ResultSet.c:65 65 (*R)->op->free(&(*R)->I);
this only happens when I use the mysql driver. The postgres driver has exhibited some infrequent crashes but those have been very difficult to reproduce.
My guess is it has to do with the query involved:
INSERT IGNORE INTO dbmail_headername (headername) VALUES (?)
a construct not used for postgres.
Perhaps this type of query affects the kind of MYSQL_RES returned. I'm not sure. I'll try to investigate myself.
Found what caused the segfault.
A bug in my own code. Thing is I assigned a value to an out-of-bounds parameter. Doing so causes segfaults in the ResetSet_free code. Would it be possible to raise an SQLException in those cases instead?
demo code attached.
Paul J Stevens wrote:
Jan-Henrik,
I've started using threads in my application and now the mysql driver starts to segfault:
Starting program: /home/paul/git/dbmail/src/dbmail-imapd -D [Thread debugging using libthread_db enabled] [New process 27692] [New Thread -1217631440 (LWP 27692)] [New Thread -1218696304 (LWP 27694)] [Thread -1218696304 (LWP 27694) exited]
Program received signal SIGSEGV, Segmentation fault. [Switching to Thread -1217631440 (LWP 27692)] 0xb7d80175 in ResultSet_free (R=0x81526e4) at src/db/ResultSet.c:65 65 (*R)->op->free(&(*R)->I);
this only happens when I use the mysql driver. The postgres driver has exhibited some infrequent crashes but those have been very difficult to reproduce.
My guess is it has to do with the query involved:
INSERT IGNORE INTO dbmail_headername (headername) VALUES (?)
a construct not used for postgres.
Perhaps this type of query affects the kind of MYSQL_RES returned. I'm not sure. I'll try to investigate myself.
with attachment this time
On 6. april. 2008, at 22.04, Paul J Stevens wrote:
Found what caused the segfault.
A bug in my own code. Thing is I assigned a value to an out-of-bounds parameter. Doing so causes segfaults in the ResetSet_free code. Would it be possible to raise an SQLException in those cases instead?
libzdb should already throw an exception in that case, but unfortunately, both in mysql and postgresql we had an off-by-one bug when testing parameter and column index, so your invalid parameter index slipped through. A bug fix is checked in.
-
Jan-Henrik Haukeland -
Paul J Stevens