View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0001566 | GNUnet | hostlist daemon | public | 2010-06-08 00:17 | 2010-07-03 20:55 |
| Reporter | Christian Grothoff | Assigned To | Matthias Wachs | ||
| Priority | urgent | Severity | major | Reproducibility | have not tried |
| Status | closed | Resolution | fixed | ||
| Summary | 0001566: hostlist service (in conjunction with stats) went CPU crazy (100%) | ||||
| Description | The calculated 'delay' in 'task_check' is always zero, causing the code to spin. Interesting data: (gdb) print delay $3 = {value = 0} (gdb) print hostlist_delay $4 = {value = 0} (gdb) print stat_connection_count $5 = 4294967295 Note that the actual connection count should be zero (AFAIK peer is not connected, might have been though -- did we get two disconnect events for one connect event?). | ||||
| Additional Information | Trace: (gdb) ba #0 0x0034d7e4 in gettext () from /lib/tls/i686/cmov/libc.so.6 #1 0x0034e810 in gettext () from /lib/tls/i686/cmov/libc.so.6 #2 0x0034d523 in dcgettext () from /lib/tls/i686/cmov/libc.so.6 #3 0x0034d550 in dgettext () from /lib/tls/i686/cmov/libc.so.6 #4 0x0804c490 in task_check (cls=0x0, tc=0xbf895968) at hostlist-client.c:1103 #5 0x00e88041 in run_ready (sched=0xbf8959cc) at scheduler.c:559 #6 0x00e884dd in GNUNET_SCHEDULER_run (task=0xe84704 <program_main>, task_cls=0xbf895b04) at scheduler.c:687 #7 0x00e84d2e in GNUNET_PROGRAM_run (argc=4, argv=0xbf895c44, binaryName=0x804ee33 "hostlist", binaryHelp=0x804ee6c "GNUnet hostlist server and client", options=0x80511c0, task=0x804a21a <run>, task_cls=0x0) at program.c:227 #8 0x0804a48a in main (argc=4, argv=0xbf895c44) at gnunet-daemon-hostlist.c:354 | ||||
| Tags | No tags attached. | ||||
|
|
Added assertion to disconnect handler to prevent underflow |
|
|
stat_connection_count is - incremented when a peer connects - decremented when a peer disconnects - is of type static unsigned int - 4294967295 is (2^32)-1 |
|
|
Any progress on finding out why core might give us more disconnect callbacks than connect callbacks? |
|
|
Fixed in SVN 12083. Disconnect notifications were given even if the peer had not even managed to do a session key exchange, but in that case connect notifications had not been created. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2010-06-08 00:17 | Christian Grothoff | New Issue | |
| 2010-06-08 00:17 | Christian Grothoff | Status | new => assigned |
| 2010-06-08 00:17 | Christian Grothoff | Assigned To | => Matthias Wachs |
| 2010-06-08 10:30 | Matthias Wachs | Note Added: 0004030 | |
| 2010-06-08 10:30 | Matthias Wachs | Status | assigned => acknowledged |
| 2010-06-08 10:43 | Matthias Wachs | Note Added: 0004031 | |
| 2010-06-18 00:14 | Christian Grothoff | Note Added: 0004034 | |
| 2010-07-02 09:21 | Christian Grothoff | Note Added: 0004054 | |
| 2010-07-02 09:21 | Christian Grothoff | Status | acknowledged => resolved |
| 2010-07-02 09:21 | Christian Grothoff | Resolution | open => fixed |
| 2010-07-03 20:55 | Christian Grothoff | Status | resolved => closed |
