View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0002494 | GNUnet | testbed service | public | 2012-07-09 22:19 | 2012-11-05 17:34 |
| Reporter | Sree Harsha Totakura | Assigned To | Sree Harsha Totakura | Due Date | |
| Priority | normal | Severity | major | Reproducibility | always |
| Status | closed | Resolution | fixed | ||
| Platform | Linux-x86 | OS | Ubuntu | OS Version | 12.04 |
| Product Version | Git master | ||||
| Target Version | 0.9.4 | Fixed in Version | 0.9.4 | ||
| Summary | 0002494: testbed_api test is misbehaving (was: stdin is closed in subprocesses started by GNUNET_OS_start_process_vap()) | ||||
| Description | After recent change SVN 22578 sub-processes started with GNUNET_OS_start_process_vap() don't seem to have an opened stdin. This is confirmed by testbed API function GNUNET_TESTBED_host_run_() in testbed_api_hosts.c which starts gnunet-service-arm as a sub-process through GNUNET_OS_start_process_vap(). | ||||
| Additional Information | When GNUNET_FORCE_LOG is set to "testbed*;;;;DEBUG/;;;;WARNING" it can be observed that gnunet-service-tested is restarted repeatedly when running test_testbed_api. Furthermore, the reason for the shutdown trigger in testbed service is observed to be stdin close. The repeated restarts are due to gnunet-arm restarting the service. | ||||
| Tags | No tags attached. | ||||
| The reason is that this gnunet-service-testbed process is started by gnunet-service-arm and not from testbed_api_hosts.c. ARM is not supposed to pass a stdin to the services, thus this fails (even with my new code which passes /dev/null for stdin instead). Changing 'testbed.conf' BINARY to /bin/true shows that it is really ARM starting the process and not testbed_api_hosts.c. Not sure why the testcase runs gnunet-service-testbed from ARM, as that should never be done. In fact, as testbed is not a regular service it should probably not have the arm-specific options in testbed.conf in the first place. | |
| Looking at test_testbed_api.c, the issue seems clear: the code should not begin with 'GNUNET_TESTING_peer_run' and start ARM and then try to use 'TESTBED_controller_connect'. Instead, TESTBED_controller_start should be used. | |
| GNUNET_TESTBED_controller_start API changed, doesn't it need to have 'cfg' as an argument!? The controller process needs to have a configuration... | |
| Yes, the problem was with the test case. I was using GNUNET_TESTING_peer_run and making ARM to start the process as I wanted to debug the service. Without this we have to invent a new way to start the testbed service under gdb for debugging. | |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2012-07-09 22:19 | Sree Harsha Totakura | New Issue | |
| 2012-07-10 14:04 | Christian Grothoff | Note Added: 0006234 | |
| 2012-07-10 14:10 | Christian Grothoff | Note Added: 0006235 | |
| 2012-07-10 14:13 | Christian Grothoff | Note Added: 0006236 | |
| 2012-07-10 14:17 | Christian Grothoff | Assigned To | => Sree Harsha Totakura |
| 2012-07-10 14:17 | Christian Grothoff | Status | new => assigned |
| 2012-07-10 14:17 | Christian Grothoff | Category | util library => testbed service |
| 2012-07-10 14:17 | Christian Grothoff | Summary | stdin is closed in subprocesses started by GNUNET_OS_start_process_vap() => testbed_api test is misbehaving (was: stdin is closed in subprocesses started by GNUNET_OS_start_process_vap()) |
| 2012-07-10 15:39 | Sree Harsha Totakura | Note Added: 0006237 | |
| 2012-07-10 15:39 | Sree Harsha Totakura | Status | assigned => resolved |
| 2012-07-10 15:39 | Sree Harsha Totakura | Fixed in Version | => Git master |
| 2012-07-10 15:39 | Sree Harsha Totakura | Resolution | open => fixed |
| 2012-07-17 22:07 | Christian Grothoff | Fixed in Version | Git master => 0.9.4 |
| 2012-07-17 22:07 | Christian Grothoff | Target Version | => 0.9.4 |
| 2012-11-05 17:34 | Christian Grothoff | Status | resolved => closed |