Opened 4 years ago

Last modified 3 years ago

#554 new defect

provoking ARB fatal error does not cause termination

Reported by: westram Owned by: devel
Priority: major Milestone: arb6.1
Component: no idea Version: SVN
Keywords: Cc:

Description (last modified by westram)

Reproduce:

cd $ARBHOME ; rm lib/arb_tcp.dat ; arb demo.arb

Instantly fails with

ARB: Internal ARB Error: ARB_DB_CLIENT_OPEN

but remains locked in launcher.

Notes:

  • possibly related to #492
  • arb apparently does not terminate!
  • test #436 when fixed

Change History (5)

comment:1 Changed 4 years ago by westram

Console output:

ARB: Internal ARB Error: ARB_DB_CLIENT_OPEN
(Reason: OPEN_ARB_DB_CLIENT ERROR: ARB ERROR: File $ARBHOME/lib/arb_tcp.dat missing or unreadable)
ARB: ARB is most likely unstable now (due to this error).
If you've made changes to the database, consider to save it using a different name.
Try to fix the cause of the error and restart ARB.

-------------------- ARB-backtrace 'Internal ARB Error: ARB_DB_CLIENT_OPEN
(Reason: OPEN_ARB_DB_CLIENT ERROR: ARB ERROR: File $ARBHOME/lib/arb_tcp.dat missing or unreadable)':
/home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/lib/libCORE.so(GBK_dump_backtrace(_IO_FILE*, char const*)+0x34)[0x7ff1b1825c54]
/home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/lib/libCORE.so(GB_internal_error(char const*)+0x51)[0x7ff1b18268b1]
/home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/lib/libCORE.so(GB_internal_errorf(char const*, ...)+0x95)[0x7ff1b1826965]
/home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/lib/libARBDB.so(gbcmc_open(char const*)+0xa9)[0x7ff1b1bd4b89]
/home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/lib/libARBDB.so(GBCMS_open(char const*, long, GBDATA*)+0xbd)[0x7ff1b1afd26d]
arb_ntree(startup_dbserver(AW_root*, char const*, GBDATA*)+0x3f)[0x609e1f]
arb_ntree[0x48137b]
arb_ntree[0x4815f0]
arb_ntree[0x4834cb]
arb_ntree(ARB_main(int, char**)+0x47a)[0x483cba]
/lib/libc.so.6(__libc_start_main+0xfd)[0x7ff1af68ac8d]
arb_ntree[0x461ec9]
-------------------- End of backtrace
ARB_DB_SERVER_ERROR: OPEN_ARB_DB_CLIENT ERROR: ARB ERROR: File $ARBHOME/lib/arb_tcp.dat missing or unreadable

^C^C^C^C^C^C^C^C

Process tree:

  \_ -csh
  |   \_ /bin/bash -u /home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/bin/arb demo.arb
  |       \_ /bin/bash /home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/bin/arb_launcher arb_ntree demo.arb
  |           \_ /bin/bash /home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/bin/arb_launcher arb_ntree demo.arb
  |           |   \_ /bin/bash /home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/bin/arb_launcher arb_ntree demo.arb
  |           |   |   \_ arb_ntree demo.arb
  |           |   |       \_ arb_ntree demo.arb
  |           |   \_ /bin/bash /home/ralf/ARB-bilbo/ARB.trunk.test.NDEBUG/bin/arb_launcher arb_ntree demo.arb
  |           \_ tee /home/ralf/.arb_prop/logs/20140604_100359.7947/run.log

comment:2 Changed 4 years ago by epruesse

I think unrelated.

The error is triggered in gbcmc_open and passed down into load_and_startup_main_window, where it is ignored:

NT_main.cxx:204

        GB_ERROR problem = startup_mainwindow_and_dbserver(aw_root, autorun_macro);
        aw_message_if(problem); // no need to terminate ARB

Thereafter, the main loop simply continues to run w/o window.

I'm not sure what exactly is going on there. The startup of NT is split into a gazillion functions, so I cannot tell, what terminating on problem!=NULL would mean. It looks like macro replay is also done here (maybe NT is doing too many different things?), which is probably why the error is deemed non-fatal.

comment:3 Changed 3 years ago by westram

  • Milestone set to arb6.1

comment:4 Changed 3 years ago by westram

  • Description modified (diff)
  • Summary changed from provoking ARB fatal error blocks launcher to provoking ARB fatal error does not cause termination

comment:5 Changed 3 years ago by westram

  • Description modified (diff)
Note: See TracTickets for help on using tickets.