Bacula 3.0.2 crashes on transfer in Snow Leopard 64bit

[claunia@…]

Log of bacula console on director (Linux, zeus-fd is Snow Leopard 64-bit)

05-oct 19:22 hades-sd JobId 8789: Ready to append to end of Volume "000182" at file=1.
05-oct 19:22 zeus-fd: ERROR in find_one.c:706 Failed ASSERT: name_max+1 > (int)sizeof(struct dirent) + (int)NAMELEN(entry)
05-oct 19:22 hades-dir JobId 8789: Fatal error: Network error with FD during Backup: ERR=No hay datos disponibles
05-oct 19:22 hades-sd JobId 8789: Fatal error: append.c:159 Error reading data header from FD. ERR=No hay datos disponibles
05-oct 19:22 hades-sd JobId 8789: Job write elapsed time = 00:00:01, Transfer rate = 0  bytes/second
05-oct 19:22 hades-sd JobId 8789: Fatal error: append.c:292 Fatal append error on device "TapeDevice" (/dev/nst0): ERR=
05-oct 19:22 hades-sd JobId 8789: Job zeus.claunia.com:_Backup_iPhoto_Library.2009-10-05_19.22.32_16 marked to be canceled.
05-oct 19:22 hades-sd JobId 8789: Fatal error: fd_cmds.c:166 Command error with FD, hanging up. Append data error.

05-oct 19:22 hades-dir JobId 8789: Fatal error: No Job status returned from FD.
05-oct 19:22 hades-dir JobId 8789: Error: Bacula hades-dir 3.0.2 (18Jul09): 05-oct-2009 19:22:40

/opt/local/var/bacula/working/bacula.14607.traceback 
Reading symbols for shared libraries ............. done
/opt/local/var/bacula/working/14604: No such file or directory
Attaching to program: `/opt/local/sbin/bacula-fd', process 14604.
Reading symbols for shared libraries .++++++++++++.... done
0x00007fff800bf9f2 in select$DARWIN_EXTSN ()
$1 = <unknown type>
$2 = <unknown type>
$3 = <unknown type>
$4 = <unknown type>
$5 = <unknown type>
$6 = <unknown type>
$7 = <unknown type>
$8 = <unknown type>
$9 = <unknown type>
#0  0x00007fff800bf9f2 in select$DARWIN_EXTSN ()
#1  0x000000010005d9b8 in bnet_thread_server ()
#2  0x0000000100001e8f in main ()

Thread 4 (process 14604):
#0  0x00007fff800bf9f2 in select$DARWIN_EXTSN ()
#1  0x000000010005f05d in BSOCK::wait_data_intr ()
#2  0x000000010000a47d in sd_heartbeat_thread ()
#3  0x00007fff800b4f66 in _pthread_start ()
#4  0x00007fff800b4e19 in thread_start ()

Thread 3 (process 14604):
#0  0x00007fff800ca032 in __wait4 ()
#1  0x000000010007a688 in signal_handler ()
#2  <signal handler called>
#3  0x0000000100042a2f in find_one_file ()
#4  0x0000000100040966 in find_files ()
#5  0x0000000100004551 in blast_data_to_storage_daemon ()
#6  0x000000010000cb8c in backup_cmd ()
#7  0x000000010000ae0c in handle_client_request ()
#8  0x00000001000849f2 in workq_server ()
#9  0x00007fff800b4f66 in _pthread_start ()
#10 0x00007fff800b4e19 in thread_start ()

Thread 2 (process 14604):
#0  0x00007fff800b69c6 in __semwait_signal ()
#1  0x00007fff800ba801 in _pthread_cond_wait ()
#2  0x0000000100084347 in watchdog_thread ()
#3  0x00007fff800b4f66 in _pthread_start ()
#4  0x00007fff800b4e19 in thread_start ()

Thread 1 (process 14604):
#0  0x00007fff800bf9f2 in select$DARWIN_EXTSN ()
#1  0x000000010005d9b8 in bnet_thread_server ()
#2  0x0000000100001e8f in main ()
#0  0x00007fff800bf9f2 in select$DARWIN_EXTSN ()
No symbol table info available.
#1  0x000000010005d9b8 in bnet_thread_server ()
No symbol table info available.
#2  0x0000000100001e8f in main ()
No symbol table info available.
#0  0x0000000000000000 in ?? ()
No symbol table info available.
#0  0x0000000000000000 in ?? ()
No symbol table info available.
#0  0x0000000000000000 in ?? ()
No symbol table info available.
#0  0x0000000000000000 in ?? ()
No symbol table info available.
#0  0x0000000000000000 in ?? ()
No symbol table info available.

It's a not stop, tested on another snow leopard machine and no way, same error. It's unusable in this status.

Installed simply with macports install bacula. Have not changed the macports configuration files at all (they are as default) and have not used variants.

[mf2k (Frank Schima)]

What is the result of the following command?

port installed bacula

[claunia@…]

zeus:working claunia$ port installed bacula The following ports are currently installed:

bacula @3.0.2_0+client_only+darwin bacula @3.0.2_2+client_only+darwin (active)

and

hecate:~ claunia$ /opt/local/bin/port installed bacula The following ports are currently installed:

bacula @3.0.2_2+client_only+darwin (active)

(Boths machines do the same fail)

[mf2k (Frank Schima)]

I see a problem here too. I believe a patch for building 64-bit from ticket #21674 was causing the crashes. Reverted in r58814. Please let me know if this works for you or not.

[claunia@…]

Things now are worse. bacula-fd doesn't even exists anymore!

hecate:local claunia$ bin/port installed bacula The following ports are currently installed:

bacula @3.0.2_3+client_only+darwin (active)

hecate:local claunia$ file sbin/bacula-fd sbin/bacula-fd: cannot open `sbin/bacula-fd' (No such file or directory)

[mf2k (Frank Schima)]

I re-applied the patch from r58828. I'll keep looking into this. Any patches appreciated.

[claunia@…]

It's strange that in gentoo linux x86-64 it does work correctly with same GCC version.

Gentoo does apply a patch, but it seems to be only for configuration.

[mf2k (Frank Schima)]

Can you attach that patch or a link to it?

[claunia@…]

Gentoo's Bacula 3.0.2 ebuild

[claunia@…]

Gentoo's Bacula 3.0.2 patch

[claunia@…]

I've attached both the ebuild that configures and compiles it as well as the patch that's getting applied. There is another patch for making it work with GNOME's g2ksu

[mf2k (Frank Schima)]

Thanks for the files. So an ebuild file is the gentoo version of a Macports Portfile apparently. I'm not able to glean anything useful from it that will help us. Also, the other patch just has to do with the location of the conf files. I don't see that as being helpful for fixing our 64-bit problem.

So I'm not sure how to fix this right now.

[claunia@…]

is there at least a way to instruct it creating a 32-bit executable just for bacula and its dependences so I can continue doing backups meanwhile?

[mf2k (Frank Schima)]

I have filed a ​bug report upstream.

[mf2k (Frank Schima)]

Replying to claunia@…:

is there at least a way to instruct it creating a 32-bit executable just for bacula and its dependences so I can continue doing backups meanwhile?

All of the bacula dependencies would also have to be built 32-bit - or with +universal. I have not tried this yet. So you could try building bacula +universal, but you will also need to compile at least zlib, openssl, libiconv, gettext, ncurses that way too. See also #21865, wine-devel can only run 32-bit.

[claunia@…]

Replying to macsforever2000@…:

Replying to claunia@…:

is there at least a way to instruct it creating a 32-bit executable just for bacula and its dependences so I can continue doing backups meanwhile?

All of the bacula dependencies would also have to be built 32-bit - or with +universal. I have not tried this yet. So you could try building bacula +universal, but you will also need to compile at least zlib, openssl, libiconv, gettext, ncurses that way too. See also #21865, wine-devel can only run 32-bit.

Let me try and I'll comment you. If it does still crash or fail it's not a 64-bit issue.

wine in gentoo amd64 also is compiled as a x86 binary, there is an option to compile it as a x86-64 binary but then only win64 executables can be run.

[claunia@…]

I installed all the dependencies as universal, changed buildarch to i386, installed bacula, and, still no executable found O.o

[claunia@…]

Changing buildarch to i386 does not work, let buildarch be commented, remove x86_64 from universal (JUST FOR bacula, let it from the dependencies), install bacula, executable is created.

zeus:~ claunia$ file /opt/local/sbin/bacula-fd /opt/local/sbin/bacula-fd: Mach-O executable i386

Testing, I got the same error: 08-oct 01:35 zeus-fd: ERROR in find_one.c:706 Failed ASSERT: name_max+1 > (int)sizeof(struct dirent) + (int)NAMELEN(entry)

It's not a 64-bit issue, it's a runtime or compiler one. The 64-bit patch that was applied seems to not be a problem as it allows it to compile and give the same error as in 32-bit.

Is there anyway to disable optimizations?

DEBUG: Environment: CFLAGS='-O2 -arch i386' CPPFLAGS='-I/opt/local/include -I/opt/local/ncurses' CXXFLAGS='-O2 -arch i386'

Maybe that -O2 is the problem, my gentoo CFLAGS are CFLAGS="-march=nocona -mfpmath=sse -msse -msse2 -mmmx -msse3 -O3 -pipe" and it works.

[mf2k (Frank Schima)]

Good news. This is a known issue and fixed for 3.0.3 which is not released yet. There is a patch on the bacula ticket and I will try to apply that today.

[claunia@…]

Great! Please comment me when it is in the port so I can continue the backups!

[mf2k (Frank Schima)]

Committed revision r59025. It works on one of my clients with SL. Let me know how it works for you.

[claunia@…]

I'm still stuck at bacula @3.0.2_5+client_only when doing port selfupdate, I'll retry later.

[mf2k (Frank Schima)]

I should have updated the revision, sorry. Done in r59027. You just need to wait for the port index in a few minutes.

[claunia@…]

Ok, waited a couple of minutes, uninstalled, installed, running in both clients as sweet.

Thanks!