[moved to serdev]
-- the disscution concerns only the development branch until it gets to
a new release (although the problem can be in stable, too) --
Daniel
On 11/30/04 18:20, msg wrote:
Hi Daiel,
Thanks for your reply.
could you please send the backtrace?
I grabbed the running process and wasn't in the source tree -- I'll try
it again and send the results.
Also, the line doesn't match for me, do you
have a modified version?
Yeah, it seems to be at the end of dbt_row_update_val(); the version is
cvs HEAD.
If so, please make a tarball with
all sources, binaries and core files and put it
somewhere for
download (ftp, http).
Will do (no core files since none were generated).
Also after further testing:
With only _one_ entry in 'subscriber' but additional accounting logic
added to 'ser.cfg', the timer process goes sig 10 in the same way (I'll
need to backtrace this one too). While typing this my logs also added
this entry:
10(13702) db_insert_ucontact(): Error while inserting contact
10(13702) wb_timer(): Error while inserting contact into database
This is with my original 'ser.cfg' and one entry in the 'subscriber'
file.
Please let me know if my field definitions for the database files are
correct (from the previous post regarding dbtext).
Regards,
Michael
>> 'dbtext' works well if the 'subscriber' file contains only one
>> entry; when there are two entries, the 'ser' 'timer' process
crashes
>> with
>> signal 10 (bus error) just after the first UA registers over the wire.
>> 'gdb' debugging the 'timer' process reports:
>
>> Program received signal
SIGBUS, Bus error.
>> 0x8538b5c in dbt_row_set_val () at dbt_tb.c:403