Re: Problems with database breakpoint routines.

[Andrew Johnson <anj@aps.anl.gov>]

Hi Nick,

Rees, NP (Nick) wrote:
> Either the initialization should be moved to iocInit, or a check should
> be done for a NULL pointer semaphore, or both. I would implement it if
> someone tells me exactly what is the best thing to do, but maybe the
> person who knows what to do should do it...

It sounds like the iocInit() option would make the most sense, but I 
don't know anything about that code myself - I assume there's only one 
instance of the bkpt_stack_sem semaphore.  If you'd like to work out an 
appropriate fix in the next few days and send me a patch I'd be happy to 
apply it for R3.14.9.

In playing just now I discovered some other breakpoint problems which 
you'd be welcome to follow up on.  Running a standard example app on 
linux-x86:

> epics> dbb anjHost:aiExample1
>    BKPT> Stopped at:  anjHost:aiExample1  within Entrypoint:  anjHost:calcExample1
> ->
> epics> dbc
>    BKPT> Continuing:  anjHost:aiExample1
> epics>
>    BKPT> Stopped at:  anjHost:aiExample1  within Entrypoint:  anjHost:calcExample1
> ->
> epics> dbd anjHost:aiExample1
> epics> dbc
> epics>
>    BKPT> End debug of lockset 2
> -> epicsMutex pthread_mutex_destroy failed: error Device or resource busy
>
> epics> dbstat

At this point my IOC shell hangs up - the prompt never returns.

- Andrew
--
There is considerable overlap between the intelligence of the smartest
bears and the dumbest tourists -- Yosemite National Park Ranger