Re: [ECOS] gdb "step" behavior in multi-threaded programs

[Jonathan Larmour <jifl at eCosCentric dot com>]

Kevin Hilman wrote:
> I'm working on the eCos gdb stub for new architecture, and I'm a bit
> confused about the correct gdb behavior for breaking/stepping for
> multi-threaded programs.  For example, if you have a function being
> called by multiple threads and you set a breakpoint at that function.
>
> I presume that for breakpoints, whenever ANY thread calls that
> function, the breakpoint should be hit, right?
Correct, although GDB has the concept of per-thread breakpoints. See the 
GDB documentation for details...

Although in practice per-thread breakpoints are implemented by letting 
every thread hit the breakpoint anyway and GDB just resuming the target 
without user input.

> What about stepping in that function (via the "step" command)? Should
> gdb stop whenever ANY thread hits the next source line? or just the
> thread that hit the breakpoint?  And what about "stepi"?
Since GDB has per-thread breakpoints it certainly can do it for only one 
thread, and I believe it does (bugs aside).

The GDB info files have stuff about scheduler locking too to allow threads 
to be stepped without the risk of unnecessary task switches. This is 
supported in the eCos GDB stub.

Jifl
--
eCosCentric       http://www.eCosCentric.com/       <info@eCosCentric.com>
--[ "You can complain because roses have thorns, or you ]--
--[  can rejoice because thorns have roses." -Lincoln   ]-- Opinions==mine


--
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss