[PATCH] Running the inferior from breakpoint commands

Eli Zaretskii eliz@delorie.com
Wed Mar 22 01:30:00 GMT 2000


> 	Any
> 	other commands in the command list are ignored, after
> 	a command that resumes execution.
> 
> Note the last sentence.  So despite the fact that the 
> testsuite seems to imply that this should work, it is
> not expected to.  I have no idea what that test is 
> doing in there.  The commands-on-breakpoint behavior
> is NOT intended to be recursive.

Yes, somebody already pointed out that this appears to be a documented
limitation.  Unfortunately, this info came after I already did the
changes and posted them; my original question when I first discovered
the problem remained unanswered...

(I did look in the manual, but since the snippet you cited isn't
indexed, I didn't find it.  [Yes, I will submit a manual change to add
an index entry.])

To me, the fact that the test suite tried to do this was a sign that
it was supposed to work.

As an aside: is there any place I can find the list of tests which
fail, and on what systems do they fail?

> If you seriously want to undertake this project, we can
> work on a list of criteria that should be tested (things
> that can go wrong).

I will put it on my todo (thanks for the list of things to test; some
of them were already tested, I just didn't report it).  I would like
at least to probe the issues, to get a feeling how badly broken is
what I did ;-).

> On top of that, I have grave concerns about being able
> to correctly save and restore all of the internal
> debugger state necessary to make this work (the
> infrun execution state, the expression chain, etc.)

Err... I'm not sure this should be of concern here (but perhaps I
don't see the subtle issues clearly enough): if the breakpoint
commands change any of these global entities, I think the user expects
the changes to be visible after these commands are processed.  For
example, if the commands delete the breakpoint where you stopped, the
user will expect the breakpoint to remain deleted after that (the
changes I submitted do handle this specific case).  Am I missing
something?


More information about the Gdb-patches mailing list