On Thu, Mar 4, 2010 at 1:03 PM, Marc Khouzam
<marc.khouzam@xxxxxxxxxxxx> wrote:
> I'm of the opinion this is very serious. Bizarre behavior on
> interrupt is one of the main reasons I consider using EDC for
> Windows instead of gdb. If gdb 7, the version we can
> recommend for MinGW users and what will be in Wascana 1.0,
> has a solution we should go for it. And I don't see this as a
> major change. The only thing affected should be gdb startup
> and the interrupt code. We don't need to introduce any
> additional async functionality, at least not right away.
Ok, we should give it a shot, if it is a bad problem for our users.
However, I agree with John, that although the change is minor (in code)
it will affect everything. Basically, we are changing the way GDB
is behaving throughout the debug session, from beginning to end.
Maybe GDB won't actually change much it's behaviour as Dan mentions.
If GDB did change, that would raise further alarm bells with me.
One solution could be to only turn on Async mode for Windows,
since that's the platform that really needs it. Once that proves to be safe
we can make the general change. But maybe I'm complicating things too much...
Doing it Windows only would add complexity. Keeping all comms on the MI channel is a much cleaner approach than this out-of-band interrupt signal to get gdb's attention. Just my opinion, though.
At the root of everyone's concern seems to be mistrust that gdb will do the right thing. I hope there are ways we can mitigate that, like bridging the community gap as Dan is doing.
Doug.
