This is the mail archive of the
gdb@sourceware.org
mailing list for the GDB project.
Re: GDB Focus Group at the 2008 GCC Summit
- From: Mark Kettenis <mark dot kettenis at xs4all dot nl>
- To: brobecker at adacore dot com
- Cc: gdb at sourceware dot org
- Date: Tue, 24 Jun 2008 00:08:17 +0200 (CEST)
- Subject: Re: GDB Focus Group at the 2008 GCC Summit
- References: <20080619190942.GA3744@adacore.com>
> Date: Thu, 19 Jun 2008 15:09:42 -0400
> From: Joel Brobecker <brobecker@adacore.com>
>
> Hello,
>
> we had a couple of 45min sessions where we discussed various items
> related to GDB. I took some brief notes, although I might have forgotten
> one or two. Anyway, here is what I wrote:
>
> | * Transition to SVN:
> |
> | Required version (for the client) just released a couple of hours ago:
> | feature allows to checkout a subset of the module (?). There is still
> | no one-line command that allows to check gdb out, so maybe will have
> | a script.
> |
> | Pb: If we want to have a combined tree, we will need to convert
> | binutils, gnulib, etc.
Objection! I have played with SVN for work and I really dilike the
fact that a checked out repository is very grep unfriendly.
> | * Using threads inside GDB:
> |
> | Problem: Expression evaluation is synchronous and blocking.
> | While GDB is doing that work, it is not handling other events,
> | which can be a problem in non-stop mode.
> |
> | Typical problem is inferior function call that causes two issues:
> | - potential length of time it takes to evaluate
> | - having to re-enter the event loop to wait for function to return.
> |
> | Pedro to think about it and write a proposal. Idea we're gearing
> | towards is one thread that runs the event loop all the time.
I guess something like this is inevitable, but I'm not happy with
using threads. Threads and signals don't mix really well, and
ptrace(2) depends heavily on threads.