This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: PATCH: Fork event updates, part the first
- From: Daniel Jacobowitz <drow at mvista dot com>
- To: gdb-patches at sources dot redhat dot com
- Date: Sat, 16 Nov 2002 14:25:41 -0500
- Subject: Re: PATCH: Fork event updates, part the first
- References: <20021113224815.GA10059@nevyn.them.org>
On Wed, Nov 13, 2002 at 05:48:15PM -0500, Daniel Jacobowitz wrote:
> I'm going to have a series of patches over the next few days to bring the
> fork support back into shape. Here's the first one.
>
> Current status summary: These operations (fork, exec, and vfork events) are
> supported under HP/UX only; fork and exec events work; vfork events, which
> constitute the bulk of the unpleasant code, do not. It appears that the
> HP/UX kernel bug (?) referenced in foll-vfork.exp is still present, or that
> the fixes to it sufficiently changed the interface that GDB's
> vfork-following code no longer works. Even HP's wdb debugger (it claims to
> be GDB 3.0, ha ha) hangs if you step over a vfork.
>
> I intend to preserve support for fork/exec on HP/UX; I'm testing on HP/UX 11
> as I go along. I intend to rip out any piece of the vfork support that I
> notice and consider intrusive; I have the design in my head for how to
> "properly" implement this feature on HP/UX, although I've no intention of
> doing it myself. In a couple of patches I'll stick an explanation of the
> technique in hppah-nat.c. If anyone else is motivated to do it I'll lend a
> hand.
>
> In three or four more I'll be ready to actually start adding the Linux
> support! (BTW: Yes, the missing GNU there was intentional... this is a
> kernel feature...)
>
> This patch removes the target target_can_follow_vfork_prior_to_exec method.
> I believe it was never in the right place; it's a wart very reminiscent of
> DECR_PC_AFTER_BREAK. It should be handled completely in the target-specific
> code and left entirely out of infrun. Then I kill a bunch of dead comments
> and simplify a lot of if statements based on the assumption that if the
> target told us about the vfork we are prepared to follow it.
>
> Unless there's any objection I'll commit this on Friday.
Committed.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer