Mail Archives: cygwin/2008/08/04/20:35:05
Christopher Faylor wrote:
> Does anyone actually use insight on Cygwin?
I do. I am a complete idiot when it comes to using gdb in "normal" mode.
Without insight, I'm lost.
> Keith Seitz, the insight
> maintainer, has recently made changes which will allow insight to work
> with a non-special version of tcl/tk rather than the hacked version that
> it had previously been built with. That opens the door to building a
> real version of tcl/tk for cygwin and linking insight to it.
> Unfortunately, I believe that would mean that insight would need to run
> in an X window rather than natively.
Not a problem for me; I don't mind running an Xserver. And I routinely
run both cygwin's very old Xserver and the 'free' XMing version.
However, I remember this issue came up several years ago. There were a
few arguments in favor of the hacked tcl/tk version of insight:
(1) How do you debug the Xserver if your debugger depends on it?
Of course, there is nobody currently trying to debug the Xserver on
cygwin -- or develop it in any way -- so that's most likely a moot
point. And besides, this same issue faces *nix developers: there are
ways around it. For instance, use a stable Xserver on a different
display, and run insight on that display to debug the unstable target
Xserver (or a program running on that Xserver).
(2) Red Hat's paying customers expected a standalone debugger, and would
balk at a Xserver requirement.
This is even more true now, I would imagine, as cygwin-xfree has been
all-but-dead for years. (What? Your debugger requires an Xserver, but
you don't provide a current one?) Hopefully some of these problems will
soon improve, at least on the X-library, if not X-server, side --
</raises beer vaguely in Yaakov's direction>
OTOH, maybe Red Hat is now providing a different graphical debugger to
its embedded customers -- one that uses the MI interface to gdb
directly. Like, say, Eclipse does -- and therefore their customers may
no longer care about insight. Maybe. I dunno.
Also, this may impact the ability to do
error_start=X:\somewhere\insight.exe -- but that's probably not a big deal.
> The other alternative is to nuke insight entirely. Is anyone really
> relying on it? Since I barely test it when I release gdb, I'd expect
> that there would be problem reports but I don't remember seeing any for
> quite some time.
That's 'cause it works. <g>
I could certainly see splitting the distribution into two packages. This
means that 'gdb -w' would fail in strange and wonderful ways if the
"insight" package with all of its tcl scripts were not installed,
but...maybe that's okay.
--
Chuck
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
- Raw text -