Re: Various problems and/or questions on Insight 5.2.1

[Christopher Faylor <cgf@redhat.com>]

On Fri, Sep 06, 2002 at 10:41:17AM +0300, Kai Ruottu wrote:
>Christopher Faylor <cgf@redhat.com> wrote:
>>>On Wed, 4 Sep 2002, Christopher Faylor wrote:
>>>>We really can't require people to install X in order to get a graphical
>>>>debugger for cygwin.
>
>Basically this issue is about supporting Linux, not Windows.  Which one
>RedHat tries to do?

No, that is not the issue at all.  The issue for this thread was getting
insight working on Windows with the current version of tcl/tk in
sourceware.

The discussion has been hijacked by people who want to talk about X but
X is really irrelevant to this particular discussion.

>I see the X11 being one way to finally get a debugger for cygwin (and
>mingw), which one can remotely run on Linux.

I am talking about fixing insight for Windows.  Insight has worked on
Windows for *years*.  There is no way in the world that anyone is going
to eliminate it's existing behavior.  That is so breathtakingly
ludicrous that I can't believe that anyone would even argue in its
favor.

Cygwin's XFree86 implementation is a wonderful thing.  It does not
necessarily work "out of the box", though.  So, I am not going to
say to any insight user "Oh, you want to debug?  Well, first you're
going to have to start X.  Doesn't work?  Ok.  Let's debug your X
and find out why."

If you want to argue that "someone" should offer insight/X in the cygwin
distribution or that the configury should be changed to allow an X
version for Windows then feel free.  Just don't argue that an X version
of insight will solve any existing insight problems because, there is no
evidence that this is the case.

I should point out that you will undoubtedly be very disappointed with
running your X version of insight remotely on linux since, unless you
are debugging another X application, all output will still be showing
up on your Windows system in console windows or Windows GUI screens.

So, in other words, to paraphrase:

"X11!"

"You keep using that word.  I do not think it means what you think it
means."

>AFAIK Cygwin shouldn't have much to do with 'supporting native- Windows
>with using only the Win32-API', but trying to implement Unix on
>Windows, so X11 belongs to this idea, not applications using the native
>Win32-API.  For this purpose there is the Mingw- port.

You "K" wrong.  Cygwin does support running and debugging native windows
applications.  I use it for that regularly.  So do many others.

Mingw is not actually officially supported by insight/gdb.

>>>>I suppose but, while I can't direct people's time, it sure seems like
>>>>focusing on fixing the native insight is a much much higher priority.
>
>In all sanity the 'native' must mean Mingw, not Cygwin...

Ok.  If it makes you happy, change 'native' to 'cygwin' above.

>>Well, IMO, suggesting that cygwin users would want to install X in
>>order to do debugging really is rather of a derailment of the issue at
>>hand.
>
>They haven't understood the 'Cygwin' idea if they don't want to convert
>their Windozes into Unixes and don't expect Cygwin really doing this...
>As funny as it sounds, the Cygwin-users on the net really are trying to
>use their Windoze/Cygwin's as 'Unixes', for instance trying to build
>GNU-sources on this platform...  And not using Linux or something more
>Unix-like with Windoze-target cross- tools for this purpose.  Maybe
>they even try to build Insights on Windoze/Cygwin...

I think I know what Cygwin is used for and what its intentions are:

	  http://cygwin.com/who.html

>>I really don't see what X has to do with the discussion of fixing
>>tcl/tk/insight.  It seems to me like you are dragging an entirely
>>different issue into this discussion.
>
>Years ago it was possible to start X11-apps on a single Solaris2
>station and control these apps from Windoze-stations...  If now people
>want to start Insight on a Windoze-station and control it from a
>Linux-station because all the GNU sources and development tools,
>besides GDB now, are there, why RedHat is against this idea?  I assume
>Chris telling what the RedHat'ers think about this issue...

What is it with you freenet.hut.fi people?  You all seem to be missing
the point.  Or is it just .fi people?

Hopefully, you see the point here.  If you infer a Red Hat position from
people with redhat.com as their email address then you are generalizing
way too much.

I do run the cygwin project, so you can assume that I speak
authoritatively for that.  However, cygwin is extremely far from being a
core focus of Red Hat.  It isn't even sold as a product.  So, it is
unlikely that you could infer much of a Red Hat position from anything
that I say.

And, what have I said authoritatively here?  I've said that I think that
the priority should be fixing the current cygwin version of insight.
I've said that I, personally, have no interest in an X version and
wasn't personally going to be providing it.

As far as insight (and cygwin for that matter) is concerned, I devote MY
OWN TIME to the packaging, release, and bug fixing.  Red Hat is savvy
enough to understand that a business model where I sit around all day,
answering email, and packaging up insight wouldn't really fly but they
obviously have no problem with me doing things in my free time.  Also,
although Cygwin/XFree86 is not my project, I have actively supported it.

We *do* have customers who would be understandably perplexed if insight
started acting differently.  And I *know* that we will have existing
"net release" insight users who will not want to install X just to debug
things.  They already have a GUI desktop.  Whether you hate the Windows
interface or not there is no reason to force *them* to use X.

If someone else wants to provide an insight-for-X or a tcl-for-X for
Windows they are welcome to do so.  I see this as a major support
headache and, so, have no interest in doing this myself.  In fact,
you've amply demonstrated to me that there would be all sorts of
misconceptions surrounding the project.  I have enough to do without
taking on something where I have to explain "I understand that you
have an insight screen on your linux workstation but that doesn't
mean that debugging output from your application will show up there
too." on a daily basis.

>Many Linux-users think Windoze being an 'embedded' system, which is not
>capable to serve as a sane development/build environment.

Actually, I doubt that that is true.  Cygwin has been around for years.
It is a very popular project.

>Normally the debugger works in the sane host, but in the Windoze-
>case it has been obligatory to run it on the  target system...

But adding X awareness to insight doesn't solve this problem.

Also, you can easily use a Windows-hosted insight to debug target
applications on embedded boards.

You could even, with a not inconceivable amount of effort, probably
get gdbserver to run on Windows, allowing you to build a linux x
windows cross-build environment -- which is how things are normally
done in an "embedded situation".

>But are there some serious problems in using a Windoze-station as an
>X11-client when Cygwin/X11 has been installed there?

If you are asking questions about Cygwin/XFree86 then you should use
the cygwin-xfree mailing list.  Otherwise, I think I've answered this
above.

FWIW, I have said all that I care to about this subject.  I would be
interested in hearing what happens with Keith's (voluntary, unpaid)
efforts to incorporate Mumit Khan's patches into sourceware, though.

If you are interested in providing an insight for XFree86 package for
cygwin, then your first step would be to fix the insight configury
to allow this.  Then you'll need to go to the cygwin web page to
see what steps are required to submit a package.  It's really pretty
easy.

cgf