Getting an insight 6.1 executable on Cygwin

Getting an insight 6.1 executable on Cygwin

[E. Weddington]

Hello!

I'm building Insight 6.1, host=cygwin, target=avr. I noticed this thread:
<http://sources.redhat.com/ml/insight/2004-q2/msg00030.html>
which leads to a patch to get an avr-insight.exe. This *mostly* works. Starting 
from a clean install directory, it doesn't build avr-insight.exe (but I do get 
an avr-gdb.exe and an avr-gdbtui.exe [that's a new one for me]). But if I build 
it again without cleaning the install directory, I then get an avr-inisght.exe.

Any thoughts on how this can be fixed? Is there a later patch that I'm missing?

Thanks
Eric

Re: Getting an insight 6.1 executable on Cygwin

[Keith Seitz]

On Fri, 2004-07-09 at 09:05, E. Weddington wrote:
> This *mostly* works. Starting from a clean install directory, it
> doesn't build avr-insight.exe (but I do get an avr-gdb.exe and an
> avr-gdbtui.exe [that's a new one for me]). But if I build it again
> without cleaning the install directory, I then get an avr-inisght.exe.

You mean a clean build directory? When you apply the patch, you must
regenerate the toplevel configure, then completely rebuild everything
(in other words, delete your build directory and reconfigure and build).
[Perhaps you've already done this. If you have, I apologize: all I can
say is "it worked for me". :-(]

> Any thoughts on how this can be fixed? Is there a later patch that I'm
> missing?

As far as I know, there are no other patches. Of course, I don't think I
ever tried a cross-gdb. I only did native. Maybe that's the ticket? Try
a clean (cross-avr) build. If that still fails, you can probably force
gdb to make you an insight executable (run make avr-insight.exe in the
gdb build directory). If that still fails, try a native cygwin build. If
that works, then we'll have something to go on.

Keith

Re: Getting an insight 6.1 executable on Cygwin

[E. Weddington]

On 9 Jul 2004 at 10:00, Keith Seitz wrote:

> On Fri, 2004-07-09 at 09:05, E. Weddington wrote:
> > This *mostly* works. Starting from a clean install directory, it
> > doesn't build avr-insight.exe (but I do get an avr-gdb.exe and an
> > avr-gdbtui.exe [that's a new one for me]). But if I build it again
> > without cleaning the install directory, I then get an avr-inisght.exe.
> 
> You mean a clean build directory? When you apply the patch, you must
> regenerate the toplevel configure, then completely rebuild everything
> (in other words, delete your build directory and reconfigure and build).
> [Perhaps you've already done this. If you have, I apologize: all I can
> say is "it worked for me". :-(]
> 
> > Any thoughts on how this can be fixed? Is there a later patch that I'm
> > missing?
> 
> As far as I know, there are no other patches. Of course, I don't think I
> ever tried a cross-gdb. I only did native. Maybe that's the ticket? Try
> a clean (cross-avr) build. If that still fails, you can probably force
> gdb to make you an insight executable (run make avr-insight.exe in the
> gdb build directory). If that still fails, try a native cygwin build. If
> that works, then we'll have something to go on.
> 

Here's what I did:
1. Clean install directory (prefix)
2. Clean build directory.
3. Unpack the source tarball.
4. Apply patch.
5. configure (build,host=cygwin; target=avr)
6. make all install

In my $(prefix)/bin directory I get avr-gdb.exe, avr-gdbtui.exe, and *no* avr-
insight.exe

In my $(build)/gdb directory I get gdb.exe, gdbtui.exe, and *no* insight.exe.

If I cd to the $(build)/gdb directory and do make insight.exe, it builds 
insight.exe in the $(build)/gdb directory. If I then do a make install in the 
$(build)/gdb directory, it looks like it installs avr-gdb.exe and avr-
gdbtui.exe, but *no* avr-insight.exe.

Perhaps this is only an issue with a cross?

HTH
Eric

Re: Getting an insight 6.1 executable on Cygwin

[E. Weddington]

On 9 Jul 2004 at 10:00, Keith Seitz wrote:

> On Fri, 2004-07-09 at 09:05, E. Weddington wrote:
> > This *mostly* works. Starting from a clean install directory, it
> > doesn't build avr-insight.exe (but I do get an avr-gdb.exe and an
> > avr-gdbtui.exe [that's a new one for me]). But if I build it again
> > without cleaning the install directory, I then get an avr-inisght.exe.

<snip> 

> As far as I know, there are no other patches. Of course, I don't think I
> ever tried a cross-gdb. I only did native. Maybe that's the ticket? Try
> a clean (cross-avr) build. If that still fails, you can probably force
> gdb to make you an insight executable (run make avr-insight.exe in the
> gdb build directory). If that still fails, try a native cygwin build. If
> that works, then we'll have something to go on.

Oh, I should also mention that building a cross insight worked fine for me for 
6.0. It installed everything correctly. So something broke between 6.0 and 6.1 
for cross building.

Eric

Re: Getting an insight 6.1 executable on Cygwin

[Keith Seitz]

On Fri, 2004-07-09 at 11:38, E. Weddington wrote:
> Here's what I did:
> 1. Clean install directory (prefix)
> 2. Clean build directory.
> 3. Unpack the source tarball.
> 4. Apply patch.

Did you regenerate top-level configure? The patch (as I recall) is for
configure.in. So you need to run autoconf to regenerate configure from
configure.in.

> 5. configure (build,host=cygwin; target=avr)
> 6. make all install

Otherwise, that all sounds correct.

BTW (I saw your other email), this is happening because of changes
outside of gdb, i.e., top-level configure. I don't even know if the
patch is correct. All I know is that it works. (And it is only needed on
cygwin; linux works without it.)

Hopefully the whole acceptance mess will get straightened out soon, and
I can spend some serious time fixing this and a host of other problems
creeping up in insight now.

Keith

Re: Getting an insight 6.1 executable on Cygwin

[E. Weddington]

On 9 Jul 2004 at 11:57, Keith Seitz wrote:

> On Fri, 2004-07-09 at 11:38, E. Weddington wrote:
> > Here's what I did:
> > 1. Clean install directory (prefix)
> > 2. Clean build directory.
> > 3. Unpack the source tarball.
> > 4. Apply patch.
> 
> Did you regenerate top-level configure? The patch (as I recall) is for
> configure.in. So you need to run autoconf to regenerate configure from
> configure.in.

Ah, that explains it. No, I don't think I did. My fault for not looking close 
enough at the patch. I'll have to get back to you on how this works out....
 
> BTW (I saw your other email), this is happening because of changes
> outside of gdb, i.e., top-level configure. I don't even know if the
> patch is correct. All I know is that it works. (And it is only needed on
> cygwin; linux works without it.)

Hey, if it works, it's correct for me. :-)
 
> Hopefully the whole acceptance mess will get straightened out soon, and
> I can spend some serious time fixing this and a host of other problems
> creeping up in insight now.

I've only just heard about this. Do you have any knowledge of the status of 
this?

Thanks for your help and clarification.
Eric

Re: Getting an insight 6.1 executable on Cygwin

[Keith Seitz]

On Fri, 2004-07-09 at 12:56, E. Weddington wrote:
> > Hopefully the whole acceptance mess will get straightened out soon, and
> > I can spend some serious time fixing this and a host of other problems
> > creeping up in insight now.
> 
> I've only just heard about this. Do you have any knowledge of the status of 
> this?

Andrew Cagney (FSF gdb maintainer) is trying his best to get Red Hat and
the FSF to iron out any problems. He's been trying for months. I don't
really understand what the problem, and I'm not even sure I want to
know.

I may go back and try to rip insight away from tcl, tk, and itcl. At
least then I'll be able to use any version of tcl/tk, and I could make
RPMs for linux, and maybe I would pick up the package management on
cygwin.

Who knows. Maybe now that I'm talking about it, I'll go off and do
something, just for posterity's sake...

Keith

Re: Getting an insight 6.1 executable on Cygwin

[E. Weddington]

On 9 Jul 2004 at 13:03, Keith Seitz wrote:

> On Fri, 2004-07-09 at 12:56, E. Weddington wrote:
> > > Hopefully the whole acceptance mess will get straightened out soon, and I
> > > can spend some serious time fixing this and a host of other problems
> > > creeping up in insight now.
> > 
> > I've only just heard about this. Do you have any knowledge of the status of
> > this?
> 
> Andrew Cagney (FSF gdb maintainer) is trying his best to get Red Hat and
> the FSF to iron out any problems. He's been trying for months. I don't
> really understand what the problem, and I'm not even sure I want to
> know.

I'm sorry there's frustration with this situation. I can't imagine what the 
issue(s) might be. But who knows with the FSF/RMS.
 
> I may go back and try to rip insight away from tcl, tk, and itcl. At
> least then I'll be able to use any version of tcl/tk, and I could make
> RPMs for linux, and maybe I would pick up the package management on
> cygwin.

What do you mean by rip insight away from tcl, tk, and itcl? Do you mean away 
from the sourceware repository versions of these?

If you can make a version of Insight that can use any version tcl/tk, I would 
bet that others would be interested in this. I know that several people would 
love to be able to put together a GDB/Insight (usually for cross toolsets) with 
host=mingw and to get it *away* from requiring Cygwin. Having Insight require 
Cygwin's tcl/tk libs is a big PIA.

Eric

Re: Getting an insight 6.1 executable on Cygwin

[Keith Seitz]

On Fri, 2004-07-09 at 13:29, E. Weddington wrote:
 
> What do you mean by rip insight away from tcl, tk, and itcl? Do you mean away 
> from the sourceware repository versions of these?

Yes, right now, we rely on the sources version of tcl, tk, and itcl. We
are constantly having to do updates, and blah, blah, blah. This would
also allow us to make RPMs available for linux users.

> If you can make a version of Insight that can use any version tcl/tk, I would 
> bet that others would be interested in this. I know that several people would 
> love to be able to put together a GDB/Insight (usually for cross toolsets) with 
> host=mingw and to get it *away* from requiring Cygwin. Having Insight require 
> Cygwin's tcl/tk libs is a big PIA.

Unfortunately, it will be impossible to completely get away from cygwin:
gdb requires it. And since Insight requires gdb, insight is also a
cygwin app.

Of course, I've heard rumors that people are attempting to write
mingw-"native" extensions to gdb, but I admit that I've not followed it
too much.

In the end, I believe it's the "right thing to do". Since I'm the (sole)
maintainer (ignore gdb/MAINTAINERS), I can do whatever I want! :-P

Good luck with Insight. Hopefully better days lie ahead.

Keith

Re: Getting an insight 6.1 executable on Cygwin

[E. Weddington]

On 9 Jul 2004 at 13:38, Keith Seitz wrote:

BTW, my build just finished, and yes your patch works! Thanks!

> > If you can make a version of Insight that can use any version tcl/tk, I would
> > bet that others would be interested in this. I know that several people would
> > love to be able to put together a GDB/Insight (usually for cross toolsets)
> > with host=mingw and to get it *away* from requiring Cygwin. Having Insight
> > require Cygwin's tcl/tk libs is a big PIA.
> 
> Unfortunately, it will be impossible to completely get away from cygwin:
> gdb requires it. And since Insight requires gdb, insight is also a
> cygwin app.

I knew that insight->gdb->cygwin, but is that the only dependency? I'm sorry, I 
don't know enough about the architecture of Insight. 

In other words, if gdb could be built without cygwin, then is it true that 
insight could be built without cygwin?

Doing a cygcheck on the the insight executable, I get tcl84.dll depending on 
cygwin1.dll. I would have thought that this implies that insight, being the 
tcl/tk part, is directly dependent upon cygwin as well.
 
> Of course, I've heard rumors that people are attempting to write
> mingw-"native" extensions to gdb, but I admit that I've not followed it
> too much.

I haven't heard much either about a concerted effort to get gdb host=mingw.
 
> In the end, I believe it's the "right thing to do". Since I'm the (sole)
> maintainer (ignore gdb/MAINTAINERS), I can do whatever I want! :-P

More power to you!

Eric 

Re: Getting an insight 6.1 executable on Cygwin

[Keith Seitz]

On Fri, 2004-07-09 at 13:47, E. Weddington wrote:

> BTW, my build just finished, and yes your patch works! Thanks!

Excellent.

> I knew that insight->gdb->cygwin, but is that the only dependency? I'm sorry, I 
> don't know enough about the architecture of Insight. 
>
> In other words, if gdb could be built without cygwin, then is it true that 
> insight could be built without cygwin?

Yes, that's correct (if we used mingw versions of tcl et al).

> Doing a cygcheck on the the insight executable, I get tcl84.dll depending on 
> cygwin1.dll. I would have thought that this implies that insight, being the 
> tcl/tk part, is directly dependent upon cygwin as well.

Yes, but only because the tcl we use is cygwin-based. If gdb changed, we
could use a mingw version of tcl.
 
Keith

Re: Getting an insight 6.1 executable on Cygwin

[E. Weddington]

On 9 Jul 2004 at 13:56, Keith Seitz wrote:

> On Fri, 2004-07-09 at 13:47, E. Weddington wrote:
> 
> > I knew that insight->gdb->cygwin, but is that the only dependency? I'm sorry,
> > I don't know enough about the architecture of Insight. 
> >
> > In other words, if gdb could be built without cygwin, then is it true that
> > insight could be built without cygwin?
> 
> Yes, that's correct (if we used mingw versions of tcl et al).
> 
> > Doing a cygcheck on the the insight executable, I get tcl84.dll depending on
> > cygwin1.dll. I would have thought that this implies that insight, being the
> > tcl/tk part, is directly dependent upon cygwin as well.
> 
> Yes, but only because the tcl we use is cygwin-based. If gdb changed, we
> could use a mingw version of tcl.

Ok.... Now I'm confused.

If I go and get just the gdb package, it doesn't seem (to me anyway) to require 
tcl. If I get the insight package, then it requires tcl.

So why can't insight go ahead and use a mingw version of tcl, even when gdb 
still requires cygwin? I was under the assumption that the tcl/tk part of it 
insight was just for the GUI, which was separate, somehow, from the underlying 
gdb program. But I could be real wrong about this, as everything seems to get 
bundled into one executable and a couple of dlls for tcl and tk.

(I would bet that the answer has something to do with the architecture of 
insight that I don't know about.)

Thanks for taking the time to answer.
Eric

Re: Getting an insight 6.1 executable on Cygwin

[Keith Seitz]

On Fri, 2004-07-09 at 14:12, E. Weddington wrote:
> If I go and get just the gdb package, it doesn't seem (to me anyway) to require 
> tcl. If I get the insight package, then it requires tcl.

Right.

> So why can't insight go ahead and use a mingw version of tcl, even when gdb 
> still requires cygwin? I was under the assumption that the tcl/tk part of it 
> insight was just for the GUI, which was separate, somehow, from the underlying 
> gdb program. But I could be real wrong about this, as everything seems to get 
> bundled into one executable and a couple of dlls for tcl and tk.
> 
> (I would bet that the answer has something to do with the architecture of 
> insight that I don't know about.)

Nope. Cygwin. It's not nice to mix cygwin and msft runtimes. It's bad
juju. Since gdb requires cygwin, we use a cygwin version of tcl. If gdb
did not rely on cygwin, we would use a version of tcl that did not
require cygwin.

:-)
Keith

Re: Getting an insight 6.1 executable on Cygwin

[E. Weddington]

On 9 Jul 2004 at 14:15, Keith Seitz wrote:

> > So why can't insight go ahead and use a mingw version of tcl, even when gdb
> > still requires cygwin? I was under the assumption that the tcl/tk part of it
> > insight was just for the GUI, which was separate, somehow, from the underlying
> > gdb program. But I could be real wrong about this, as everything seems to get
> > bundled into one executable and a couple of dlls for tcl and tk.
> > 
> > (I would bet that the answer has something to do with the architecture of
> > insight that I don't know about.)
> 
> Nope. Cygwin. It's not nice to mix cygwin and msft runtimes. It's bad
> juju. 

Oh, right! D'uh.

>Since gdb requires cygwin, we use a cygwin version of tcl. If gdb
> did not rely on cygwin, we would use a version of tcl that did not
> require cygwin.
> 
> :-)

Thanks for clarifying the issue! 
Now it's time to talk to the mingw folks.....

Eric