Re: Insight on Windows NT w/cygwin

Re: Insight on Windows NT w/cygwin

[MWCruess]

Thanks for the advice.  I was thinking that it must
be something about the tcl code.  I did try building
the whole thing from scratch with no optimization and
the behavior was still odd (in the sense that the
newly built gdb would just mysteriously exit, the
gdb console window would print a message that the
program exited normally.

I also saw some behavior in which it made a diference
if I built a program from a directory I got to with
"cd /usr/play" or "/cygdrive/d/usr/play".

Thanks,
Michael

Re: Insight on Windows NT w/cygwin

[Chris Faylor]

On Fri, Jun 09, 2000 at 02:54:26PM -0400, MWCruess@cs.com wrote:
>Thanks for the advice.  I was thinking that it must be something about
>the tcl code.  I did try building the whole thing from scratch with no
>optimization and the behavior was still odd (in the sense that the
>newly built gdb would just mysteriously exit, the gdb console window
>would print a message that the program exited normally.

I just fixed another problem in libgui.c which also resulted in the "silent
exit" behavior.  I'm waiting for approval on the checkin now.  I should have
this checked in shortly.

>I also saw some behavior in which it made a diference if I built a
>program from a directory I got to with "cd /usr/play" or
>"/cygdrive/d/usr/play".

Unfortunately, insight has always had only minimal ability to recognize
Cygwin paths.  This is due to the fact that the tcl and tk DLLs are 98%
pure Windows and only the bare minimum support for cygwin paths required
to search for things like init.tcl, etc.

cgf

Re: Insight on Windows NT w/cygwin

[John A. Turner]

>>>>> "CF" == Chris Faylor <cgf@cygnus.com>:

CF> Unfortunately, insight has always had only minimal ability to recognize
CF> Cygwin paths.  This is due to the fact that the tcl and tk DLLs are 98%
CF> pure Windows and only the bare minimum support for cygwin paths
CF> required to search for things like init.tcl, etc.

could someone discuss the pros and cons of using a fully cygwin-aware
version of tcl/tk?  seems to me that's what most people using insight on
cygwin would want anyway

I just know that after using a native build of XEmacs for some time, I was
much happier when I started using a cygwin build - seems it would be
similar with insight

-- 
John A. Turner, Senior Research Associate
Blue Sky Studios, One South Rd, Harrison, NY 10528
http://www.blueskystudios.com/      (914) 825-8319

Re: Insight on Windows NT w/cygwin

[Chris Faylor]

On Mon, Jun 12, 2000 at 02:03:32PM -0400, John A. Turner wrote:
>>>>>> "CF" == Chris Faylor <cgf@cygnus.com>:
>CF> Unfortunately, insight has always had only minimal ability to recognize
>CF> Cygwin paths.  This is due to the fact that the tcl and tk DLLs are 98%
>CF> pure Windows and only the bare minimum support for cygwin paths
>CF> required to search for things like init.tcl, etc.
>
>could someone discuss the pros and cons of using a fully cygwin-aware
>version of tcl/tk?  seems to me that's what most people using insight on
>cygwin would want anyway
>
>I just know that after using a native build of XEmacs for some time, I was
>much happier when I started using a cygwin build - seems it would be
>similar with insight

Pro: insight would be able to use all UNIX-like features of Cygwin.

Con: Somebody has to do it.

cgf

Re: Insight on Windows NT w/cygwin

[Syd Polk]

At 03:30 PM 6/12/00 -0400, Chris Faylor wrote:
>On Mon, Jun 12, 2000 at 02:03:32PM -0400, John A. Turner wrote:
> >>>>>> "CF" == Chris Faylor <cgf@cygnus.com>:
> >CF> Unfortunately, insight has always had only minimal ability to recognize
> >CF> Cygwin paths.  This is due to the fact that the tcl and tk DLLs are 98%
> >CF> pure Windows and only the bare minimum support for cygwin paths
> >CF> required to search for things like init.tcl, etc.
> >
> >could someone discuss the pros and cons of using a fully cygwin-aware
> >version of tcl/tk?  seems to me that's what most people using insight on
> >cygwin would want anyway
> >
> >I just know that after using a native build of XEmacs for some time, I was
> >much happier when I started using a cygwin build - seems it would be
> >similar with insight
>
>Pro: insight would be able to use all UNIX-like features of Cygwin.
>
>Con: Somebody has to do it.

Other con: People using Insight for native Windows development could get 
confused.


Syd Polk		spolk@redhat.com
Engineering Manager	+1 415 777 9810 x 241
Red Hat, Inc.

Re: Insight on Windows NT w/cygwin

[Chris Faylor]

On Mon, Jun 12, 2000 at 12:33:32PM -0700, Syd Polk wrote:
>At 03:30 PM 6/12/00 -0400, Chris Faylor wrote:
>>On Mon, Jun 12, 2000 at 02:03:32PM -0400, John A. Turner wrote:
>> >>>>>> "CF" == Chris Faylor <cgf@cygnus.com>:
>> >CF> Unfortunately, insight has always had only minimal ability to recognize
>> >CF> Cygwin paths.  This is due to the fact that the tcl and tk DLLs are 98%
>> >CF> pure Windows and only the bare minimum support for cygwin paths
>> >CF> required to search for things like init.tcl, etc.
>> >
>> >could someone discuss the pros and cons of using a fully cygwin-aware
>> >version of tcl/tk?  seems to me that's what most people using insight on
>> >cygwin would want anyway
>> >
>> >I just know that after using a native build of XEmacs for some time, I was
>> >much happier when I started using a cygwin build - seems it would be
>> >similar with insight
>>
>>Pro: insight would be able to use all UNIX-like features of Cygwin.
>>
>>Con: Somebody has to do it.
>
>Other con: People using Insight for native Windows development could get 
>confused.

We've tried to minimize incompatibilities with recent cygwin DLLs such
that a path which contains a colon or backslash is assumed to be a pure
Windows path and is untouched by the cygwin path-handling code.

This was because we'd had some complaints from people who didn't care
about cygwin and just wanted to compile and debug their programs using
standard Windows pathnames.  This seemed like a good compromise, although
gdb in general doesn't really like to use \ as a path separator.

Of course, now we are seeing complaints in the cygwin mailing list
from people who had routinely done things like

#include "myincludedir\\bob.h"

and are concerned that they can no longer find bob.h...

cgf

Re: Insight on Windows NT w/cygwin

[MWCruess]

OK.  Thanks.  I did get to work by doing a "make install".  In the past
I had always tested gdb builds in the build directory.  Following up on
your suggestion about the paths I went ahead and did the install and it
came up.  It seems to be slower than your precompiled binaries, but
it works.  I was also able to build the version that I am really interested
in for one of our microcontroller boards.  Unfortunately, I now have a COM
port problem and can't talk to the board...

Thanks for your assistance.  I'll download the newer version with your
changes Monday.

Michael

Insight on Windows NT w/cygwin

[MWCruess]

I have am having trouble building a working version of Insight for a
Windows NT host.  Insight builds, but when I run it nothing appears
to happen.  I just get a shell prompt.  When I use the -nw option the
command line interface seems to work fine.

I downloaded and installed a completely new cygwin environment a
few days ago.  The binary of gdb/insight that comes with cygwin works,
although debugging the binary that I built gives odd results.  The program
under debug will suddenly exit with no indication of what happened.  When
single-stepping I have seen a function entered with the return statement
highlighted, then one more step causes an earlier statement to be
highlighted.  The behavior changes from run to run on the same binary.
This has me wondering if I have a compiler problem.

If any of this sounds familiar, I would appreciate some advice.  The Solaris
version built first time and runs great, but I don't have much experience
with the Windows system.

Thanks,
Michael

Re: Insight on Windows NT w/cygwin

[Chris Faylor]

On Thu, Jun 08, 2000 at 10:57:57PM -0400, MWCruess@cs.com wrote:
>I have am having trouble building a working version of Insight for a
>Windows NT host.  Insight builds, but when I run it nothing appears to
>happen.  I just get a shell prompt.  When I use the -nw option the
>command line interface seems to work fine.
>
>I downloaded and installed a completely new cygwin environment a few
>days ago.  The binary of gdb/insight that comes with cygwin works,
>although debugging the binary that I built gives odd results.  The
>program under debug will suddenly exit with no indication of what
>happened.  When single-stepping I have seen a function entered with the
>return statement highlighted, then one more step causes an earlier
>statement to be highlighted.  The behavior changes from run to run on
>the same binary.  This has me wondering if I have a compiler problem.

I've just uploaded a new gdb binary to cygwin/latest/gdb.  I don't know
if it will work any better for you but what you are describing sounds
like standard behavior when attempting to debug a program that was
compiled with -O and -g.  If you drop the optimization from your compiles
I'll bet things will work better.

The problem of gdb returning silently to the windows prompt is usually
due to an inability to find init.tcl in any of the standard places that
gdb normally searches.  I have recently checked in a patch which should
make gdb a little smarter, at least with regard to standard cygwin
locations for this file.  This is in CVS now, is also in the source tar
ball on sourceware, and should show up in the next gdb/insight snapshot.

>If any of this sounds familiar, I would appreciate some advice.  The
>Solaris version built first time and runs great, but I don't have much
>experience with the Windows system.

If you have further problems the people at the cygwin@sourceware.cygnus.com
mailing list will probably be able to offer more insight.  No pun
intended.

cgf