gdb.ini vs. .gdbinit on Cygwin

gdb.ini vs. .gdbinit on Cygwin

[Corinna Vinschen]

Hi,

is there still a reason why the .gdbinit file is called gdb.ini
on Cygwin?  Are there any issues with that?  8.3 filesystems
are rather dead since Win95/NT3...

Corinna

-- 
Corinna Vinschen
Cygwin Developer
Red Hat, Inc.
mailto:vinschen@redhat.com

Re: gdb.ini vs. .gdbinit on cygwin

[Christopher Faylor]

On Sun, Nov 11, 2001 at 08:53:16PM +0100, Corinna Vinschen wrote:
>Hi,
>
>is there still a reason why the .gdbinit file is called gdb.ini
>on Cygwin?

No.

>Are there any issues with that?  8.3 filesystems are rather dead since
>Win95/NT3...

No.  Is it actually documented as gdb.ini for Windows?

I'd like to just nuke this behavior.

cgf

Re: gdb.ini vs. .gdbinit on cygwin

[Corinna Vinschen]

On Sun, Nov 11, 2001 at 04:01:50PM -0500, Christopher Faylor wrote:
> On Sun, Nov 11, 2001 at 08:53:16PM +0100, Corinna Vinschen wrote:
> >Hi,
> >
> >is there still a reason why the .gdbinit file is called gdb.ini
> >on Cygwin?
> 
> No.
> 
> >Are there any issues with that?  8.3 filesystems are rather dead since
> >Win95/NT3...
> 
> No.  Is it actually documented as gdb.ini for Windows?
> 
> I'd like to just nuke this behavior.

Yeah, me too.

Corinna

-- 
Corinna Vinschen
Cygwin Developer
Red Hat, Inc.
mailto:vinschen@redhat.com

Re: gdb.ini vs. .gdbinit on Cygwin

[Eli Zaretskii]

On Sun, 11 Nov 2001, Corinna Vinschen wrote:

> is there still a reason why the .gdbinit file is called gdb.ini
> on Cygwin?

You might consider the fact that the Windows Explorer still disallows
creation of files with a leading dot.

I don't know if this is important for Cygwin users, though.

Also, if you remove gdb.ini, I'd suggest to have a transitional period
where gdb.ini is still supported.  Otherwise, users might get mad at
you for breaking their setup.

In any case, please change the docs accordingly (type "i gdb.ini RET"
in an Info reader, to find where it is mentioned).

Re: gdb.ini vs. .gdbinit on Cygwin

[Fernando Nasser]

Eli Zaretskii wrote:
> 
> On Sun, 11 Nov 2001, Corinna Vinschen wrote:
> 
> > is there still a reason why the .gdbinit file is called gdb.ini
> > on Cygwin?
> 
> You might consider the fact that the Windows Explorer still disallows
> creation of files with a leading dot.
> 
> I don't know if this is important for Cygwin users, though.
> 
> Also, if you remove gdb.ini, I'd suggest to have a transitional period
> where gdb.ini is still supported.  Otherwise, users might get mad at
> you for breaking their setup.
> 
> In any case, please change the docs accordingly (type "i gdb.ini RET"
> in an Info reader, to find where it is mentioned).


I agree with Eli.  

My first thought was: look for ".gdbinit", if not found
look for "gdb.ini" after issuing a warning that "gdb.ini" should be
renamed to ".gdbinit" and that "gdb.ini" will not be supported in future
versions.

But then I remembered that we may have native Windows GDB (i.e., not
Cygwin,
but Win32 GDB programs).  Can those also look for ".gdbinit"?

And the Windows Explorer issue, has it gone on the XP version?


-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: gdb.ini vs. .gdbinit on Cygwin

[Corinna Vinschen]

On Tue, Nov 13, 2001 at 11:46:08AM -0500, Fernando Nasser wrote:
> Eli Zaretskii wrote:
> > 
> > On Sun, 11 Nov 2001, Corinna Vinschen wrote:
> > 
> > > is there still a reason why the .gdbinit file is called gdb.ini
> > > on Cygwin?
> > 
> > You might consider the fact that the Windows Explorer still disallows
> > creation of files with a leading dot.
> > 
> > I don't know if this is important for Cygwin users, though.
> > 
> > Also, if you remove gdb.ini, I'd suggest to have a transitional period
> > where gdb.ini is still supported.  Otherwise, users might get mad at
> > you for breaking their setup.
> > 
> > In any case, please change the docs accordingly (type "i gdb.ini RET"
> > in an Info reader, to find where it is mentioned).
> 
> 
> I agree with Eli.  
> 
> My first thought was: look for ".gdbinit", if not found
> look for "gdb.ini" after issuing a warning that "gdb.ini" should be
> renamed to ".gdbinit" and that "gdb.ini" will not be supported in future
> versions.
> 
> But then I remembered that we may have native Windows GDB (i.e., not
> Cygwin,
> but Win32 GDB programs).  Can those also look for ".gdbinit"?

We're talking about the Cygwin version.  A native DOS/Windows version
can keep gdb.ini.  It doesn't matter.

> And the Windows Explorer issue, has it gone on the XP version?

Of course not.  However, we should use .gdbinit the same way as for
any other host on Cygwin.  There's no reason to treat the Cygwin GDB
special in that case.  If you think it's really necessary, we can
keep supporting gdb.ini as well but a Cygwin GDB doesn't have to
take care for Explorer disabilities, IMO.  As a resort we could
begin to support an environment variable GDBINIT or similar which
contains the name/path of the GDB init file.  That would have the
advantage to be mostly host independent.

Corinna

-- 
Corinna Vinschen
Cygwin Developer
Red Hat, Inc.
mailto:vinschen@redhat.com

Re: gdb.ini vs. .gdbinit on Cygwin

[Eli Zaretskii]

> Date: Tue, 13 Nov 2001 11:46:08 -0500
> From: Fernando Nasser <fnasser@redhat.com>
> 
> But then I remembered that we may have native Windows GDB (i.e., not
> Cygwin,
> but Win32 GDB programs).  Can those also look for ".gdbinit"?

There's no problem whatsoever for native Windows programs, including a
Win32 GDB port, to access files with a leading dots in their names.
The fact that the Explorer disallows is simply a terrible misfeature
(if not a bug), but it has no technical reasons behind it.

So a Win32 port of GDB could easily use .gdbinit as well.

Re: gdb.ini vs. .gdbinit on Cygwin

[Eli Zaretskii]

> Date: Tue, 13 Nov 2001 18:04:34 +0100
> From: Corinna Vinschen <vinschen@redhat.com>
> 
> We're talking about the Cygwin version.  A native DOS/Windows version
> can keep gdb.ini.

That would be a mistake, I think.  For starters, how would we explain
in the manual which versions of GDB use what file name for the init
file, without confusing the users?  It would be painful enough to
tell that DOS versions use gdb.ini while Windows versions use
.gdbinit, since the DOS version runs on Windows as well.

> However, we should use .gdbinit the same way as for
> any other host on Cygwin.  There's no reason to treat the Cygwin GDB
> special in that case.

Yes, there is: Cygwin executables run on Windows, not on Unix or
GNU/Linux.

> If you think it's really necessary, we can keep supporting gdb.ini
> as well but a Cygwin GDB doesn't have to take care for Explorer
> disabilities, IMO.

I'd suggest to take those disabilities into consideration.  Users
will bump into this whether you want it or not; when they do, you
will be harming your users, not Microsoft.

Personally, I like Fernando's suggestion: look for .gdbinit, and if
not found, look for gdb.ini.  I think this would satisfy everybody.

As another data point, consider this: the Windows port of Emacs looks
for .emacs first, and if not found, looks for _emacs (the name used
on DOS).

> As a resort we could
> begin to support an environment variable GDBINIT or similar which
> contains the name/path of the GDB init file.  That would have the
> advantage to be mostly host independent.

??? But GDB doesn't support $GDBINIT on other platforms, so this
suggestion seems to actually _introduce_ OS-specific features.  Is
this really a good idea?

Re: gdb.ini vs. .gdbinit on Cygwin

[Fernando Nasser]

Corinna Vinschen wrote:
> 
> We're talking about the Cygwin version.  A native DOS/Windows version
> can keep gdb.ini.  It doesn't matter.
> 

As per Eli note, it can go as well.


> > And the Windows Explorer issue, has it gone on the XP version?
> 
> Of course not.  However, we should use .gdbinit the same way as for
> any other host on Cygwin.  There's no reason to treat the Cygwin GDB
> special in that case.

OK.

>  If you think it's really necessary, we can
> keep supporting gdb.ini as well 

We need it for backward compatibility.  Users may have a gdb.ini
right now and may not be aware of the change.


>but a Cygwin GDB doesn't have to
> take care for Explorer disabilities, IMO. 

OK with me.

> As a resort we could
> begin to support an environment variable GDBINIT or similar which
> contains the name/path of the GDB init file.  That would have the
> advantage to be mostly host independent.
> 

That would have to go through a public gdb discussion as it would 
affect other platforms as well.

And what we need is to accommodate users with an older setup and 
give them an incentive to switch.  The warning message (if no
.gdbinit was fond and a gdb.ini was found instead) accomplishes
both goals.

Eli, what do you think of that solution?  
Does it address your concerns?

-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: gdb.ini vs. .gdbinit on Cygwin

[Corinna Vinschen]

On Tue, Nov 13, 2001 at 07:52:50PM +0200, Eli Zaretskii wrote:
> > However, we should use .gdbinit the same way as for
> > any other host on Cygwin.  There's no reason to treat the Cygwin GDB
> > special in that case.
> 
> Yes, there is: Cygwin executables run on Windows, not on Unix or
> GNU/Linux.

That's right.  And GNU/Linux binaries don't run on Windows and most
flavours of Unix.  So what?

> Personally, I like Fernando's suggestion: look for .gdbinit, and if
> not found, look for gdb.ini.  I think this would satisfy everybody.

Which is a special treatment of Cygwin.

> As another data point, consider this: the Windows port of Emacs looks
> for .emacs first, and if not found, looks for _emacs (the name used
> on DOS).
> 
> > As a resort we could
> > begin to support an environment variable GDBINIT or similar which
> > contains the name/path of the GDB init file.  That would have the
> > advantage to be mostly host independent.
> 
> ??? But GDB doesn't support $GDBINIT on other platforms, so this
> suggestion seems to actually _introduce_ OS-specific features.  Is
> this really a good idea?

My suggestion was to introduce $GDBINIT as a generic way to set
the path to the init file.  So it's not OS dependent.

Corinna

Re: gdb.ini vs. .gdbinit on Cygwin

[Eli Zaretskii]

> Date: Tue, 13 Nov 2001 13:06:38 -0500
> From: Fernando Nasser <fnasser@redhat.com>
> 
> > As a resort we could
> > begin to support an environment variable GDBINIT or similar which
> > contains the name/path of the GDB init file.  That would have the
> > advantage to be mostly host independent.
> > 
> 
> That would have to go through a public gdb discussion as it would 
> affect other platforms as well.
> 
> And what we need is to accommodate users with an older setup and 
> give them an incentive to switch.  The warning message (if no
> .gdbinit was fond and a gdb.ini was found instead) accomplishes
> both goals.
> 
> Eli, what do you think of that solution?  

I generally don't like environment variables unless they are
absolutely necessary.  This doesn't seem to be the case.  Other
platforms don't need it.  Users will have to set the variable in
order to get the functionality.

It's simply a back compatibility issue, that's all.  I like your
suggestion, to support both the old and the new names, best.  It
seems to be silently doing the Right Thing, which is what I expect
from good solutions: you shouldn't be aware that there was a problem
to begin with ;-)

But I don't want to make this an issue between myself and the Cygwin
maintainers.  I just made a comment in the hope that it will be
useful; if the Cygwin maintainers have strong feelings about tossing
gdb.ini and never looking back, it's between them and the Cygwin
users.  Just be sure to document the change in the manual.

Re: gdb.ini vs. .gdbinit on cygwin

[Christopher Faylor]

On Tue, Nov 13, 2001 at 07:41:56PM +0200, Eli Zaretskii wrote:
>> Date: Tue, 13 Nov 2001 11:46:08 -0500
>> From: Fernando Nasser <fnasser@redhat.com>
>> 
>> But then I remembered that we may have native Windows GDB (i.e., not
>> Cygwin,
>> but Win32 GDB programs).  Can those also look for ".gdbinit"?
>
>There's no problem whatsoever for native Windows programs, including a
>Win32 GDB port, to access files with a leading dots in their names.
>The fact that the Explorer disallows is simply a terrible misfeature
>(if not a bug), but it has no technical reasons behind it.
>
>So a Win32 port of GDB could easily use .gdbinit as well.

Right.  gdb.ini was a bad choice for a name from the beginning.
Apparently whomeever did the original cygwin port was a little confused
about file names under Windows NT/95.

FWIW, I don't like the idea of modifying gdb to search for multiple
init files.  It seems rather intrusive a change to correct one host's
brain dead behavior.

If we decide to do this, then I don't think that it will be a big deal
for users to figure out the new name.  It's a momentary confusion, easily
rectified with a 'mv'.

cgf

Re: gdb.ini vs. .gdbinit on cygwin

[Stan Shebs]

Christopher Faylor wrote:
> 
> Right.  gdb.ini was a bad choice for a name from the beginning.
> Apparently whomeever did the original cygwin port was a little confused
> about file names under Windows NT/95.

It was originally set up for DOS (Cygnus had a now-long-forgotten
DOS tools business...), and retained into Windows out of inertia and
some desire for continuity.  Not much reason to keep it any longer.

Stan