Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Corinna Vinschen

On Mon, Aug 12, 2002 at 06:45:32PM -0400, Joe Buehler wrote:
 Joe Buehler wrote:
 
 New GNU emacs packages are available for upload to sources.redhat.com at:
 
 http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs/emacs-21.2-3-src.tar.bz2
 http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs/emacs-21.2-3.tar.bz2
 http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs/setup.hint
 http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-el/emacs-el-21.2-3.tar.bz2
 http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-el/setup.hint
 http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-X11/emacs-X11-21.2-3.tar.bz2
 http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-X11/setup.hint
 
 This version fixes tty subprocess functionality: telnet, ange-ftp, gnus and 
 the like.

Uploaded.  Please add a note to your announcement that the deinstallation
of the old package will drop ctags exectuables so that a reinstall of that
package would be necessary.

Thanks,
Corinna

-- 
Corinna Vinschen  Please, send mails regarding Cygwin to
Cygwin Developermailto:[EMAIL PROTECTED]
Red Hat, Inc.



Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Earnie Boyd

Corinna Vinschen wrote:
 
 On Mon, Aug 12, 2002 at 06:45:32PM -0400, Joe Buehler wrote:
  Joe Buehler wrote:
 
  New GNU emacs packages are available for upload to sources.redhat.com at:
 
  http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs/emacs-21.2-3-src.tar.bz2
  http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs/emacs-21.2-3.tar.bz2
  http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs/setup.hint
  http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-el/emacs-el-21.2-3.tar.bz2
  http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-el/setup.hint
  http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-X11/emacs-X11-21.2-3.tar.bz2
  http://68.100.179.154:3000/cygwin/emacs-21.2-3/emacs-X11/setup.hint
 
  This version fixes tty subprocess functionality: telnet, ange-ftp, gnus and
  the like.
 
 Uploaded.  Please add a note to your announcement that the deinstallation
 of the old package will drop ctags exectuables so that a reinstall of that
 package would be necessary.
 

Would it drop ctags with

requires: ctags cygwin

in the hint file?

Earnie.



Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread jbuehler

Corinna Vinschen wrote:

Uploaded.  Please add a note to your announcement that the deinstallation
of the old package will drop ctags exectuables so that a reinstall of that
package would be necessary.

Corinna, maybe I misunderstand, but only the main package for
emacs-21.2-3 is on sources.redhat.com.  The others are not there.
I split the single original package into several, like XFree86
does it, with subdirectories.  One of my concerns was whether
I did that correctly.

Joe Buehler



Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Corinna Vinschen

On Tue, Aug 13, 2002 at 08:05:13AM -0400, [EMAIL PROTECTED] wrote:
 Corinna Vinschen wrote:
 
 Uploaded.  Please add a note to your announcement that the deinstallation
 of the old package will drop ctags exectuables so that a reinstall of that
 package would be necessary.
 
 Corinna, maybe I misunderstand, but only the main package for
 emacs-21.2-3 is on sources.redhat.com.  The others are not there.

Yes, they are.  release/emacs/emacs-el, release/emacs/emacs-X11.

Corinna

-- 
Corinna Vinschen  Please, send mails regarding Cygwin to
Cygwin Developermailto:[EMAIL PROTECTED]
Red Hat, Inc.



Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Corinna Vinschen

On Tue, Aug 13, 2002 at 07:50:32AM -0400, Nicholas Wourms wrote:
 Earnie Boyd wrote:
 Would it drop ctags with
 
 requires: ctags cygwin
 
 in the hint file?
 
 Yes, because the ctags version hasn't been bumped, so the installer 
 thinks it is still there eventhough the removal of the old emacs package 
 deletes them.  The new package, of course, doesn't reinstall them.

Hmm, perhaps it *is* better to bump the version?!?

Corinna

-- 
Corinna Vinschen  Please, send mails regarding Cygwin to
Cygwin Developermailto:[EMAIL PROTECTED]
Red Hat, Inc.



Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Robert Collins

On Tue, 2002-08-13 at 22:26, Corinna Vinschen wrote:

 Hmm, perhaps it *is* better to bump the version?!?

Yes.

Rob




signature.asc
Description: This is a digitally signed message part


Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Corinna Vinschen

On Tue, Aug 13, 2002 at 10:32:13PM +1000, Robert Collins wrote:
 On Tue, 2002-08-13 at 22:28, Robert Collins wrote:
  On Tue, 2002-08-13 at 22:26, Corinna Vinschen wrote:
  
   Hmm, perhaps it *is* better to bump the version?!?
  
  Yes.
 
 But it won't help. Sorry.
 The reason is that ctags will get reinstalled *first*, leaving emacs to
 still break the user environment.

Doesn't setup first deinstall all old and then install all new
packages? In that case everything should be fine.

Corinna

-- 
Corinna Vinschen  Please, send mails regarding Cygwin to
Cygwin Developermailto:[EMAIL PROTECTED]
Red Hat, Inc.



Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Nicholas Wourms

Corinna Vinschen wrote:

On Tue, Aug 13, 2002 at 10:32:13PM +1000, Robert Collins wrote:

On Tue, 2002-08-13 at 22:28, Robert Collins wrote:

On Tue, 2002-08-13 at 22:26, Corinna Vinschen wrote:


Hmm, perhaps it *is* better to bump the version?!?

Yes.

But it won't help. Sorry.
The reason is that ctags will get reinstalled *first*, leaving emacs to
still break the user environment.


Doesn't setup first deinstall all old and then install all new
packages? In that case everything should be fine.

It does, I think Robert's having a wee bit of memory loss.  He better 
look into some parity memory :-D.

Cheers,
Nicholas




Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Nicholas Wourms

Robert Collins wrote:

On Tue, 2002-08-13 at 23:22, Nicholas Wourms wrote:


It does, I think Robert's having a wee bit of memory loss.  He better 
look into some parity memory :-D.


An upgrade is != to an uninstall and install. Setup treats upgrades as
atomic actions per package.

Then why is it that when I upgrade, it always says it is removing all 
the files from each previous packages?  It does this for each package 
I've got checked to upgrade.  Only after it finishes that, does it 
install the upgraded packages.  Either the dialogue lies or something 
isn't right with my installer.

Cheers,
Nicholas




Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Robert Collins

On Wed, 2002-08-14 at 00:11, Nicholas Wourms wrote:
 Robert Collins wrote:
 
 On Wed, 2002-08-14 at 00:02, Nicholas Wourms wrote:
 
   
 
 Well that may be the way it should be, but the reality of the situation 
 is this:
 
 
 
 Check the source luke. 
 
 Source != Reality

Ha!. Source released under the GPL had d**n well better == reality, or
we're breaking the licence.
 
 Why on earth would setup be saying it is uninstalling but not really 
 uninstalling?  

Is is uninstalling, but not as an uninstall, rather as the first part of
an upgrade. I've already said this in this conversation.

 It sure does act like it is uninstalling, including
 long pauses of activity for the larger packages.  And if your
 assertions were correct, wouldn't the uninstall of the next package
 happen right after the install of the previously updated package? 

Yes. And it should. You've also prodded me into finding a bug. Thanks.

 AFAICT, this is not what is happening.  Be aware that I'm using the
 release version of setup, not HEAD or a snapshot.

It's been like this for 3 full releases. See install.cc. 
See your log files. They'll show I was completely wrong!

FYI there are three loops in install.cc.
One for uninstalls
One for inplace replacements 
One for installs.

The uninstall checks is coded wrongly and triggering on replacements.

Doh.

So emacs will squeeze by for now.

Rob



signature.asc
Description: This is a digitally signed message part


Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Nicholas Wourms

Robert Collins wrote:

On Wed, 2002-08-14 at 00:11, Nicholas Wourms wrote:

Robert Collins wrote:


On Wed, 2002-08-14 at 00:02, Nicholas Wourms wrote:

 


Well that may be the way it should be, but the reality of the situation 
is this:
   


Check the source luke. 


Source != Reality


Ha!. Source released under the GPL had d**n well better == reality, or
we're breaking the licence.

What I mean to say is that, despite one's best efforts, compiled source 
doesn't always behave as one had intended.  Of course it will act as it 
is written, it just may not seem apparent that the way it acts != the 
way you intended it to act.

 

Why on earth would setup be saying it is uninstalling but not really 
uninstalling?  


Is is uninstalling, but not as an uninstall, rather as the first part of
an upgrade. I've already said this in this conversation.

I wasn't disputing the fact that this is the way you intended it.

It sure does act like it is uninstalling, including
long pauses of activity for the larger packages.  And if your
assertions were correct, wouldn't the uninstall of the next package
happen right after the install of the previously updated package? 


Yes. And it should. You've also prodded me into finding a bug. Thanks.

Why is this behaviour considiered a bug?  It seems quite logical to me. 
 It allows for the usage that Corinna had desired.  Otherwise, there is 
no way to garuntee that updating a package later in the alphabet won't 
trump an earlier update's install.  Pretty handy when you want to fix a 
conflicting package f*$kup.


AFAICT, this is not what is happening.  Be aware that I'm using the
release version of setup, not HEAD or a snapshot.


It's been like this for 3 full releases. See install.cc. 
See your log files. They'll show I was completely wrong!

FYI there are three loops in install.cc.
One for uninstalls
One for inplace replacements 
One for installs.

The uninstall checks is coded wrongly and triggering on replacements.

Doh.

So emacs will squeeze by for now.

Well, I guess given the fact that this is not a permanent feature, the 
whole point is rather moot [I would, however, suggest that this become a 
permanent feature...].  People will just have to remeber to reinstall ctags.

Cheers,
Nicholas




Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Robert Collins

On Wed, 2002-08-14 at 00:34, Nicholas Wourms wrote:
 Robert Collins wrote:
 
 What I mean to say is that, despite one's best efforts, compiled source 
 doesn't always behave as one had intended.  Of course it will act as it 
 is written, it just may not seem apparent that the way it acts != the 
 way you intended it to act.

Now that, that I agree with.
 

 Yes. And it should. You've also prodded me into finding a bug. Thanks.
 
 Why is this behaviour considiered a bug?  It seems quite logical to me. 
  It allows for the usage that Corinna had desired.  Otherwise, there is 
 no way to garuntee that updating a package later in the alphabet won't 
 trump an earlier update's install.  Pretty handy when you want to fix a 
 conflicting package f*$kup.

Because it's actually worse. The cause of the conflicting package f***up
is setup not checking for conflicts. So that can and will be addressed
in other fashions.

The reason to make upgrades atomic and done one package at a time is
to deal with cases like the following:
A pre-removal script (which *should* trigger on upgrades) may require
binaries from another package being upgraded. Unless that other package
is installed again at the time of the pre-removal script triggering, bad
things will happen.

Rob



signature.asc
Description: This is a digitally signed message part


Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Gareth Pearce



Argh - sorry bad gareth sent copys to authors as well as list.

Gareth - adds to your spam by appologising :P.

_
Join the world’s largest e-mail service with MSN Hotmail. 
http://www.hotmail.com




Re: GNU emacs 21.2-3 packages available

2002-08-13 Thread Robert Collins

On Wed, 2002-08-14 at 12:41, Gareth Pearce wrote:

 
 alphabetical order can still obviously screw this up anyway, no way to work 
 this perfectly until the full versioned dependency set comes in - 
 pre-removal-depends post-install-depends ... etc.

That level of depends tracking is not needed. The depends is versioned
in HEAD already. 
 
 Hope someone has a nice algorithm for multi-dependency ordering satifaction 
 and circular dependency detection.

It's a DAG, and there are libraries for that. Still it is an excercise.

Rob



signature.asc
Description: This is a digitally signed message part