Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Charles Wilson]

Christopher Faylor wrote:
> I've just done this for lzma to prove that it works and it does.

Great! I'll make that change for each of the packages I've already
"forked" for cygwin-1.7

> But then, I don't think it makes a lot of sense to be doing much package
> reorganization in 1.5 since it is approaching the end of its life.  When
> we shuffle things around in the 1.5 release we increase the possibility
> of annoying support problems when 1.7 is released.

Well, that's really off-point, because...

> IMO, we should be moving towards stabilization of the 1.5 package base
> and the 1.7 Cygwin DLL code base.  I don't see either of those happening
> right now, though.

... the problem in *this* case was, even if I had left 1.5 completely
alone, and updated only lzma/xz for cygwin-1.7, then we'd have "package
reorganization" in the *cygwin-1.7* area, not cygwin-1.5 (AND, the
problems that started this thread were from mistakes in the
cygwin-1.7/lzma/setup.hint -- the cygwin-1.5 version was fine.)  Worse,
if we were to *also* enforce the rule about "no obsolete packages in
cygwin-1.7", then this sort of package reorganization is exactly the
thing that we would not be able to automate in cygwin-1.7; you could
ONLY automate it in cygwin-1.5!

I can see re-wording the rule to be: "no _obsolete packages in
cygwin-1.7 --- that WERE ALREADY obsolete at the time the release-2
repository was first set up". THAT's just common sense.

However, if I "shuffle things around" only in cygwin-1.7 -- because it's
privileged and gets the new xz package, but cygwin-1.5 is the red-headed
stepchild and doesn't -- then current users who are sticking with
cygwin-1.5 *for now* will see a LOT of immediate package changes (like
"where'd my lzma go? what's this xz thing?") over and above the expected
transition issues related to the DLL itself.  To me -- since I have to
support the lzma and xz packages during that transition -- it makes
sense to try to keep both -1.5 and -1.7 in sync -- at least until the
official -1.7 release.  That way, people can work thru whatever little
kinks may happen, related to my insignificant packages, over some course
of time, rather than all at once AND bound up together with each user's
growning pains with their personal -1.7 transition.   Therefore, if I
choose to "shuffle things around" in cygwin-1.7...then I'm gonna do the
same "shuffle" on cygwin-1.5.

But that's just me.

Anyway, regarding cygwin-1.5, hasn't it been the policy that after
cygwin-1.7 is released, then maintainance of cygwin-1.5 packages is at
the package maintainer's discretion?  Calling a partial "slush" on
cygwin-1.5 packages -- /when cygwin-1.7 hasn't even been officially
release/ -- runs counter to that.

--
Chuck

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Christopher Faylor]

On Mon, Mar 23, 2009 at 02:02:00PM -0400, Charles Wilson wrote:
>Christopher Faylor wrote:
>> I've just done this for lzma to prove that it works and it does.
>
>Great! I'll make that change for each of the packages I've already
>"forked" for cygwin-1.7

Too bad you couldn't just stop there.

>> But then, I don't think it makes a lot of sense to be doing much package
>> reorganization in 1.5 since it is approaching the end of its life.  When
>> we shuffle things around in the 1.5 release we increase the possibility
>> of annoying support problems when 1.7 is released.
>
>Well, that's really off-point, because...

No, actually it's not.

But, a digression...

With your suggestion of an upset bug, I spent 15-20 minutes of my time
to debug and attempt to fix a problem and discovered a simple cockpit
error on your part.  Yet in your two voluminous responses on this
subject you haven't bothered to include a "Thank you" or "I'm sorry to
have bothered you".

I probably would just let this slide if this interchange hadn't followed
fairly closely on the heels of you publicly scolding me for not showing
the proper appreciation to someone whom you thought had provided a
patch.

I sense a comportment disparity here.

End digression...

So now that I'm not digressing, I'm disheartened that I have to argue
for code stabilization prior to a major release point.  I fully
understand that no one likes having to exercise discipline around a
release.  I don't enjoy waiting or managing branches myself.  And, I can
understand that most people would not voluntarily adopt a release
"freeze" if no one else was doing it.  So, since there was no policy in
place, there was no reason for any package maintainer to be thinking
along those lines.  I didn't think I'd have to argue for the wisdom of
such a proven release policy, though.

But, I can't say that I actually care that much.  My opinion was
indirectly solicited so I offered it.  The bottom line is that, no, I'm
not going to change release-2 from being a unionfs.  We can easily and
slowly transition in this direction without the requirement of a
wholesale switch.

I have not yet decided if sourceware will continue to run a copy of
upset on the old Cygwin 1.5 release directory when 1.7 goes live.  I
suspect that it will not.

cgf

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Charles Wilson]

Christopher Faylor wrote:
> On Mon, Mar 23, 2009 at 02:02:00PM -0400, Charles Wilson wrote:
>> Well, that's really off-point, because...
> 
> No, actually it's not.
> 
> But, a digression...
> you haven't bothered to include a "Thank you" or "I'm sorry to
> have bothered you".

You're right.  Thank you.  Sorry to have bothered you.

> So now that I'm not digressing, I'm disheartened that I have to argue
> for code stabilization prior to a major release point.  I fully
> understand that no one likes having to exercise discipline around a
> release.  I don't enjoy waiting or managing branches myself.  And, I can
> understand that most people would not voluntarily adopt a release
> "freeze" if no one else was doing it.  So, since there was no policy in
> place, there was no reason for any package maintainer to be thinking
> along those lines.  I didn't think I'd have to argue for the wisdom of
> such a proven release policy, though.

If we really -- really and truly, actually, honest-to-god
swear-on-the-bible -- intend to release cygwin-1.7 real soon now (and
not "Real Soon Now(tm)" -- then by all means, let's go to package
freeze.  (Notwithstanding Yaakov's desire to do a flag day,
backwards-incompatible, "rebuild everything with the spiffy new gcc4 for
the spiffy new cygwin-1.7)

But I'm game for a freeze. The reason I've been doing my flurry of
rebuilds was specifically to prepare FOR the cygwin-1.7 release.  But if
we go to package freeze, I've got plenty of other things to do with my time.

> But, I can't say that I actually care that much.  My opinion was
> indirectly solicited so I offered it.  The bottom line is that, no, I'm
> not going to change release-2 from being a unionfs.  We can easily and
> slowly transition in this direction without the requirement of a
> wholesale switch.

OK.  I didn't complain about the unionfs -- I only described *my*
mistake in working *with* unionfs. And now that you've pointed how I
should have been doing it all along, I doubt I'll have any more hassles
along that line.

> I have not yet decided if sourceware will continue to run a copy of
> upset on the old Cygwin 1.5 release directory when 1.7 goes live.  I
> suspect that it will not.

At the risk of continuing a contentious thread, is this is a policy
change?  I'm honestly not sure:

Corinna said (2008 Apr 24) (in response to Yaakov's question):
>>* for how long will maintainers be expected to maintain two sets of
>> releases?
>
> Until we are all more or less confident that we can release 1.7 and
> its packages to the world.

Eric Blake issued his opinion:
> IMHO, only until 1.7 is released.  After that, the 1.5 tree will be left
> to grow weeds and drift off into the mists of time.  Of course, if you
> feel like releasing updates into the 1.5 tree (e.g. security patches)
> I'm sure that would be fine, but I don't think there's any expectation
> of service once 1.7 is finally out the door and made current.  But this
> is all just MHO, others may have a different view.

I thought there was some mention of supporting updates in the (legacy)
release/ area -- especially security fixes -- at the maintainer's
discretion.  But (a) if upset doesn't regen setup.ini, then the
maintainer can upload new packages all day long to no effect, and (b) I
can't find any actual statement in the archives supporting my recollection.

So, maybe "once 1.7 goes live, 1.5 is dead. deceased. joined the choir
invisible. frozen solid and kept for historical interest only. Want
newer stuff, security fixes, bug fixes?  Leave win9x and use 1.7"
is/was/has-been the policy all along and doesn't represent a change.
Can you or Corinna clarify that?

--
Chuck

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Charles Wilson wrote:
> If we really -- really and truly, actually, honest-to-god
> swear-on-the-bible -- intend to release cygwin-1.7 real soon now (and
> not "Real Soon Now(tm)" -- then by all means, let's go to package
> freeze.  (Notwithstanding Yaakov's desire to do a flag day,
> backwards-incompatible, "rebuild everything with the spiffy new gcc4 for
> the spiffy new cygwin-1.7)

I've already began this with my packages, and in doing so I have come
across a number of bugs which I have been reporting to the list as I
find them.  How else do you propose to properly test cygwin-1.7 until
packages have been rebuilt for it?


Yaakov

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAknLChsACgkQpiWmPGlmQSNzpACfTC4iK9VRignrD+UomU44HrJs
BqQAoN3814Qdy5/SO+rQvFRV7CY/0KoY
=cgkb
-----END PGP SIGNATURE-----

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Christopher Faylor]

On Wed, Mar 25, 2009 at 11:52:43PM -0500, Yaakov (Cygwin/X) wrote:
>Charles Wilson wrote:
>>If we really -- really and truly, actually, honest-to-god
>>swear-on-the-bible -- intend to release cygwin-1.7 real soon now (and
>>not "Real Soon Now(tm)" -- then by all means, let's go to package
>>freeze.  (Notwithstanding Yaakov's desire to do a flag day,
>>backwards-incompatible, "rebuild everything with the spiffy new gcc4
>>for the spiffy new cygwin-1.7)
>
>I've already began this with my packages, and in doing so I have come
>across a number of bugs which I have been reporting to the list as I
>find them.  How else do you propose to properly test cygwin-1.7 until
>packages have been rebuilt for it?

I'm not sure if this is what you're asking but I was talking about a
package freeze for 1.5, not 1.7.

cgf

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Christopher Faylor wrote:
> I'm not sure if this is what you're asking but I was talking about a
> package freeze for 1.5, not 1.7.

As long we won't be expected to support the 1.5 packages for too much
longer, that's fine with me.


Yaakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAknLHhMACgkQpiWmPGlmQSN82QCeLKsOwd1gc1N/t/kf+FocQPDk
aXsAn0Et7uJEEb9TPCWLg26B0yh0tI8W
=l3VY
-----END PGP SIGNATURE-----

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Dave Korn]

Christopher Faylor wrote:

> Stop obsoleting packages and adding major new package releases.  In
> fact, I'd say stop anything but serious bugfixes or trivial upgrades to
> 1.5 packages at this point.

  I was considering doing one final release for 1.5 so they can have default
shared-libgcc and default v4 alternatives, but I'm undecided.

    cheers,
      DaveK

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Corinna Vinschen]

On Mar 23 10:52, Christopher Faylor wrote:
> On Mon, Mar 23, 2009 at 03:33:26PM +0100, Corinna Vinschen wrote:
> >On Mar 23 10:18, Christopher Faylor wrote:
> >>On Mon, Mar 23, 2009 at 08:29:33AM -0500, Yaakov (Cygwin/X) wrote:
> >>>Corinna Vinschen wrote:
> >>>>Isn't it a good time to split the release and release-2 areas into two
> >>>>separate directories?  Now that many maintainers create separate
> >>>>packages for 1.5 and 1.7, updating 1.5 and 1.7 directories and always
> >>>>having to delete the 1.5 stuff from the 1.7 release-2 directory
> >>>>afterwards is troublesome and bound to produce mistakes.  And the union
> >>>>of the two directories makes increasingly less sense.
> >>>
> >>>+1
> >>
> >>For my packages, I still want the convenience of the unionfs.  [...]
> >>IMO, we should be moving towards stabilization of the 1.5 package base
> >>and the 1.7 Cygwin DLL code base.  I don't see either of those
> >>happening right now, though.
> >
> >I don't understand what you mean.  What package stabilization has to be
> >done in 1.5?
> 
> Stop obsoleting packages and adding major new package releases.  In
> fact, I'd say stop anything but serious bugfixes or trivial upgrades to
> 1.5 packages at this point.

Sure.  That's what we get practically automatically by splitting the
unionfs into two separate directories, plus asking maintainers to
concentrate on 1.7 from now on.

> >And the Cygwin DLL is quite stable now.  I'm only fixing bugs and
> >changing the wide char/multibyte
> 
> "wide char/multibyte" - that's a huge change to be adding to a product
> which is supposed to be in beta.  It's worth a separate beta all by
> itself.

I planned to have the wide char/multibyte stuff in 1.7 from the
beginning.  I expected more help with this to begin with, but
unfortunately I had to do it alone so it took more time than I had hoped
for.  When the newlib patch I created yesterday is approved, then that,
together with the accompanying cygwin patch which is waiting locally, is
the last big change in this area.

> In fact, as I have previously stated, this extended 1.7 release should
> have, IMO, been a series of 1.7.x releases where the base functionality
> was laid in 1.7.0 and improvements showed up at a regular basis in 1.7.1,
> 1.7.2, etc.  That mirrors the way that Linux does things and it's one of
> the reasons why Cygwin adopted a similar versioning strategy.

After 1.7.1 is released, the old scheme will be used again.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Christopher Faylor]

On Mon, Mar 23, 2009 at 03:33:26PM +0100, Corinna Vinschen wrote:
>On Mar 23 10:18, Christopher Faylor wrote:
>>On Mon, Mar 23, 2009 at 08:29:33AM -0500, Yaakov (Cygwin/X) wrote:
>>>Corinna Vinschen wrote:
>>>>Isn't it a good time to split the release and release-2 areas into two
>>>>separate directories?  Now that many maintainers create separate
>>>>packages for 1.5 and 1.7, updating 1.5 and 1.7 directories and always
>>>>having to delete the 1.5 stuff from the 1.7 release-2 directory
>>>>afterwards is troublesome and bound to produce mistakes.  And the union
>>>>of the two directories makes increasingly less sense.
>>>
>>>+1
>>
>>For my packages, I still want the convenience of the unionfs.  [...]
>>IMO, we should be moving towards stabilization of the 1.5 package base
>>and the 1.7 Cygwin DLL code base.  I don't see either of those
>>happening right now, though.
>
>I don't understand what you mean.  What package stabilization has to be
>done in 1.5?

Stop obsoleting packages and adding major new package releases.  In
fact, I'd say stop anything but serious bugfixes or trivial upgrades to
1.5 packages at this point.

>And the Cygwin DLL is quite stable now.  I'm only fixing bugs and
>changing the wide char/multibyte

"wide char/multibyte" - that's a huge change to be adding to a product
which is supposed to be in beta.  It's worth a separate beta all by
itself.

In fact, as I have previously stated, this extended 1.7 release should
have, IMO, been a series of 1.7.x releases where the base functionality
was laid in 1.7.0 and improvements showed up at a regular basis in 1.7.1,
1.7.2, etc.  That mirrors the way that Linux does things and it's one of
the reasons why Cygwin adopted a similar versioning strategy.

cgf

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Corinna Vinschen]

On Mar 23 10:18, Christopher Faylor wrote:
> On Mon, Mar 23, 2009 at 08:29:33AM -0500, Yaakov (Cygwin/X) wrote:
> >Corinna Vinschen wrote:
> >>Isn't it a good time to split the release and release-2 areas into two
> >>separate directories?  Now that many maintainers create separate
> >>packages for 1.5 and 1.7, updating 1.5 and 1.7 directories and always
> >>having to delete the 1.5 stuff from the 1.7 release-2 directory
> >>afterwards is troublesome and bound to produce mistakes.  And the union
> >>of the two directories makes increasingly less sense.
> >
> >+1
> 
> For my packages, I still want the convenience of the unionfs.
> [...]
> IMO, we should be moving towards stabilization of the 1.5 package base
> and the 1.7 Cygwin DLL code base.  I don't see either of those happening
> right now, though.

I don't understand what you mean.  What package stabilization has to
be done in 1.5?

And the Cygwin DLL is quite stable now.  I'm only fixing bugs and
changing the wide char/multibyte stuff now, plus sweating over
documentation.  Except maybe for the libc/libm problem I don't see any
big block for a release in May or June.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Christopher Faylor]

On Mon, Mar 23, 2009 at 08:29:33AM -0500, Yaakov (Cygwin/X) wrote:
>Corinna Vinschen wrote:
>>Isn't it a good time to split the release and release-2 areas into two
>>separate directories?  Now that many maintainers create separate
>>packages for 1.5 and 1.7, updating 1.5 and 1.7 directories and always
>>having to delete the 1.5 stuff from the 1.7 release-2 directory
>>afterwards is troublesome and bound to produce mistakes.  And the union
>>of the two directories makes increasingly less sense.
>
>+1

For my packages, I still want the convenience of the unionfs.

I would think that, as 1.7 approaches, there would be fewer packages
released for 1.5 so this wouldn't be a real problem.  However, if a
maintainer doesn't want their directory to mirror the 1.5 release they
can just remove the directory in release-2 and recreate it with whatever
they want in it.  That will stop the mirroring.

I've just done this for lzma to prove that it works and it does.

But then, I don't think it makes a lot of sense to be doing much package
reorganization in 1.5 since it is approaching the end of its life.  When
we shuffle things around in the 1.5 release we increase the possibility
of annoying support problems when 1.7 is released.

IMO, we should be moving towards stabilization of the 1.5 package base
and the 1.7 Cygwin DLL code base.  I don't see either of those happening
right now, though.

cgf

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Yaakov (Cygwin/X)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Corinna Vinschen wrote:
> Isn't it a good time to split the release and release-2 areas into
> two separate directories?  Now that many maintainers create separate
> packages for 1.5 and 1.7, updating 1.5 and 1.7 directories and always
> having to delete the 1.5 stuff from the 1.7 release-2 directory
> afterwards is troublesome and bound to produce mistakes.  And the
> union of the two directories makes increasingly less sense.

+1


Yaakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEAREIAAYFAknHjr0ACgkQpiWmPGlmQSMpjACgj9vhRgw3/8/N1mryA3iBnfGi
3MQAn0XTi31BkjWBpAsqLfAdcuMWgCr/
=RY6g
-----END PGP SIGNATURE-----

Re: Bug in upset? [Was: Re: R: Problem [1.7]: link /bin/lzma -> xz]

[Corinna Vinschen]

[Redirected to cygwin-apps]

On Mar 22 17:39, Charles Wilson wrote:
> Christopher Faylor wrote:
> > Since release-2 isn't supposed to have obsolete stuff in it can't we
> > just remove this directory entirely?
> 
> No.
> 
> How do you propose to accomodate people -- esp. testers who have
> accepted Corinna's plea to "please begin testing cygwin-1.7" -- that
> have an existing "lzma" package into their cygwin-1.7 installation? 
> They need to (automatically?) get that existing one removed and the new
> version from xz installed in its place (ignoring the fact that I screwed
> up the setup.hints -- TWICE -- this time).
> 
> Is the policy about obsolete files so hard-and-fast that we want to
> require current 1.7 users to manually perform these actions, "uninstall
> lzma and then install xz..."?

Isn't it a good time to split the release and release-2 areas into
two separate directories?  Now that many maintainers create separate
packages for 1.5 and 1.7, updating 1.5 and 1.7 directories and always
having to delete the 1.5 stuff from the 1.7 release-2 directory
afterwards is troublesome and bound to produce mistakes.  And the
union of the two directories makes increasingly less sense.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat