Choosing a moment for the Bazaar switchover.

Choosing a moment for the Bazaar switchover.

[Karl Fogel]

Thanks to Jason Earl and others, we've solved most of our Bazaar
conversion issues.  There are still a few things to clean up, but I
think we're at the stage where it would help to plan for a switchover
date.

When should we do this?  Some options:

   1) Now, i.e., as soon as the conversion is ready (in which case we'll
      start prepping knowing that it's for real this time).
   2) When we branch for release.
   3) Immediately after the release.

I advocate for (1), on the grounds that there is never a good time for
these things anyway, and the sooner they are done the better.  By going
through a release process with a new version control system, we know
we'll work out all the kinks, and by the end of it, everyone will be
comfortable with Bazaar.

(No doubt it will interfere with the release a bit -- but that would
happen whenever we first release with a new VC system.)

Also, (1) is the option that doesn't tie the switchover to things that
could take an unknown amount of time.  (2) or (3) create a dependency
that can delay the switchover indefinitely.

References: http://www.emacswiki.org/emacs-en/EmacsBzrSwitchover

Thoughts?

-Karl

Re: Choosing a moment for the Bazaar switchover.

[Stefan Monnier]

> Thanks to Jason Earl and others, we've solved most of our Bazaar
> conversion issues.

Can we have a quick description of the state of the conversion business?
IIUC we now have a conversion that preserves the history, branches, and
tags from the CVS repository, plus adds the merge info coming from the
Arch (via the git) repository (and half-manually stuffed into the git
repository, thanks to Andreas), plus adds the rename info
guessed/inferred by git itself.

And that conversion can be done "easily" but cannot be updated later on
to incorporate changes in the CVS repository, forcing us into a "flag
day" scenario.

Right?

Also, what's the status of the Savannah hosting?  Looking at
http://savannah.gnu.org/support/index.php?106612, it's not clear whether
"bzr+ssh" access is available, and whether emacs-commit is working.
Do we have a test branch setup there?  Anything else that might be
missing?
[ BTW ]

> When should we do this?  Some options:

>    1) Now, i.e., as soon as the conversion is ready (in which case we'll
>       start prepping knowing that it's for real this time).
>    2) When we branch for release.
>    3) Immediately after the release.

I think I'd prefer 2.  It should be a time where the release branch
(currently "trunk") doesn't see too much activity, which would help.


        Stefan

Re: Choosing a moment for the Bazaar switchover.

[Karl Fogel]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:
>> Thanks to Jason Earl and others, we've solved most of our Bazaar
>> conversion issues.
>
> Can we have a quick description of the state of the conversion business?
> IIUC we now have a conversion that preserves the history, branches, and
> tags from the CVS repository, plus adds the merge info coming from the
> Arch (via the git) repository (and half-manually stuffed into the git
> repository, thanks to Andreas), plus adds the rename info
> guessed/inferred by git itself.
>
> And that conversion can be done "easily" but cannot be updated later on
> to incorporate changes in the CVS repository, forcing us into a "flag
> day" scenario.
>
> Right?

Yes, "flag day" is what we're looking at.  

Your description of the conversion sounds right; Jason can respond
authoritatively.

> Also, what's the status of the Savannah hosting?  Looking at
> http://savannah.gnu.org/support/index.php?106612, it's not clear whether
> "bzr+ssh" access is available, and whether emacs-commit is working.
> Do we have a test branch setup there?  Anything else that might be
> missing?

All the todo items I know about are listed at
http://www.emacswiki.org/emacs-en/EmacsBzrSwitchover, and I'm happy to
push on them.  We don't have a test branch yet; making one would be part
of that.

(I'm trying to avoid a situation where Sylvain works really hard to get
everything set up, and then it's months before we actually switch, so it
bitrots and some of the work has to be redone.)

>> When should we do this?  Some options:
>
>>    1) Now, i.e., as soon as the conversion is ready (in which case we'll
>>       start prepping knowing that it's for real this time).
>>    2) When we branch for release.
>>    3) Immediately after the release.
>
> I think I'd prefer 2.  It should be a time where the release branch
> (currently "trunk") doesn't see too much activity, which would help.

Do we have a rough idea when (2) is?  (It's probably been talked about
here, but... there's a lot of mail, and there are few on-web clues about
the schedule.)

Thanks,
-Karl

Re: Choosing a moment for the Bazaar switchover.

[Stefan Monnier]

> All the todo items I know about are listed at
> http://www.emacswiki.org/emacs-en/EmacsBzrSwitchover, and I'm happy to
> push on them.  We don't have a test branch yet; making one would be
> part of that.

Then we should add "test branch setup", "test branch tested",
"emacs-commit working" to that list.

> (I'm trying to avoid a situation where Sylvain works really hard to get
> everything set up, and then it's months before we actually switch, so it
> bitrots and some of the work has to be redone.)

Yes, that's fine by me.

>>> When should we do this?  Some options:
>>> 1) Now, i.e., as soon as the conversion is ready (in which case we'll
>>> start prepping knowing that it's for real this time).
>>> 2) When we branch for release.
>>> 3) Immediately after the release.
>> 
>> I think I'd prefer 2.  It should be a time where the release branch
>> (currently "trunk") doesn't see too much activity, which would help.

> Do we have a rough idea when (2) is?  (It's probably been talked about
> here, but... there's a lot of mail, and there are few on-web clues about
> the schedule.)

AFAIK it hasn't been talked about, really.  My rough guess is "sometime
in 2009".  Also, probably on the later part of 2009.


        Stefan

release (was: Re: Choosing a moment for the Bazaar switchover.)

[Dan Nicolaescu]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

  > > Do we have a rough idea when (2) is?  (It's probably been talked about
  > > here, but... there's a lot of mail, and there are few on-web clues about
  > > the schedule.)
  > 
  > AFAIK it hasn't been talked about, really.  My rough guess is "sometime
  > in 2009".  Also, probably on the later part of 2009.

Is there a list of things that need to happen for the release?

We have a bug tracker now, can the bugs that need to be fixed for the
release be marked somehow?
Are all the must-fix bugs in the bug tracker?

IMHO, providing clear criteria would give people extra incentive to work
towards the release.

Re: release

[Chong Yidong]

Dan Nicolaescu <dann@ics.uci.edu> writes:

> Is there a list of things that need to happen for the release?
>
> We have a bug tracker now, can the bugs that need to be fixed for the
> release be marked somehow?  Are all the must-fix bugs in the bug
> tracker?

As we get closer to the release, we will use the severity tag in the bug
tracker to indicate release-blocking bugs.  I think we are not close
enough to the release for that, yet.  The main outstanding issues, off
the top of my head, are: (1) bugs in the new internationalization and
font-handling code.  Bug reports are still slowly trickling in,
indicating that the code has not converged sufficiently yet.  (2)
Several files in the Emacs and Emacs Lisp reference manual still need to
be checked (progress seems good, but this is a slow process.)

Re: release

[Dan Nicolaescu]

Chong Yidong <cyd@stupidchicken.com> writes:

  > Dan Nicolaescu <dann@ics.uci.edu> writes:
  > 
  > > Is there a list of things that need to happen for the release?
  > >
  > > We have a bug tracker now, can the bugs that need to be fixed for the
  > > release be marked somehow?  Are all the must-fix bugs in the bug
  > > tracker?
  > 
  > As we get closer to the release, we will use the severity tag in the bug
  > tracker to indicate release-blocking bugs.  I think we are not close
  > enough to the release for that, yet.

If there are bugs that are known to be release blocking, why not mark
them ASAP?  If  they prove later to not be so important, they can be
unmarked.

Again, knowing what needs to be done provides an extra incentive for
people to do it.

Too bad that the number of people that can fix the most critical of bugs
(unicode) is so tiny...

Re: release

[Stefan Monnier]

> We have a bug tracker now, can the bugs that need to be fixed for the
> release be marked somehow?

Supposedly there is a way to do such things (mark a bug as "blocking"
some release), tho I haven't figured it out yet.  I think using the
severity for it is OK, as well.

> If there are bugs that are known to be release blocking, why not mark
> them ASAP?

No reason at all.

> Are all the must-fix bugs in the bug tracker?

AFAIK, yes.


        Stefan

Re: Choosing a moment for the Bazaar switchover.

[David Reitter]

[-- Attachment #1: Type: text/plain, Size: 286 bytes --]

What format is the repository going to be in?  --1.9?
(I was just told that --1.9-rich-root is not recommended for  
production.)

Are the file IDs in the repository on Savannah going to be aligned  
with those currently used on notengoamigos.org?
(and if so, which of the archives?)



[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 2193 bytes --]

Re: Choosing a moment for the Bazaar switchover.

[Stefan Monnier]

> What format is the repository going to be in?  --1.9?

Why would it matter?  We can always easily change it later on.

> Are the file IDs in the repository on Savannah going to be aligned with
> those currently used on notengoamigos.org?

AFAIK, no.


        Stefan

Re: Choosing a moment for the Bazaar switchover.

[David Reitter]

[-- Attachment #1: Type: text/plain, Size: 476 bytes --]

On Mar 22, 2009, at 10:04 AM, Stefan Monnier wrote:

>> What format is the repository going to be in?  --1.9?
>
> Why would it matter?  We can always easily change it later on.

The reason why I asked is that I'm setting up a branch right now,  
possibly to be rebased later on.  And it seems that rebase is  
problematic (or unbearably slow) when repository formats differ.

But I would hope that it's not going to be 0.92 (emacs on  
notengoamigos is 0.92, emacs-ce is 1.9).

[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 2193 bytes --]

Re: Choosing a moment for the Bazaar switchover.

[Stefan Monnier]

> I have, however, created a new repository, containing all of the
> branches from CVS available at the usual location:

> http://bzr.notengoamigos.org/emacs-merges-ce.tar.[gz,lzma]

> I don't believe that this repository has rename info.

Is Ian's aware of the rename problem?


        Stefan

Re: Choosing a moment for the Bazaar switchover.

[Andreas Schwab]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> I have, however, created a new repository, containing all of the
>> branches from CVS available at the usual location:
>
>> http://bzr.notengoamigos.org/emacs-merges-ce.tar.[gz,lzma]
>
>> I don't believe that this repository has rename info.
>
> Is Ian's aware of the rename problem?

I'm not aware of a single changeset in the CVS repository that
represents a rename.  Whenever a file was moved there were separate
commits that remove and add the file.  For an extreme example, take a
look at
<http://repo.or.cz/w/emacs.git?a=commitdiff;h=9446f6ee8b2b0d758da9eb4b34627a33dacf8922;hp=d0b8e1ae34a5a0c6be87b291002a8a86db8be5c9>,
which is the summary of the move of all the texi files from man/ to
doc/.  This summary consists of a total of 428 commits, almost all of
them touching only a single file.  Moreover, each file was first removed
in one commit and readded only much later.  A proper move would probably
have produced less than 5 commits.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Choosing a moment for the Bazaar switchover.

[Stefan Monnier]

>> Is Ian's aware of the rename problem?
> I'm not aware of a single changeset in the CVS repository that
> represents a rename.  Whenever a file was moved there were separate
> commits that remove and add the file.  For an extreme example, take
> a look at <...> which is the summary of the move of all the texi files
> from man/ to doc/.

Hmm... indeed that was botched.


        Stefan

Re: Choosing a moment for the Bazaar switchover.

[Alan Mackenzie]

Hi, everybody!

On Fri, Mar 20, 2009 at 01:49:41PM -0400, Karl Fogel wrote:
> Thanks to Jason Earl and others, we've solved most of our Bazaar
> conversion issues.  There are still a few things to clean up, but I
> think we're at the stage where it would help to plan for a switchover
> date.

> When should we do this?  Some options:

>    1) Now, i.e., as soon as the conversion is ready (in which case we'll
>       start prepping knowing that it's for real this time).
>    2) When we branch for release.
>    3) Immediately after the release.

> I advocate for (1), on the grounds that there is never a good time for
> these things anyway, and the sooner they are done the better.  By going
> through a release process with a new version control system, we know
> we'll work out all the kinks, and by the end of it, everyone will be
> comfortable with Bazaar.

> (No doubt it will interfere with the release a bit -- but that would
> happen whenever we first release with a new VC system.)

I think there is doubt here.

> Also, (1) is the option that doesn't tie the switchover to things that
> could take an unknown amount of time.  (2) or (3) create a dependency
> that can delay the switchover indefinitely.

Yes.  You and the other heroes doing this switch are full of adrenalin
at the moment.  That's going to go sour if you an indefinite 6 - 12
months wait.

> References: http://www.emacswiki.org/emacs-en/EmacsBzrSwitchover

> Thoughts?

I say, convert as soon as possible.  Clearly Emacs 23 isn't anywhere
near ready for release yet, and as you say, there's no good time to
convert to a new VCS, just like there's no good time to have a baby.

If we're assuming that bzr will help us work more efficiently than we're
doing with CVS, then the sooner we make the change, the longer we'll be
working more efficiently.  :-)  My gut feeling is that the saving we'll
get from bzr will outweigh the effort of learning it well within the
time up till 23.1's release.

Let's do it now!

> -Karl

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: Choosing a moment for the Bazaar switchover.

[Karl Fogel]

Alan Mackenzie <acm@muc.de> writes:
> Yes.  You and the other heroes doing this switch are full of adrenalin
> at the moment.  That's going to go sour if you an indefinite 6 - 12
> months wait.

Oh, the heroes are Jason Earl, Andreas Schwab, and Sylvain Beucler.
(I'm enjoying doing this, but I'm also able to do it on work time, which
helps a lot of course.)

There is such a thing as momentum, yes...  My motivation won't sour, but
I have time to do this right now; can't speak for Jason, Andreas, and
Sylvain.

> I say, convert as soon as possible.  Clearly Emacs 23 isn't anywhere
> near ready for release yet, and as you say, there's no good time to
> convert to a new VCS, just like there's no good time to have a baby.
>
> If we're assuming that bzr will help us work more efficiently than we're
> doing with CVS, then the sooner we make the change, the longer we'll be
> working more efficiently.  :-)  My gut feeling is that the saving we'll
> get from bzr will outweigh the effort of learning it well within the
> time up till 23.1's release.
>
> Let's do it now!

As you know, I favor this, for all the above reasons.  But Stefan and
Chong have to be on board with it.  Stefan has indicated that he prefers
to wait until we branch for release.  I'm waiting to see if they're
swayed by your excellent argument above :-).

All other questions in this thread, like what format will we use, etc,
I'm going to defer until we are "all systems go".  If I try to answer
the question now, then a) some of the answers will be obsolete by the
time we switch, and b) I will have forgotten a lot of them anyway.

Re: Choosing a moment for the Bazaar switchover.

[Karl Fogel]

Jason Earl <jearl@notengoamigos.org> writes:
> That being the case, I don't think that it would be wise to switch to
> bzr right now.  For one thing a minor bug in bzr's fast-import plugin
> has made it impossible for me to generate new emacs-merges repositories
> from Andreas' git repository in anything but the bleeding edge
> brisbane-core formats.  I tried poking at this bug a bit myself, but I
> quickly found myself out of my depth.  Ian Clatworthy is working on it
> now.  Here's the bug report if you are interested.
>
> https://bugs.launchpad.net/bzr-fastimport/+bug/345602

*nod*  Thank you; yes, that's very helpful to know.

> The other advantage of waiting is that bzr really is still in a state of
> considerable flux.  I have really come to love bzr, but it is hard to
> ignore the fact that the bzr devs make considerable changes on a monthly
> basis.  I think that it might well be in Emacs' best interests to at
> least wait until the new repository format lands.

I'd love to know what causes your impression of "a state of flux".  I
have some ideas, but don't want to taint your impressions by stating
them yet -- would rather hear from you.

NOTE: I think the Bazaar developers would like to know, too, so I've set
Reply-to on this mail to go to the Bazaar list.  The thread would get
off-topic for emacs-devel at this point anyway -- anyone who does a
follow-up-to-all, please take emacs-devel@ off the recipient list.

-Karl