Mailing List Archive: Perl: porters

FEATURE FREEZE for 5.12 on SATURDAY && The plan from there

Index | Next | Previous | View Threaded

jesse at fsck

Nov 20, 2009, 7:47 AM

Post #1 of 9 (428 views)
Permalink

FEATURE FREEZE for 5.12 on SATURDAY && The plan from there

As I mentioned last month, it's time to start the process of getting the
next stable series released. To that end, new features will not be
accepted after ** November 21, 2009 at 11:59PM Pacific/Midway (GMT+11) **

Don't worry if you don't have a commit bit: If your feature-change patch
hits p5p before then but isn't committed, it will still be considered
for 5.12.

Feature freeze is not code freeze. In particular, the following sorts of
changes are still 100% fair game:

* Tests to demonstrate new bugs
* Tests to better cover existing bugs or features
* Fixes for security issues
* Fixes for things that make perl crash
* Fixes for things that make perl give the wrong answer
* Fixes for things that make perl not build
* Fixes for pathological performance problems
* Fixes for regressions since 5.10
* ..and bug fixes of just about any kind

Until blead reopens after 5.12 is shipped, the following sorts of changes
_require_ an RT ticket and my signoff before they're committed to blead:

* CPAN Module updates or changes of any kind
* Binary-incompatible changes
* Performance improvements
* New language features
* New deprecations
* New core APIs
* New perl ports
* ...anything else you can't justify as a "bugfix"

There are already a couple of functionality changes (particularly in
regard to the unicode regex matching brouhaha and tooolchain updates)
that I've given provisional blessing to.

After this, changes more serious than a typo fix really ought to have RT
tickets associated with them, but I'd rather see a bugfix with a good
commit message than not see a fix.

As we have rather a lot of outstanding tickets and have had a period of
fairly rapid development, I am _not_ yet pegging a release date for
5.12.0RC1. There would be something rather poetic about hitting Christmas
Eve, but that will require a rather superhuman effort on the part of many,
many people.

Right now, there are 1581 open tickets in perl5's RT queue. Many of
these are "can't fix", "not a bug" or "oh, we fixed that ages ago", but
there are certainly a bunch of bugs in there that need to be fixed
before we can ship 5.12.

One of the things I need to see before 5.12.0 can go out the door is
for the first two lists below to be fully cleared. If you can help make
that happen, you will win my undying gratitude.

PJF and his accomplices at Sydney.pm cut through about 5% of our entire
outstanding database in an evening. If your local .pm group or other
conglomeration of perl hackers is up for a bug triaging session, I'd be
happy to help you make that happen.

Known blocking bugs:

http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!="rejected"+AND+Status!="resolved"+AND+DependedOnBy=69710

Bugs that might be blocking and need further triage:

http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!="rejected"+AND+Status!="resolved"+AND+DependedOnBy=70369

Non-critical bugs which should probably be fixed anyway:

http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!="rejected"+AND+Status!="resolved"+AND+DependedOnBy=70421

We also really need to walk through the rest of the open Perl 5 tickets looking for things we need to address before 5.12.0:

http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Queue='perl5'+AND+(Status='open'+OR+Status='new'+OR+Status='stalled')

--

heiko at hexco

Nov 20, 2009, 1:41 PM

Post #2 of 9 (394 views)
Permalink

Re: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

jesse wrote:

> Known blocking bugs:
>
> http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!="reject
> ed"+AND+Status!="resolved"+AND+DependedOnBy=69710
>
>
> Bugs that might be blocking and need further triage:
>
> http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!="reject
> ed"+AND+Status!="resolved"+AND+DependedOnBy=70369
>
>
> Non-critical bugs which should probably be fixed anyway:
>
> http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!="reject
> ed"+AND+Status!="resolved"+AND+DependedOnBy=70421
>
>
> We also really need to walk through the rest of the open Perl 5
> tickets looking for things we need to address before 5.12.0:
>
> http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Queue='perl5'+A
> ND+(Status='open'+OR+Status='new'+OR+Status='stalled')

The first three links yield no results. If you however replace " with '
in these URLs they will.

http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!='rejected
'+AND+Status!='resolved'+AND+DependedOnBy=69710

http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!='rejected
'+AND+Status!='resolved'+AND+DependedOnBy=70369

http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!='rejected
'+AND+Status!='resolved'+AND+DependedOnBy=70421

greetings hexcoder
--

john.peacock at havurah-software

Nov 20, 2009, 8:41 PM

Post #3 of 9 (408 views)
Permalink

Re: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

On 11/20/2009 10:47 AM, jesse wrote:
> As I mentioned last month, it's time to start the process of getting the
> next stable series released. To that end, new features will not be
> accepted after ** November 21, 2009 at 11:59PM Pacific/Midway (GMT+11) **

Was there some discussion of this freeze that I missed, or was it off
list? I have a pending version.pm release based on an issue that David
Golden opened:

https://rt.cpan.org/Ticket/Display.html?id=50841

I had to get over H1N1 first, so I haven't finished the feature, but I
have the pure Perl code working. This is a backwards compatible feature
that extends version objects to handle ActiveState's ill-considered munging.

I would also like to make one behavioral change: currently when you
explicitly numify() an alpha version object, it comes back with
underscore(s) intact. You'd have to eval() it to get what Perl would
normally do with bare numbers containing underscores. I'd like to
propose that numify() do that for you (i.e. strip the underscores). It
means that numify() will really return a number for all version objects.

I doubt that I'll be able to get this released by midnight tomorrow
night (or by my measure 5:59AM Sunday morning). I'm a little cranky
that this was sprung on us this morning (which I didn't even notice
because $WORK is crazy). I appreciate the hard work that you are doing
to get our collective butts in gear, but some more warning would have
been better...

John

craig.a.berry at gmail

Nov 21, 2009, 6:02 AM

Post #4 of 9 (402 views)
Permalink

Re: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

On Fri, Nov 20, 2009 at 10:41 PM, John Peacock
<john.peacock [at] havurah-software> wrote:
> On 11/20/2009 10:47 AM, jesse wrote:
>>
>> As I mentioned last month, it's time to start the process of getting the
>> next stable series released. � To that end, new features will not be
>> accepted after ** November 21, 2009 at 11:59PM Pacific/Midway (GMT+11) **
>
> Was there some discussion of this freeze that I missed, or was it off list?

It was in Jesse's pumpking announcement on 31 October:

http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2009-10/msg01584.html

where he said:

"Blead will feature-freeze on November 21. (just after Yves ships 5.11.2)
After that point, no feature should be added or removed from blead.
After that point, dual-lifed modules should only be updated from CPAN
versions to fix major bugs or security issues."

So for version.pm, you do have an out in that you can describe what's
in your RT queue as a "major bug" :-).

Three weeks is a pretty short window, and I don't understand what the
plan is for CPAN modules that are already out of synch with blead and
don't get updated by the deadline. I have not seen any public
hounding of people who are out of synch, or even any public analysis
of what needs synching up, both of which typically take many
iterations to get things in reasonable shape. I don't think we ever
finished getting all the little changes from the ext/ and cpan/
migration propagated outward. I don't think MakeMaker is in synch.

Most of the smoke I'm seeing *came* from CPAN in the last week or two,
so a strict interpretation of the policy would mean I'm SOL unless I
get those things fixed today and get the upstream authors to cut new
releases today. Which isn't going to happen.

Of course reasons for delaying, even good ones, are among the reasons
that Perl releases have been less frequent than people think they
ought to be. On the other hand, there was also plenty of language in
the pumpking announcement of the its-not-ready-until-its-ready
variety, so these things can perhaps be reconciled; thankfully it's
not my job to reconcile them.

jesse at fsck

Nov 21, 2009, 8:33 AM

Post #5 of 9 (398 views)
Permalink

Re: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

> Three weeks is a pretty short window, and I don't understand what the
> plan is for CPAN modules that are already out of synch with blead and
> don't get updated by the deadline. I have not seen any public
> hounding of people who are out of synch, or even any public analysis
> of what needs synching up, both of which typically take many
> iterations to get things in reasonable shape.

Indeed it does. And most of the hounding I've had to do, I've done in
private. Thankfully, 5.10.1 wasn't so long ago and Dave did a fantastic
job setting the stage for getting things in sync.

> I don't think we ever
> finished getting all the little changes from the ext/ and cpan/
> migration propagated outward.

That analysis is part of the release process, which means we've done it
three times since October 1, improving the state of the world each time.
Very few dists have unpropagated changes to anything that gets
installed. (There _are_ lots of changes to test setup that haven't
propagated, but that isn't a release showstopper)

> I don't think MakeMaker is in synch.

MakeMaker is a very special case and I've been working with Schwern to
get that sorted out.

> Most of the smoke I'm seeing *came* from CPAN in the last week or two,
> so a strict interpretation of the policy would mean I'm SOL unless I
> get those things fixed today and get the upstream authors to cut new
> releases today. Which isn't going to happen.

Well, not exactly. "Makes blead fail tests" is one of the things that I
consider a reasonable reason for pulling in a new CPAN distribution
update. The important change (for me) is that updates of CPAN modules
become carefully considered choices rather than a default of "oh hey,
there's a new dev release on CPAN! Let's check it in!"

> Of course reasons for delaying, even good ones, are among the reasons
> that Perl releases have been less frequent than people think they
> ought to be. On the other hand, there was also plenty of language in
> the pumpking announcement of the its-not-ready-until-its-ready
> variety, so these things can perhaps be reconciled; thankfully it's
> not my job to reconcile them.

*laugh* I'm still getting my bearings about a reasonable level of facism
for this. I'll almost certainly make some huge missteps that will draw
ire and criticism. But hey, hopefully we'll get a solid 5.12 out of
this.

-Jesse

--

john.peacock at havurah-software

Nov 21, 2009, 4:19 PM

Post #6 of 9 (399 views)
Permalink

Re: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

On 11/21/2009 09:02 AM, Craig A. Berry wrote:
> It was in Jesse's pumpking announcement on 31 October:
>
> http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2009-10/msg01584.html
>
> where he said:
>
> "Blead will feature-freeze on November 21. (just after Yves ships 5.11.2)
> After that point, no feature should be added or removed from blead.
> After that point, dual-lifed modules should only be updated from CPAN
> versions to fix major bugs or security issues."

So he did. I was only just starting to recover from the flu, so I can
use that as an excuse, :(

> So for version.pm, you do have an out in that you can describe what's
> in your RT queue as a "major bug" :-).

I just set it to critical, since without the fix, it will be hard(er)
for those poor lost souls running Windows to use version.pm...

John

jand at activestate

Nov 22, 2009, 2:46 AM

Post #7 of 9 (397 views)
Permalink

RE: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

On Fri, 20 Nov 2009, John Peacock wrote:
>
> https://rt.cpan.org/Ticket/Display.html?id=50841
>
> I had to get over H1N1 first, so I haven't finished the feature, but I
> have the pure Perl code working. This is a backwards compatible feature
> that extends version objects to handle ActiveState's ill-considered munging.

If you are talking about the double underscore in lib/ExtUtils/Install.pm
in ActivePerl build 1006, then this is an artifact of an integration error
and not an intentional use. It will be gone again in the next ActivePerl
release.

Note that I have not seen any bug reports (against ActivePerl) about it
either; I only noticed it myself while updating Module::Build.

So I think adding support for multiple underscores in version objects
is not a blocker for Perl 5.12.

Cheers,
-Jan

jesse at fsck

Nov 23, 2009, 6:12 AM

Post #8 of 9 (395 views)
Permalink

Re: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

On Fri, Nov 20, 2009 at 09:41:50PM -0000, Heiko Ei?feldt wrote:
> jesse wrote:
>
> > Known blocking bugs:
> >
> > http://rt.perl.org/rt3/Public/Search/Simple.html?Query=Status!="reject
> > ed"+AND+Status!="resolved"+AND+DependedOnBy=69710
> >
> >
> The first three links yield no results. If you however replace " with '
> in these URLs they will.
>

Thanks. Though, hmm, they seem to work fine for me. (Admittedly, it's
chromium, so god knows what sort of special things are going on)

john.peacock at havurah-software

Nov 23, 2009, 6:22 AM

Post #9 of 9 (393 views)
Permalink

Re: FEATURE FREEZE for 5.12 on SATURDAY && The plan from there [In reply to]

Jan Dubois wrote:
> If you are talking about the double underscore in lib/ExtUtils/Install.pm
> in ActivePerl build 1006, then this is an artifact of an integration error
> and not an intentional use. It will be gone again in the next ActivePerl
> release.

That is excellent news, as long as the next ActivePerl release is sooner
rather than later. Module::Build has aleady be fixed, so there is no
rush to get this in. I will hopefully put out an alpha this week
including the new feature, but the only integration of version.pm to
5.10.2 that needs to happen is just a version bump (so that 5.10.2 ships
with 0.78, even though no XS/C code changes were included after 0.77).

John

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads