2.2.0 release plan

2.2.0 release plan

[Andy Wingo]

Hi,

I think we are ready to go for 2.2.0.  I am going to make a provisional
2.2.0 tag today and update the version.  I will dist a tarball and
upload to somewhere provisional.

However!  Even though the code says 2.2.0 and the tag says 2.2.0, the
release is not until Thursday.  I will not upload the tag yet.

This will give us a chance to update the web site, prepare nice release
notes, make a binary build using the new "guix pack", and test
everything out.  It could be we find some showstopper like a libtool age
issue or something.  If that's the case, we can fix it and re-spin a
release.  Since I won't have uploaded the tag, all should be good.

Thoughts welcome :)

Andy

Re: 2.2.0 release plan

[Greg Troxel]

[-- Attachment #1: Type: text/plain, Size: 886 bytes --]


Andy Wingo <wingo@pobox.com> writes:

> However!  Even though the code says 2.2.0 and the tag says 2.2.0, the
> release is not until Thursday.  I will not upload the tag yet.

I have just barely caught up with 2.0.x in pkgsrc, and haven't gotten to
testing 2.1.x yet.  I wonder how many platforms 2.1.x has been tested
on, and really I mean "has it been tested on other than GNU/Linux".

So if you think it's ready, I would suggest preparing a 2.2.0rc1, or
simply a new 2.1.x with the notion that aside from version numbers it's
an RC, and call for all people engaged in packaging to try to package it
as 2.2, and to give them a few weeks.

The basic issue is that getting widespread testing relies on a bunch of
people for whom hacking on guile is one little thing among many, and not
their main Free Software focus.  This is not about guile; I see this in
a vast number of projects.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 162 bytes --]

Re: 2.2.0 release plan

[Andy Wingo]

On Tue 14 Mar 2017 16:25, Greg Troxel <gdt@lexort.com> writes:

> Andy Wingo <wingo@pobox.com> writes:
>
>> However!  Even though the code says 2.2.0 and the tag says 2.2.0, the
>> release is not until Thursday.  I will not upload the tag yet.
>
> I have just barely caught up with 2.0.x in pkgsrc, and haven't gotten to
> testing 2.1.x yet.  I wonder how many platforms 2.1.x has been tested
> on, and really I mean "has it been tested on other than GNU/Linux".

Things are in the early days, you are right.  GNU/Linux and macOS have
been tested.  Cygwin builds and works though it has some test errors; we
are working on it.  I don't know about mingw yet.  I think there might
be some errors on the other BSDs but I don't know yet.

> So if you think it's ready, I would suggest preparing a 2.2.0rc1, or
> simply a new 2.1.x with the notion that aside from version numbers it's
> an RC, and call for all people engaged in packaging to try to package it
> as 2.2, and to give them a few weeks.

I have done exactly that in the past weeks with 2.1.7 and 2.1.8 :) I
think a 2.2.1 will follow very soon (within a month or so from 2.2.0, I
would imagine), so any portability fixes will naturally be included
there.  That's about the same timeframe you are asking for, so all is
well in that regard.

Happy hacking,

Andy

Re: 2.2.0 release plan

[Matt Wette]

> On Mar 14, 2017, at 8:50 AM, Andy Wingo <wingo@pobox.com> wrote:
> 
> On Tue 14 Mar 2017 16:25, Greg Troxel <gdt@lexort.com> writes:
> 
>> Andy Wingo <wingo@pobox.com> writes:
>> 
>>> However!  Even though the code says 2.2.0 and the tag says 2.2.0, the
>>> release is not until Thursday.  I will not upload the tag yet.
>> 
>> I have just barely caught up with 2.0.x in pkgsrc, and haven't gotten to
>> testing 2.1.x yet.  I wonder how many platforms 2.1.x has been tested
>> on, and really I mean "has it been tested on other than GNU/Linux".
> 
> Things are in the early days, you are right.  GNU/Linux and macOS have
> been tested.  Cygwin builds and works though it has some test errors; we
> are working on it.  I don't know about mingw yet.  I think there might
> be some errors on the other BSDs but I don't know yet.

Just got my desktop to extra-boot FreeBSD.  I’m taking a shot at that target (amd64).

Re: 2.2.0 release plan

[Matt Wette]

[-- Attachment #1: Type: text/plain, Size: 1826 bytes --]


> On Mar 14, 2017, at 5:35 PM, Matt Wette <matt.wette@gmail.com> wrote:
> 
>> 
>> On Mar 14, 2017, at 8:50 AM, Andy Wingo <wingo@pobox.com> wrote:
>> 
>> On Tue 14 Mar 2017 16:25, Greg Troxel <gdt@lexort.com> writes:
>> 
>>> Andy Wingo <wingo@pobox.com> writes:
>>> 
>>>> However!  Even though the code says 2.2.0 and the tag says 2.2.0, the
>>>> release is not until Thursday.  I will not upload the tag yet.
>>> 
>>> I have just barely caught up with 2.0.x in pkgsrc, and haven't gotten to
>>> testing 2.1.x yet.  I wonder how many platforms 2.1.x has been tested
>>> on, and really I mean "has it been tested on other than GNU/Linux".
>> 
>> Things are in the early days, you are right.  GNU/Linux and macOS have
>> been tested.  Cygwin builds and works though it has some test errors; we
>> are working on it.  I don't know about mingw yet.  I think there might
>> be some errors on the other BSDs but I don't know yet.
> 
> Just got my desktop to extra-boot FreeBSD.  I’m taking a shot at that target (amd64).

Status: The (g)make is grinding away, now building ice-9/psyntax-pp.go, going slow ...

Notes:

To configure, besides the mentioned items (e.g., libffi), must "pkg install"
* pkgconf
* gmake
* boehm-gc-threaded

Must use “gmake”, “make” breaks on meta/Makefile

The first gmake produced "not found" link errors for
  GC_unregister_my_thread
  GC_get_suspend_signal
  GC_allow_register_threads
  GC_register_my_thread

So I reconfigured using
  BDW_GC_LIBS=-L/usr/local/lib -lgc-threaded ./configure --prefix=/usr/local


Now the HACK:
ran into issue w/ weak-set.c (?), so I
edited config.h and changed
  /* #undef HAVE_GC_MOVE_DISAPPEARING_LINK */
to
  #define HAVE_GC_MOVE_DISAPPEARING_LINK 1

I will report more later, as time permits.
Matt


[-- Attachment #2: Type: text/html, Size: 9822 bytes --]

Re: 2.2.0 release plan

[Matt Wette]

[-- Attachment #1: Type: text/plain, Size: 2427 bytes --]


> On Mar 14, 2017, at 7:59 PM, Matt Wette <matt.wette@gmail.com> wrote:
> 
> 
>> On Mar 14, 2017, at 5:35 PM, Matt Wette <matt.wette@gmail.com <mailto:matt.wette@gmail.com>> wrote:
>> 
>>> 
>>> On Mar 14, 2017, at 8:50 AM, Andy Wingo <wingo@pobox.com <mailto:wingo@pobox.com>> wrote:
>>> 
>>> On Tue 14 Mar 2017 16:25, Greg Troxel <gdt@lexort.com <mailto:gdt@lexort.com>> writes:
>>> 
>>>> Andy Wingo <wingo@pobox.com <mailto:wingo@pobox.com>> writes:
>>>> 
>>>>> However!  Even though the code says 2.2.0 and the tag says 2.2.0, the
>>>>> release is not until Thursday.  I will not upload the tag yet.
>>>> 
>>>> I have just barely caught up with 2.0.x in pkgsrc, and haven't gotten to
>>>> testing 2.1.x yet.  I wonder how many platforms 2.1.x has been tested
>>>> on, and really I mean "has it been tested on other than GNU/Linux".
>>> 
>>> Things are in the early days, you are right.  GNU/Linux and macOS have
>>> been tested.  Cygwin builds and works though it has some test errors; we
>>> are working on it.  I don't know about mingw yet.  I think there might
>>> be some errors on the other BSDs but I don't know yet.
>> 
>> Just got my desktop to extra-boot FreeBSD.  I’m taking a shot at that target (amd64).
> 
> Status: The (g)make is grinding away, now building ice-9/psyntax-pp.go, going slow ...
> 
> Notes:
> 
> To configure, besides the mentioned items (e.g., libffi), must "pkg install"
> * pkgconf
> * gmake
> * boehm-gc-threaded
> 
> Must use “gmake”, “make” breaks on meta/Makefile
> 
> The first gmake produced "not found" link errors for
>   GC_unregister_my_thread
>   GC_get_suspend_signal
>   GC_allow_register_threads
>   GC_register_my_thread
> 
> So I reconfigured using
>   BDW_GC_LIBS=-L/usr/local/lib -lgc-threaded ./configure --prefix=/usr/local
> 
> 
> Now the HACK:
> ran into issue w/ weak-set.c (?), so I
> edited config.h and changed
>   /* #undef HAVE_GC_MOVE_DISAPPEARING_LINK */
> to
>   #define HAVE_GC_MOVE_DISAPPEARING_LINK 1
> 
> I will report more later, as time permits.
> Matt
> 

“gmake” and “gmake check” completed on FreeBSD 11.0

There were some warnings in the make check regarding forking w/ multiple threads running.  I assume this is an OS issue.

The remaining task is to get configure to work correctly with the FreeBSD boehm-gc-threaded package.  Not sure how long that will take.



[-- Attachment #2: Type: text/html, Size: 10751 bytes --]

Re: 2.2.0 release plan

[Matt Wette]

[-- Attachment #1: Type: text/plain, Size: 3303 bytes --]


> On Mar 15, 2017, at 6:07 AM, Matt Wette <matt.wette@gmail.com> wrote:
> 
> 
>> On Mar 14, 2017, at 7:59 PM, Matt Wette <matt.wette@gmail.com <mailto:matt.wette@gmail.com>> wrote:
>> 
>> 
>>> On Mar 14, 2017, at 5:35 PM, Matt Wette <matt.wette@gmail.com <mailto:matt.wette@gmail.com>> wrote:
>>> 
>>>> 
>>>> On Mar 14, 2017, at 8:50 AM, Andy Wingo <wingo@pobox.com <mailto:wingo@pobox.com>> wrote:
>>>> 
>>>> On Tue 14 Mar 2017 16:25, Greg Troxel <gdt@lexort.com <mailto:gdt@lexort.com>> writes:
>>>> 
>>>>> Andy Wingo <wingo@pobox.com <mailto:wingo@pobox.com>> writes:
>>>>> 
>>>>>> However!  Even though the code says 2.2.0 and the tag says 2.2.0, the
>>>>>> release is not until Thursday.  I will not upload the tag yet.
>>>>> 
>>>>> I have just barely caught up with 2.0.x in pkgsrc, and haven't gotten to
>>>>> testing 2.1.x yet.  I wonder how many platforms 2.1.x has been tested
>>>>> on, and really I mean "has it been tested on other than GNU/Linux".
>>>> 
>>>> Things are in the early days, you are right.  GNU/Linux and macOS have
>>>> been tested.  Cygwin builds and works though it has some test errors; we
>>>> are working on it.  I don't know about mingw yet.  I think there might
>>>> be some errors on the other BSDs but I don't know yet.
>>> 
>>> Just got my desktop to extra-boot FreeBSD.  I’m taking a shot at that target (amd64).
>> 
>> Status: The (g)make is grinding away, now building ice-9/psyntax-pp.go, going slow ...
>> 
>> Notes:
>> 
>> To configure, besides the mentioned items (e.g., libffi), must "pkg install"
>> * pkgconf
>> * gmake
>> * boehm-gc-threaded
>> 
>> Must use “gmake”, “make” breaks on meta/Makefile
>> 
>> The first gmake produced "not found" link errors for
>>   GC_unregister_my_thread
>>   GC_get_suspend_signal
>>   GC_allow_register_threads
>>   GC_register_my_thread
>> 
>> So I reconfigured using
>>   BDW_GC_LIBS=-L/usr/local/lib -lgc-threaded ./configure --prefix=/usr/local
>> 
>> 
>> Now the HACK:
>> ran into issue w/ weak-set.c (?), so I
>> edited config.h and changed
>>   /* #undef HAVE_GC_MOVE_DISAPPEARING_LINK */
>> to
>>   #define HAVE_GC_MOVE_DISAPPEARING_LINK 1
>> 
>> I will report more later, as time permits.
>> Matt
>> 
> 
> “gmake” and “gmake check” completed on FreeBSD 11.0
> 
> There were some warnings in the make check regarding forking w/ multiple threads running.  I assume this is an OS issue.
> 
> The remaining task is to get configure to work correctly with the FreeBSD boehm-gc-threaded package.  Not sure how long that will take.

So I think this is a clue for FreeBSD config w/ boehm-gc-threaded, but final fix.  My total fix below does not work.  Maybe someone knows autoconf better.

Near line 1323 in configure.ac is the line
    PKG_CHECK_MODULES([BDW_GC], [bdw-gc >= 7.2])

I changed it on my FreeBSD system to
    PKG_CHECK_MODULES([BDW_GC], [bdw-gc-threaded >= 7.2])
which WORKS!

But if I try this general solution:
    PKG_CHECK_MODULES([BDW_GC], [bdw-gc-threaded >= 7.2],[],
      [PKG_CHECK_MODULES([BDW_GC],, [bdw-gc >= 7.2])])
NOT WORKING

Andy, can you get it working?  (If bdw-gc-threaded not available try bdw-gc.). I assume something like this should work for the other platforms also.

Matt


[-- Attachment #2: Type: text/html, Size: 14226 bytes --]