[gentoo-user] Any real need to switch python targets back and forth every month?

[gentoo-user] Any real need to switch python targets back and forth every month?

[Grand Duet]

After today's emerge-webrsync, I have found that python_targets
and python_single_targets use flags have been changed again
from python3_5 to python3_6, which leads to a lot of recompilation.

It already happened last month and a week later the both use flags
was changed back, again with a lot of recompilations.

And it was not the first time!

Is any real need to switch python_targets and python_single_target
back and forth every month?

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Mike Gilbert]

On Wed, Jul 25, 2018 at 7:58 AM Grand Duet <grand.duet@gmail.com> wrote:
>
> After today's emerge-webrsync, I have found that python_targets
> and python_single_targets use flags have been changed again
> from python3_5 to python3_6, which leads to a lot of recompilation.
>
> It already happened last month and a week later the both use flags
> was changed back, again with a lot of recompilations.
>
> And it was not the first time!
>
> Is any real need to switch python_targets and python_single_target
> back and forth every month?
>

It's very simple:

1. The default was switched to python3.6. This was probably a bit premature.
2. People complained that this broke their systems, so we reverted the
default back to python3.5.
3. The problems were fixed, so we switched the default back to python3.6.

Sorry to have wasted your precious computing time, but hey, people
make mistakes sometimes.

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Peter Humphrey]

On Wednesday, 25 July 2018 16:20:59 BST Mike Gilbert wrote:

> Sorry to have wasted your precious computing time, but hey, people
> make mistakes sometimes.

I'm sure we all know exactly what you mean, Mike - from experience.

-- 
Regards,
Peter.

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Bill Kenworthy]

On 25/07/18 23:20, Mike Gilbert wrote:
> On Wed, Jul 25, 2018 at 7:58 AM Grand Duet <grand.duet@gmail.com> wrote:
>> After today's emerge-webrsync, I have found that python_targets
>> and python_single_targets use flags have been changed again
>> from python3_5 to python3_6, which leads to a lot of recompilation.
>>
>> It already happened last month and a week later the both use flags
>> was changed back, again with a lot of recompilations.
>>
>> And it was not the first time!
>>
>> Is any real need to switch python_targets and python_single_target
>> back and forth every month?
>>
> It's very simple:
>
> 1. The default was switched to python3.6. This was probably a bit premature.
> 2. People complained that this broke their systems, so we reverted the
> default back to python3.5.
> 3. The problems were fixed, so we switched the default back to python3.6.
>
> Sorry to have wasted your precious computing time, but hey, people
> make mistakes sometimes.
>
Major impact on users as well as creating confusion, talk on how a lot
of packages fail on 3.6 whats to say this isn't another mistake - isn't
this what news articles are for?


BillK

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Grand Duet]

2018-07-25 18:20 GMT+03:00 Mike Gilbert <floppym@gentoo.org>:
> On Wed, Jul 25, 2018 at 7:58 AM Grand Duet <grand.duet@gmail.com> wrote:
>>
>> After today's emerge-webrsync, I have found that python_targets
>> and python_single_targets use flags have been changed again
>> from python3_5 to python3_6, which leads to a lot of recompilation.
>>
>> It already happened last month and a week later the both use flags
>> was changed back, again with a lot of recompilations.
>>
>> And it was not the first time!
>>
>> Is any real need to switch python_targets and python_single_target
>> back and forth every month?
>>
>
> It's very simple:
>
> 1. The default was switched to python3.6. This was probably a bit premature.
> 2. People complained that this broke their systems, so we reverted the
> default back to python3.5.
> 3. The problems were fixed, so we switched the default back to python3.6.
>
> Sorry to have wasted your precious computing time, but hey, people
> make mistakes sometimes.

Before switching python_targets for the first time, you could use your
news system to inform Gentoo users that
1) you are switching python_targets
2) it may be "a bit premature",
and so, those who really want to have a stable Gentoo system should
1) do such and such changes to their config files and
2) wait about 3 months untill all the dust will settle.

By the way, last time I have tried to set
PYTHON_SINGLE_TARGET="python3_5"
PYTHON_TARGETS="python2_7 python3_5"
in my /etc/portage/make.conf
and I got a vary strange errors from emerge afterwards:
every time the
# emerge --update --deep --with-bdeps=y --newuse --backtrack=120 --ask world
command was run, a *new* package complained about something
and system update aborted.

I guess from that that the syntax of setting python_targets has been
changed lately (may be, it should now be set only in package.use)
but nobody informed Gentoo users about it.

P.S. The said above together with the recent Gentoo signing key issue
and sometimes "corrupted" daily portage snapshots that I download by
emerge-webrsync make me think that even so called "stable" Gentoo
lacks quality control and cannot be considered as stable and suted for
those that really need a stable system.

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Rich Freeman]

On Thu, Jul 26, 2018 at 6:44 AM Grand Duet <grand.duet@gmail.com> wrote:
>
> Before switching python_targets for the first time, you could use your
> news system to inform Gentoo users that
> 1) you are switching python_targets
> 2) it may be "a bit premature",

If anybody thought that it was a bit premature it wouldn't have been
done at all.

> and so, those who really want to have a stable Gentoo system should
> 1) do such and such changes to their config files and
> 2) wait about 3 months untill all the dust will settle.

Did this even impact the stable branch?  I thought this was only
deployed to ~arch.  If you want stable Gentoo, you probably shouldn't
have edited make.conf to set your keywords to ~arch.

> P.S. The said above together with the recent Gentoo signing key issue
> and sometimes "corrupted" daily portage snapshots that I download by
> emerge-webrsync make me think that even so called "stable" Gentoo
> lacks quality control and cannot be considered as stable and suted for
> those that really need a stable system.

Well, most people stick with RHEL/CentOS or Debian Stable for that
kind of experience, where changes only happen on releases with
significant QA, and backporting of fixes.  That simply isn't the kind
of experience Gentoo aims to deliver.

-- 
Rich

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Grand Duet]

2018-07-26 15:01 GMT+03:00 Rich Freeman <rich0@gentoo.org>:
> On Thu, Jul 26, 2018 at 6:44 AM Grand Duet <grand.duet@gmail.com> wrote:
>>
>> Before switching python_targets for the first time, you could use your
>> news system to inform Gentoo users that
>> 1) you are switching python_targets
>> 2) it may be "a bit premature",
>
> If anybody thought that it was a bit premature it wouldn't have been
> done at all.

It is exactly what is bad: nobody expects any troubles where
they definitely should be!

Anyway, it is a good idea to write a news on changing python_targets
and major version of gcc, even if no developper expects any troubles,
because we already know their predictive (dis)ability. :)

>> and so, those who really want to have a stable Gentoo system should
>> 1) do such and such changes to their config files and
>> 2) wait about 3 months untill all the dust will settle.
>
> Did this even impact the stable branch?

Yes.

> If you want stable Gentoo, you probably shouldn't have edited make.conf
> to set your keywords to ~arch.

I did not. Moreover, currently I have no single unstable package
installed on my system.

>> P.S. The said above together with the recent Gentoo signing key issue
>> and sometimes "corrupted" daily portage snapshots that I download by
>> emerge-webrsync make me think that even so called "stable" Gentoo
>> lacks quality control and cannot be considered as stable and suted for
>> those that really need a stable system.
>
> Well, most people stick with RHEL/CentOS or Debian Stable for that
> kind of experience, where changes only happen on releases with
> significant QA, and backporting of fixes.  That simply isn't the kind
> of experience Gentoo aims to deliver.

It is sad to hear. Actually, I don't want to change the Linux distro.
A bit more "due diligence" from Gentoo devs will be enough. :)

May be, adding some additional "almost stable" level between
"stable" and "unstable" one to make "stable" stable indeed?

In any case, if I will leave Gentoo, it will be in favour of some BSD,
not a Linux disro.

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Rich Freeman]

On Thu, Jul 26, 2018 at 8:45 AM Grand Duet <grand.duet@gmail.com> wrote:
>
> Anyway, it is a good idea to write a news on changing python_targets
> and major version of gcc, even if no developper expects any troubles,
> because we already know their predictive (dis)ability. :)
>

You won't get any argument from me there - more news is usually
better.  If nothing else people want to know they don't have to do
something when they see 200 packages being upgraded.

> > Did this even impact the stable branch?
>
> Yes.

Hmm, I suspect I didn't sync before it was reverted.  Either that or I
noticed the noise on the lists and waited a day.

This is one issue with our news - it isn't really realtime.  If we
want people to hold the presses and re-sync they don't get that notice
until they've already re-synced.

> May be, adding some additional "almost stable" level between
> "stable" and "unstable" one to make "stable" stable indeed?

If anything it seems like the proposal to drop stable comes along
every few years.  I don't see anybody being eager to add another level
of QA.  A big practical issue would be that unless people are actually
using the two lower levels significantly then nothing is actually
getting checked before going to stable.

There is really no reason you couldn't have a release-based Gentoo
derivative.  Everything is in git.  All "somebody" needs to do is
start a repo with a release-driven workflow that treats Gentoo as the
upstream master branch, targeting changes for release branches and
then doing release candidates and QA/etc.  Then those release-based
users would sync from there instead of the upstream Gentoo repo.
Ideally somebody would bundle it with a reference binary repo that
people could optionally sync from to speed installs for packages where
they're not changing USE flags.

The problem is that this all takes quite a bit of work, and I'm
skeptical that it would ever happen.  However, for larger Gentoo
deployments in production environments I suspect most are doing things
more-or-less in this fashion, but just with the packages they care
about.  If somebody has 100 production servers running Gentoo I doubt
they are set to just sync from us.  Rather they would set up their own
mirror and carefully test portage snapshots before they go rolling
them out.

-- 
Rich

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Grand Duet]

2018-07-26 16:04 GMT+03:00 Rich Freeman <rich0@gentoo.org>:
> On Thu, Jul 26, 2018 at 8:45 AM Grand Duet <grand.duet@gmail.com> wrote:
>> > Did this even impact the stable branch?
>>
>> Yes.
>
> Hmm, I suspect I didn't sync before it was reverted.  Either that or I
> noticed the noise on the lists and waited a day.
>
> This is one issue with our news - it isn't really realtime.  If we
> want people to hold the presses and re-sync they don't get that notice
> until they've already re-synced.
>
>> May be, adding some additional "almost stable" level between
>> "stable" and "unstable" one to make "stable" stable indeed?
>
> If anything it seems like the proposal to drop stable comes along
> every few years.  I don't see anybody being eager to add another level
> of QA.  A big practical issue would be that unless people are actually
> using the two lower levels significantly then nothing is actually
> getting checked before going to stable.
>
> There is really no reason you couldn't have a release-based Gentoo
> derivative.  Everything is in git.  All "somebody" needs to do is
> start a repo with a release-driven workflow that treats Gentoo as the
> upstream master branch, targeting changes for release branches and
> then doing release candidates and QA/etc.  Then those release-based
> users would sync from there instead of the upstream Gentoo repo.
> Ideally somebody would bundle it with a reference binary repo that
> people could optionally sync from to speed installs for packages where
> they're not changing USE flags.
>
> The problem is that this all takes quite a bit of work, and I'm
> skeptical that it would ever happen.  However, for larger Gentoo
> deployments in production environments I suspect most are doing things
> more-or-less in this fashion, but just with the packages they care
> about.  If somebody has 100 production servers running Gentoo I doubt
> they are set to just sync from us.  Rather they would set up their own
> mirror and carefully test portage snapshots before they go rolling
> them out.

Ok. Thank you for your reply.

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 449 bytes --]

On Thu, 26 Jul 2018 09:04:13 -0400, Rich Freeman wrote:

> This is one issue with our news - it isn't really realtime.  If we
> want people to hold the presses and re-sync they don't get that notice
> until they've already re-synced.

We could do with an option to sync only the news rather than the whole
tree, then we can read new news items before syncing fully.


-- 
Neil Bothwick

"Apple I" (c) Copyright 1767, Sir Isaac Newton.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 785 bytes --]

On Thu, 26 Jul 2018 13:44:28 +0300, Grand Duet wrote:

> Before switching python_targets for the first time, you could use your
> news system to inform Gentoo users that
> 1) you are switching python_targets

Like the one on May 22nd titled "Python 3.6 to become the default target"?

> 2) it may be "a bit premature",

That sort of information is only known with hindsight or a time machine.

> and so, those who really want to have a stable Gentoo system should
> 1) do such and such changes to their config files and

Like in the aforementioned news item?

FWIW I had no problem with the profile change and locked my system at 3.6
to save switching back when the profile did so.


-- 
Neil Bothwick

Anything worth fighting for is worth fighting dirty for.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Grand Duet]

2018-07-26 16:01 GMT+03:00 Neil Bothwick <neil@digimed.co.uk>:
> On Thu, 26 Jul 2018 13:44:28 +0300, Grand Duet wrote:
>
>> Before switching python_targets for the first time, you could use your
>> news system to inform Gentoo users that
>> 1) you are switching python_targets
>
> Like the one on May 22nd titled "Python 3.6 to become the default target"?
>
>> and so, those who really want to have a stable Gentoo system should
>> 1) do such and such changes to their config files and
>
> Like in the aforementioned news item?
>
> FWIW I had no problem with the profile change and locked my system at 3.6
> to save switching back when the profile did so.

Ok. There was such a news. However, I have actually followed the following
recommendation in it, before the last python_target switch from 3.5 to 3.6:

   If you would like to postpone the switch to Python 3.6, you can copy
   the current value of PYTHON_TARGETS and/or PYTHON_SINGLE_TARGET
   to /etc/portage/make.conf or its equivalent:

      PYTHON_TARGETS="python2_7 python3_5"
      PYTHON_SINGLE_TARGET="python3_5"

but the emerge gave strange errors mentioned above and aborted.

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1116 bytes --]

On Thu, 26 Jul 2018 17:34:25 +0300, Grand Duet wrote:

> > FWIW I had no problem with the profile change and locked my system at
> > 3.6 to save switching back when the profile did so.  
> 
> Ok. There was such a news. However, I have actually followed the
> following recommendation in it, before the last python_target switch
> from 3.5 to 3.6:
> 
>    If you would like to postpone the switch to Python 3.6, you can copy
>    the current value of PYTHON_TARGETS and/or PYTHON_SINGLE_TARGET
>    to /etc/portage/make.conf or its equivalent:
> 
>       PYTHON_TARGETS="python2_7 python3_5"
>       PYTHON_SINGLE_TARGET="python3_5"
> 
> but the emerge gave strange errors mentioned above and aborted.

I got some message about USE flags when keeping it at 3.6 in the same
way. So I did it in what seemed like the more elegant way of overriding a
profile change, in /etc/portage/profile/make.defaults.

PYTHON_TARGETS="python2_7 python3_5 -python3_6"
PYTHON_SINGLE_TARGET="python3_5 -python3_6"


-- 
Neil Bothwick

Last words of a Windows user: = Where do I have to click now? - There?

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Bill Kenworthy]

On 26/07/18 21:01, Neil Bothwick wrote:
> On Thu, 26 Jul 2018 13:44:28 +0300, Grand Duet wrote:
>
>> Before switching python_targets for the first time, you could use your
>> news system to inform Gentoo users that
>> 1) you are switching python_targets
> Like the one on May 22nd titled "Python 3.6 to become the default target"?
>
>> 2) it may be "a bit premature",
> That sort of information is only known with hindsight or a time machine.
>
>> and so, those who really want to have a stable Gentoo system should
>> 1) do such and such changes to their config files and
> Like in the aforementioned news item?
>
> FWIW I had no problem with the profile change and locked my system at 3.6
> to save switching back when the profile did so.
>
>
Yes there was the original news item - but nothing about the problems
this caused, and the back and forwarding which is causing me grief.  It
is only switching targets, but when you have problems and decide to
remove 3.6 and recompile, then having to go through it all again
without  proper information it has impact.  A news item is supposed to
give information ... I am not criticising the changes itself (sh..
happens), but being left in the dark about it.

BillK

Re: [gentoo-user] Any real need to switch python targets back and forth every month?

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 920 bytes --]

On Fri, 27 Jul 2018 09:50:34 +0800, Bill Kenworthy wrote:

> Yes there was the original news item - but nothing about the problems
> this caused, and the back and forwarding which is causing me grief.  It
> is only switching targets, but when you have problems and decide to
> remove 3.6 and recompile, then having to go through it all again
> without  proper information it has impact.  A news item is supposed to
> give information ... I am not criticising the changes itself (sh..
> happens), but being left in the dark about it.

Agreed, the switch back occurred with no explanation, a news item should
have been written for that, and maybe a brief warning one when
reinstating the 3.6 setting.


-- 
Neil Bothwick

Give a man a fish and you have fed him for a day, but give him a case of
dynamite and soon the village will be showered with mud and seaweed and
unidentifiable chunks of fish.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]