Debian Bug report logs - #779340
unblock (pre-approval): marco/1.8.2+dfsg1-5

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Mike Gabriel <sunweaver@debian.org>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: unblock (pre-approval): marco/1.8.2+dfsg1-5

Date: Fri, 27 Feb 2015 13:05:17 +0100

[Message part 1 (text/plain, inline)]

Package: release.debian.org
Severity: normal
User: release.debian.org@packages.debian.org
Usertags: unblock

Please consider unblocking planned upload of package marco.

Due to a conceptual change in the Gsettings backend subscription handling
in glib 2.43, the MATE upstream team stumbled over several code flaws in
the Gsettings implemenation of MATE. The Debian MATE Packaging Team,
together with the Ubuntu MATE team, would love to see these code flaws
fixed in Debian jessie, although Debian jessie ships an older version of
glib and is not directly affected by these flaws at the moment.

In MATE, several components erroneously read from Gsettings before they
are actually connected to their signals. With the change [1] in glib,
this leads to situations, where changes of certain desktop settings do
not get propagated to the running application (Gsettings notifications
fail). The user needs to logout and login to see changes on desktop
components take effect.

These problems

  o will hit Debian, once glib 2.43 gets uploaded
  o hit derivatives of Debian that ship glib 2.43, currently
    Ubuntu MATE 15.04

Indepent from those code flaws being invisible in Debian jessie or not,
the actual code passages in affected MATE components are technically
wrong and should be prophylactically fixed in Debian. This will also
greatly ease the team cooperation with Ubuntu MATE (from where we receive
a great deal of input and impact).

+  [ Martin Wimpress ]
+  * debian/patches:
+    + Add 0006_glib_connect_first.patch. Connect to settings
+      first, then read from them. Fixes severe breakage with glib >=
+      2.43.1. (Closes: #778694). (LP: #1426327).
+

-> In Marco (MATE Window Manager) the applied patch makes sure that
setting changes in the "Settings -> Preferences -> Windows" dialog box
immediately reach the running instance of Marco, even with glib >= 2.43
installed.

light+love,
Mike


[1] https://git.gnome.org/browse/glib/commit/?id=8ff5668a458344da22d30491e3ce726d861b3619


unblock marco/1.8.2+dfsg1-5

-- System Information:
Debian Release: 8.0
  APT prefers stable
  APT policy: (990, 'stable'), (500, 'testing-updates'), (500, 'testing-proposed-updates'), (500, 'testing')
Architecture: amd64 (x86_64)
Foreign Architectures: i386

Kernel: Linux 3.16.0-4-amd64 (SMP w/4 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: sysvinit (via /sbin/init)

[marco_1.8.2+dfsg1-4_1.8.2+dfsg1-5.debdiff (text/x-diff, attachment)]

Message #10 received at 779340@bugs.debian.org (full text, mbox, reply):

From: Ivo De Decker <ivodd@debian.org>

To: Mike Gabriel <sunweaver@debian.org>, 779340@bugs.debian.org

Subject: Re: Bug#779340: unblock (pre-approval): marco/1.8.2+dfsg1-5

Date: Fri, 27 Feb 2015 19:41:54 +0100

Control: tags -1 moreinfo

Hi,

On Fri, Feb 27, 2015 at 01:05:17PM +0100, Mike Gabriel wrote:
> Please consider unblocking planned upload of package marco.
> 
> Due to a conceptual change in the Gsettings backend subscription handling
> in glib 2.43, the MATE upstream team stumbled over several code flaws in
> the Gsettings implemenation of MATE. The Debian MATE Packaging Team,
> together with the Ubuntu MATE team, would love to see these code flaws
> fixed in Debian jessie, although Debian jessie ships an older version of
> glib and is not directly affected by these flaws at the moment.
> 
> In MATE, several components erroneously read from Gsettings before they
> are actually connected to their signals. With the change [1] in glib,
> this leads to situations, where changes of certain desktop settings do
> not get propagated to the running application (Gsettings notifications
> fail). The user needs to logout and login to see changes on desktop
> components take effect.
> 
> These problems
> 
>   o will hit Debian, once glib 2.43 gets uploaded
>   o hit derivatives of Debian that ship glib 2.43, currently
>     Ubuntu MATE 15.04
> 
> Indepent from those code flaws being invisible in Debian jessie or not,
> the actual code passages in affected MATE components are technically
> wrong and should be prophylactically fixed in Debian. This will also
> greatly ease the team cooperation with Ubuntu MATE (from where we receive
> a great deal of input and impact).

So you are asking unblocks for issues that don't affect jessie at all? If
that's the case, this change is not appropriate at this point of the freeze,
sorry.

Cheers,

Ivo

Message #17 received at 779340@bugs.debian.org (full text, mbox, reply):

From: Mike Gabriel <sunweaver@debian.org>

To: Ivo De Decker <ivodd@debian.org>

Cc: 779340@bugs.debian.org

Subject: Re: Bug#779340: unblock (pre-approval): marco/1.8.2+dfsg1-5

Date: Fri, 27 Feb 2015 21:48:28 +0000

[Message part 1 (text/plain, inline)]

Control: tags -1 - moreinfo

Hi Ivo,

On  Fr 27 Feb 2015 19:41:54 CET, Ivo De Decker wrote:

> Control: tags -1 moreinfo
>
> Hi,
>
> On Fri, Feb 27, 2015 at 01:05:17PM +0100, Mike Gabriel wrote:
>> Please consider unblocking planned upload of package marco.
>>
>> Due to a conceptual change in the Gsettings backend subscription handling
>> in glib 2.43, the MATE upstream team stumbled over several code flaws in
>> the Gsettings implemenation of MATE. The Debian MATE Packaging Team,
>> together with the Ubuntu MATE team, would love to see these code flaws
>> fixed in Debian jessie, although Debian jessie ships an older version of
>> glib and is not directly affected by these flaws at the moment.
>>
>> In MATE, several components erroneously read from Gsettings before they
>> are actually connected to their signals. With the change [1] in glib,
>> this leads to situations, where changes of certain desktop settings do
>> not get propagated to the running application (Gsettings notifications
>> fail). The user needs to logout and login to see changes on desktop
>> components take effect.
>>
>> These problems
>>
>>   o will hit Debian, once glib 2.43 gets uploaded
>>   o hit derivatives of Debian that ship glib 2.43, currently
>>     Ubuntu MATE 15.04
>>
>> Indepent from those code flaws being invisible in Debian jessie or not,
>> the actual code passages in affected MATE components are technically
>> wrong and should be prophylactically fixed in Debian. This will also
>> greatly ease the team cooperation with Ubuntu MATE (from where we receive
>> a great deal of input and impact).
>
> So you are asking unblocks for issues that don't affect jessie at all? If
> that's the case, this change is not appropriate at this point of the freeze,
> sorry.

No. Martin and I actually ask for unblocks of patches fixing  
conceptual/technical flaws in the MATE upstream code [1,2,3,4,5].

The point is that these flaws currently don't bug users in jessie,  
because these flaws fabulously co-exist with another flaw / conceptual  
weakness in glib that got recently fixed there, too [6].

When MATE upstream implemented gsettings support in MATE, the order of  
accessing gsettings values and connecting to the gsettings  
notifications have been implemented in the wrong order. These flaws  
start popping up in several MATE components if glib >= 2.43.

Please introspect the proposed patches one by one [1,2,3,4,5], they  
are simple and short and swap the order of certain code lines around  
in a very systematical way. The way the gsettings notification stuff  
(signalling) is implemented in the current MATE packages in Debian is  
simply the wrong way round. Upstream has started picking up the  
related pull requests from Monsta (Vlad Orlon), who also is a member  
of our packaging team.  Getting the proposed fixes into jessie will  
protect Debian users from later changes in jessie's glib (in case  
someone comes to the idea and applies [1] to glib-2.0 in jessie  
because of performance issues). Getting the proposed fixes into Debian  
jessie will also prevent people from getting into trouble when mixing  
a jessie system with a newer version of glib (the libglib-2.0 version  
in experimental will trigger the issues I refer to here, already). The  
required changes in MATE are really low-cose (simple and minimal) and  
prevent people from ending up with non-functional desktop settings if  
they run a non-default system setups.

I consider the proposed fixes as means for stabilizing MATE in Debian  
jessie and making it a bit more robust.


light+love,
Mike

[1] libmatekbd:  
https://github.com/mate-desktop/libmatekbd/commit/7e0491525f51c5eaee95ceb5a413e4ca786ea14
[2] marco:  
https://github.com/monsta/marco/commit/f56a1ba0abceaba79efa55120a9e706768809cb4
[3] mate-notification-daemon:  
https://github.com/monsta/mate-notification-daemon/commit/2da0593cad0ef8ab3a05bffa48e5fccd02a2573a
[4] mate-session-manager:  
https://github.com/monsta/mate-session-manager/commit/0376255e93e11d06e368d468de7c1c804e4b4c9c
[5] mate-settings-daemon:  
https://github.com/monsta/mate-settings-daemon/commit/ba8249e03455af1b867298d9d5c868607ab32cdc

[6]  
https://git.gnome.org/browse/glib/commit/?id=8ff5668a458344da22d30491e3ce726d861b3619

-- 

mike gabriel aka sunweaver (Debian Developer)
fon: +49 (1520) 1976 148

GnuPG Fingerprint: 9BFB AEE8 6C0A A5FF BF22  0782 9AF4 6B30 2577 1B31
mail: sunweaver@debian.org, http://sunweavers.net

[Message part 2 (application/pgp-signature, inline)]

Message #24 received at 779340-done@bugs.debian.org (full text, mbox, reply):

From: Mehdi Dogguy <mehdi@dogguy.org>

To: Mike Gabriel <sunweaver@debian.org>, 779340-done@bugs.debian.org

Cc: Ivo De Decker <ivodd@debian.org>

Subject: Re: Bug#779340: unblock (pre-approval): marco/1.8.2+dfsg1-5

Date: Sat, 28 Feb 2015 19:14:34 +0100

On Fri, Feb 27, 2015 at 09:48:28PM +0000, Mike Gabriel <sunweaver@debian.org> wrote:
> 
> No. Martin and I actually ask for unblocks of patches fixing
> conceptual/technical flaws in the MATE upstream code [1,2,3,4,5].
> 
> The point is that these flaws currently don't bug users in jessie, because
> these flaws fabulously co-exist with another flaw / conceptual weakness in
> glib that got recently fixed there, too [6].
> 
> When MATE upstream implemented gsettings support in MATE, the order of
> accessing gsettings values and connecting to the gsettings notifications
> have been implemented in the wrong order. These flaws start popping up in
> several MATE components if glib >= 2.43.
> 

We do understand your point of view and we share your concerns to some
degree. According to what you say, Jessie users are safe wrt. the bugs
fixed by the attached patches. AFAIU, theses changes will not have any
visible effect to users and only help to get the code "more correct".
Hence, we still feel that those changes are not approprite at this
stage of the freeze where we focus on getting the list fo release
critical bugs fixed.

So, we are going to close this request and hope that we understood
each other. We are also maintainers of packages and can understand why
this is not "perfect"... But, at some point, we should better identify
our priorities: Fixing RC bugs and getting Jessie released. You can
help Jessie users also by working on testing various upgrade
scenarios, getting release notes ready, helping the installer team,
etc...

Regards,

-- 
Mehdi Dogguy

Send a report that this bug log contains spam.