Re: [Sugar-devel] datastore get fails

[Gonzalo Odiard]

I can't guess with the line alonewithout the context.
Could you upload the sources anywhere?

Gonzalo

On Thu, Sep 3, 2015 at 12:31 PM, Tony Anderson 
wrote:

> Hi, Gonzalo
>
> I may have a 'smoking gun' on the problem with the datastore.get.
>
> The relevant code snippet is attached as is the activity log.
>
> The exception occurred on:
>
> obj = datastore.get(OBJ)
>
> where OBJ is the handle.object_id.
>
> The key line in the log is:
>
> ** glibc detected *** python2: munmap_chunk(): invalid pointer: 0x007bc5d8
> ***
>
> This error message seems consistent with the reports in other runs of a
> segmentation fault and terminated on signal 11.
>
> Hope this helps.
>
> Tony
>



-- 
Gonzalo Odiard

SugarLabs - Software [for | by] children learning
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Martin Abente]

On Thu, Sep 3, 2015 at 1:12 PM, Jerry Vonau  wrote:

>
>
> > On September 3, 2015 at 11:38 AM Martin Abente
> >  wrote:
> >
> >
> > Hello everyone,
> >
> > I rebuilt 0.106.0 packages from fc23 source rpms, for fc22 x86_64.
>
> So did I using my copr account, what a great tool to build rpms with,
> perhaps SL should go back to using it.[1] Easy for testers to enable with
> dnf.
>
> > It works
> > fine, so it definitely not something with Sugar. It makes sense since
> > nothing changed in SugarExt.
> >
> > Now, this might be something with fc23, but is it on build time or run
> > time? Ideas? Maybe something related to gobject instrospection?
> >
>
> Runtime in F23, otherwise the F23 rpm would not been built.
>

Not necessarily, e.g. there could be something wrong happening while
generating the gobject introspection files. Something wrong in the sense
that the content is not generated "properly", and does not necessarily
triggers a "compilation error".

In fact, I think this might be the case: I installed the 0.106.0 packages I
built for fc22 on a fresh fc23 system and sugar works fine.

Ideas? I suspect on the gobject introspection bindings generation, but
can't think of what exactly it could be.

Refs:
[1] http://people.sugarlabs.org/~tch/tests/fedora22/


> Jerry
>
>
> 1. https://copr.fedoraproject.org/coprs/dnarvaez/sugar
>
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Peter Robinson]

>> > Hello everyone,
>> >
>> > I rebuilt 0.106.0 packages from fc23 source rpms, for fc22 x86_64.
>>
>> So did I using my copr account, what a great tool to build rpms with,
>> perhaps SL should go back to using it.[1] Easy for testers to enable with
>> dnf.
>>
>> > It works
>> > fine, so it definitely not something with Sugar. It makes sense since
>> > nothing changed in SugarExt.
>> >
>> > Now, this might be something with fc23, but is it on build time or run
>> > time? Ideas? Maybe something related to gobject instrospection?
>> >
>>
>> Runtime in F23, otherwise the F23 rpm would not been built.
>
>
> Not necessarily, e.g. there could be something wrong happening while
> generating the gobject introspection files. Something wrong in the sense
> that the content is not generated "properly", and does not necessarily
> triggers a "compilation error".
>
> In fact, I think this might be the case: I installed the 0.106.0 packages I
> built for fc22 on a fresh fc23 system and sugar works fine.
>
> Ideas? I suspect on the gobject introspection bindings generation, but can't
> think of what exactly it could be.

I don't have the time, or really the knowledge, to sort this out but
we freeze next Tuesday and I really need someone to pretty much sort
this out by then else we'll have issues with beta and it'll cause me
stress.

The latest test beta is link is below.

Peter

https://alt.fedoraproject.org/pub/alt/stage/23_Beta_TC1/Live/x86_64/Fedora-Live-SoaS-x86_64-23_Beta-TC1.iso
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

[Sugar-devel] Want to contribute

[Utkarsh Tiwari]

Hello!

 I would like to contribute to some sugar project. Could anyone
please guide me. I have developed around 5 activities. Want to work on some
core project.

Thank You.
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Jerry Vonau]

> On September 3, 2015 at 11:38 AM Martin Abente
>  wrote:
> 
> 
> Hello everyone,
> 
> I rebuilt 0.106.0 packages from fc23 source rpms, for fc22 x86_64. 

So did I using my copr account, what a great tool to build rpms with,
perhaps SL should go back to using it.[1] Easy for testers to enable with
dnf.
 
> It works
> fine, so it definitely not something with Sugar. It makes sense since
> nothing changed in SugarExt.
> 
> Now, this might be something with fc23, but is it on build time or run
> time? Ideas? Maybe something related to gobject instrospection?
> 

Runtime in F23, otherwise the F23 rpm would not been built.

Jerry


1. https://copr.fedoraproject.org/coprs/dnarvaez/sugar
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Martin Abente]

Hello everyone,

I rebuilt 0.106.0 packages from fc23 source rpms, for fc22 x86_64. It works
fine, so it definitely not something with Sugar. It makes sense since
nothing changed in SugarExt.

Now, this might be something with fc23, but is it on build time or run
time? Ideas? Maybe something related to gobject instrospection?

tch.

Refs:
[1] http://people.sugarlabs.org/~tch/tests/fedora22/


On Thu, Sep 3, 2015 at 3:45 AM, Jerry Vonau  wrote:

>
>
> > On September 3, 2015 at 1:34 AM James Cameron  wrote:
> >
> >
> > On Thu, Sep 03, 2015 at 01:21:19AM -0500, Jerry Vonau wrote:
> > >
> > >
> > > > On September 3, 2015 at 12:38 AM James Cameron 
> > > > wrote:
> > > >
> > > >
> > > > On Thu, Sep 03, 2015 at 12:11:38AM -0500, Jerry Vonau wrote:
> > > > > F22 upgrades clean to 106, feels pretty zippy in a VM. Back to
> > > > > figuring out what is now different in the dependencies given the
> > > > > logs[1] for the builds look more or less the same F21 <-> F23.
> > > > > There
> > > > > has to be a rpm package with a library that was split somewhere,
> > > > > sugar's toolkit wants to use something but is left unsatisfied.
> > > >
> > > > It might be interesting to see if the F23 build of 0.104.1 also has
> > > > the problem on F23.  As further proof that it isn't anything Sugar
> > > > did.
> > > >
> > > I can get to a cmdline via the bootprompt using the F23 iso, so I'll do
> > > the
> > > 104 on F23 test, just not tonight.
> > >
> > > It's not what sugar did, but didn't do yet, adapt to the new build
> > > environment for F23. Is the current recommended sugar development
> > > still sugar-build at EOL'd F20?
> >
> > Good question.  Don't know.  Hopefully someone from Sugar Labs can
> > tell us.
> >
> > Although sugar-build hides all these kinds of problems by building on
> > the target system instead of packaging.
> >
>
> I know I agree, that is why I'm looking at the devel packages that are
> installed there, make a change recompile and test, but a quick way to lose
> track of a dependence... in which package or is that the lib package or...
> You get the idea.
>
> > sugar-build is not used when deploying, so it seems unwise to use it
> > for development.  ;-)
> >
>
> Yea but some improvements got written while using it.
>
> > > > Comparing the logs of 0.104.1 build for x86_64 between F22 and F23,
> > > > the only interesting change that springs up is;
> > > >
> > > > -LDFLAGS='-Wl,-z,relro '
> > > > +LDFLAGS='-Wl,-z,relro -specs=/usr/lib/rpm/redhat/redhat-hardened-ld'
> > > >
> > > > Doesn't seem related to the reported problem, however.
> > >
> > > I tend to agree, but I'm not sure of the full effect of the flag. I
> > > think there is a package that was split and a dependence needed by
> > > sugar/toolkit moved to a new sub-package and is not in what is
> > > declared as (Build)Requires anymore. If this is true then the new
> > > sub-package should now be used in the spec file somewhere, just
> > > finding the package.
> >
> > Hmm, doesn't seem related to the sugarext library missing from the
> > expected directory.  Maybe I've misunderstood the problem.
> >
>
> I wanted to poke around a bit because of Peter comment about s/b /usr/lib64
> in bugzilla for info. That why I booted to the cmdline, filesystem looks
> right /lib -> /usr/lib  /lib64 -> /usr/lib64 with libsugarext and
> libsugar-eventcontrollor present.
> ___
> Sugar-devel mailing list
> Sugar-devel@lists.sugarlabs.org
> http://lists.sugarlabs.org/listinfo/sugar-devel
>
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Martin Abente]

On Thu, Sep 3, 2015 at 3:58 PM, Jerry Vonau  wrote:

>
>
> > On September 3, 2015 at 12:58 PM Martin Abente
> >  wrote:
> >
> >
> > On Thu, Sep 3, 2015 at 1:12 PM, Jerry Vonau  wrote:
> >
> > >
> > >
> > > > On September 3, 2015 at 11:38 AM Martin Abente
> > > >  wrote:
> > > >
> > > >
> > > > Hello everyone,
> > > >
> > > > I rebuilt 0.106.0 packages from fc23 source rpms, for fc22 x86_64.
> > >
> > > So did I using my copr account, what a great tool to build rpms with,
> > > perhaps SL should go back to using it.[1] Easy for testers to enable
> > > with
> > > dnf.
> > >
> > > > It works
> > > > fine, so it definitely not something with Sugar. It makes sense since
> > > > nothing changed in SugarExt.
> > > >
> > > > Now, this might be something with fc23, but is it on build time or
> > > > run
> > > > time? Ideas? Maybe something related to gobject instrospection?
> > > >
> > >
> > > Runtime in F23, otherwise the F23 rpm would not been built.
> > >
> >
> > Not necessarily, e.g. there could be something wrong happening while
> > generating the gobject introspection files. Something wrong in the sense
> > that the content is not generated "properly", and does not necessarily
> > triggers a "compilation error".
> >
>
> Agreed, looking for what changed is a pain.
>
> > In fact, I think this might be the case: I installed the 0.106.0 packages
> > I
> > built for fc22 on a fresh fc23 system and sugar works fine.
> >
>
> Interesting, what is the "fresh system"? Are you selecting sugar at the
> login manager or using sugar-runner?
>

a) fresh system == a new installation of fc23 Alpha-2 x86_64 on a VM.
b) I am selecting the Sugar option on the logging screen, but it also works
if you type "sugar" on a terminal in the gnome session.


> There could be dependencies that are already met by the installed system,
> that would mask an undeclared dependency in the sugar rpms. Kind of
> re-enforces what I'm thinking, some Build|Requires moved to a sub-package
> for a sugar dependency or there is a new unaccounted sugar dependency.
>
>
No, it doesn't seem to be a problem with the run-time environment.

I tried creating another VM with fc23, installed the fc23 sugar packages
and sugar fails to start [1].  Note that the error is slightly different,
but the is the same issue.

So, looking at all the tests I did, the only difference is the environment
in which the packages are built. If the packages are built for fc22 sugar
0.106 works even on fc23, but if its built for fc23 it doesn't work.

Refs:
[1] http://fpaste.org/263354/31582414/

> Ideas? I suspect on the gobject introspection bindings generation, but
> > can't think of what exactly it could be.
> >
> Wouldn't the recent addition of importing configparser require the
> python-configparser rpm to be declared as Required, not relying on to be
> already present? Might explain your observations with a different desktop.
>
> Jerry
>
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Peter Robinson]

>> > > Hello everyone,
>> > >
>> > > I rebuilt 0.106.0 packages from fc23 source rpms, for fc22 x86_64.
>> >
>> > So did I using my copr account, what a great tool to build rpms with,
>> > perhaps SL should go back to using it.[1] Easy for testers to enable
>> > with
>> > dnf.
>> >
>> > > It works
>> > > fine, so it definitely not something with Sugar. It makes sense since
>> > > nothing changed in SugarExt.
>> > >
>> > > Now, this might be something with fc23, but is it on build time or
>> > > run
>> > > time? Ideas? Maybe something related to gobject instrospection?
>> > >
>> >
>> > Runtime in F23, otherwise the F23 rpm would not been built.
>> >
>>
>> Not necessarily, e.g. there could be something wrong happening while
>> generating the gobject introspection files. Something wrong in the sense
>> that the content is not generated "properly", and does not necessarily
>> triggers a "compilation error".
>>
>
> Agreed, looking for what changed is a pain.
>
>> In fact, I think this might be the case: I installed the 0.106.0 packages
>> I
>> built for fc22 on a fresh fc23 system and sugar works fine.
>>
>
> Interesting, what is the "fresh system"? Are you selecting sugar at the
> login manager or using sugar-runner?

Fedora 23, either gdm (from an upgraded netbook) or lightdm for the
clean built LiveCD

> There could be dependencies that are already met by the installed system,
> that would mask an undeclared dependency in the sugar rpms. Kind of
> re-enforces what I'm thinking, some Build|Requires moved to a sub-package
> for a sugar dependency or there is a new unaccounted sugar dependency.

Tha't's possibly if there was some new dep that wasn't advertised in
the changes.

>> Ideas? I suspect on the gobject introspection bindings generation, but
>> can't think of what exactly it could be.
>>
> Wouldn't the recent addition of importing configparser require the
> python-configparser rpm to be declared as Required, not relying on to be
> already present? Might explain your observations with a different desktop.

Installing that doesn't make any difference to the crash, is it
needed? I never saw anything in any of the release notes, not that
there really were with any of the dev cycle releases.

http://paste.fedoraproject.org/263319/31103314/
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Jerry Vonau]

> On September 3, 2015 at 12:38 AM James Cameron  wrote:
> 
> 
> On Thu, Sep 03, 2015 at 12:11:38AM -0500, Jerry Vonau wrote:
> > F22 upgrades clean to 106, feels pretty zippy in a VM. Back to
> > figuring out what is now different in the dependencies given the
> > logs[1] for the builds look more or less the same F21 <-> F23. There
> > has to be a rpm package with a library that was split somewhere,
> > sugar's toolkit wants to use something but is left unsatisfied.
> 
> It might be interesting to see if the F23 build of 0.104.1 also has
> the problem on F23.  As further proof that it isn't anything Sugar did.
> 
I can get to a cmdline via the bootprompt using the F23 iso, so I'll do the
104 on F23 test, just not tonight. 

It's not what sugar did, but didn't do yet, adapt to the new build
environment for F23. Is the current recommended sugar development still
sugar-build at EOL'd F20?

> Comparing the logs of 0.104.1 build for x86_64 between F22 and F23,
> the only interesting change that springs up is;
> 
> -LDFLAGS='-Wl,-z,relro '
> +LDFLAGS='-Wl,-z,relro -specs=/usr/lib/rpm/redhat/redhat-hardened-ld'
> 
> Doesn't seem related to the reported problem, however.

I tend to agree, but I'm not sure of the full effect of the flag. I think
there is a package that was split and a dependence needed by sugar/toolkit
moved to a new sub-package and is not in what is declared as
(Build)Requires anymore. If this is true then the new sub-package should
now be used in the spec file somewhere, just finding the package.

 
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[James Cameron]

On Thu, Sep 03, 2015 at 07:12:18AM +, Sam P. wrote:
> Hi James,
> 
> On Thu, 3 Sep 2015 4:34 pm James Cameron <[1]qu...@laptop.org> wrote:
> 
> On Thu, Sep 03, 2015 at 01:21:19AM -0500, Jerry Vonau wrote:
> >
> >
> > > On September 3, 2015 at 12:38 AM James Cameron <[2]qu...@laptop.org>
> wrote:
> > >
> > >
> > > On Thu, Sep 03, 2015 at 12:11:38AM -0500, Jerry Vonau wrote:
> > > > F22 upgrades clean to 106, feels pretty zippy in a VM. Back to
> > > > figuring out what is now different in the dependencies given the
> > > > logs[1] for the builds look more or less the same F21 <-> F23. There
> > > > has to be a rpm package with a library that was split somewhere,
> > > > sugar's toolkit wants to use something but is left unsatisfied.
> > >
> > > It might be interesting to see if the F23 build of 0.104.1 also has
> > > the problem on F23.  As further proof that it isn't anything Sugar 
> did.
> > >
> > I can get to a cmdline via the bootprompt using the F23 iso, so I'll do
> the
> > 104 on F23 test, just not tonight.
> >
> > It's not what sugar did, but didn't do yet, adapt to the new build
> > environment for F23. Is the current recommended sugar development
> > still sugar-build at EOL'd F20?
> 
> Good question.  Don't know.  Hopefully someone from Sugar Labs can
> tell us.
> 
> I'm pretty sure it is.  But most of the maintainers disable broot and just run
> in natively in Fedora.
> 
> Although sugar-build hides all these kinds of problems by building on
> the target system instead of packaging.
> 
> sugar-build is not used when deploying, so it seems unwise to use it
> for development.  ;-)
> 
> Works fine if you disable broot.  Maybe we should change defaults (no broot 
> for
> Fedora ever).
> 
> Otherwise there is really nothing wrong with a bunch of build scripts and
> package install scripts IMHO :)

Oh good, so you're saying there is a way we can build Sugar into a
deployment image without having to use RPMs for it?  Or do you mean
that sugar-build can generate the RPMs?

How do we avoid placing unnecessary build dependencies into the
deployment image?

> 
> Thanks,
> Sam
> 
> > > Comparing the logs of 0.104.1 build for x86_64 between F22 and F23,
> > > the only interesting change that springs up is;
> > >
> > > -LDFLAGS='-Wl,-z,relro '
> > > +LDFLAGS='-Wl,-z,relro -specs=/usr/lib/rpm/redhat/redhat-hardened-ld'
> > >
> > > Doesn't seem related to the reported problem, however.
> >
> > I tend to agree, but I'm not sure of the full effect of the flag. I
> > think there is a package that was split and a dependence needed by
> > sugar/toolkit moved to a new sub-package and is not in what is
> > declared as (Build)Requires anymore. If this is true then the new
> > sub-package should now be used in the spec file somewhere, just
> > finding the package.
> 
> Hmm, doesn't seem related to the sugarext library missing from the
> expected directory.  Maybe I've misunderstood the problem.
> 
> --
> James Cameron
> [3]http://quozl.linux.org.au/
> ___
> Sugar-devel mailing list
> [4]Sugar-devel@lists.sugarlabs.org
> [5]http://lists.sugarlabs.org/listinfo/sugar-devel
> 
> References:
> 
> [1] mailto:qu...@laptop.org
> [2] mailto:qu...@laptop.org
> [3] http://quozl.linux.org.au/
> [4] mailto:Sugar-devel@lists.sugarlabs.org
> [5] http://lists.sugarlabs.org/listinfo/sugar-devel

-- 
James Cameron
http://quozl.linux.org.au/
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[James Cameron]

On Thu, Sep 03, 2015 at 01:21:19AM -0500, Jerry Vonau wrote:
> 
> 
> > On September 3, 2015 at 12:38 AM James Cameron  wrote:
> > 
> > 
> > On Thu, Sep 03, 2015 at 12:11:38AM -0500, Jerry Vonau wrote:
> > > F22 upgrades clean to 106, feels pretty zippy in a VM. Back to
> > > figuring out what is now different in the dependencies given the
> > > logs[1] for the builds look more or less the same F21 <-> F23. There
> > > has to be a rpm package with a library that was split somewhere,
> > > sugar's toolkit wants to use something but is left unsatisfied.
> > 
> > It might be interesting to see if the F23 build of 0.104.1 also has
> > the problem on F23.  As further proof that it isn't anything Sugar did.
> > 
> I can get to a cmdline via the bootprompt using the F23 iso, so I'll do the
> 104 on F23 test, just not tonight. 
> 
> It's not what sugar did, but didn't do yet, adapt to the new build
> environment for F23. Is the current recommended sugar development
> still sugar-build at EOL'd F20?

Good question.  Don't know.  Hopefully someone from Sugar Labs can
tell us.

Although sugar-build hides all these kinds of problems by building on
the target system instead of packaging.

sugar-build is not used when deploying, so it seems unwise to use it
for development.  ;-)

> > Comparing the logs of 0.104.1 build for x86_64 between F22 and F23,
> > the only interesting change that springs up is;
> > 
> > -LDFLAGS='-Wl,-z,relro '
> > +LDFLAGS='-Wl,-z,relro -specs=/usr/lib/rpm/redhat/redhat-hardened-ld'
> > 
> > Doesn't seem related to the reported problem, however.
> 
> I tend to agree, but I'm not sure of the full effect of the flag. I
> think there is a package that was split and a dependence needed by
> sugar/toolkit moved to a new sub-package and is not in what is
> declared as (Build)Requires anymore. If this is true then the new
> sub-package should now be used in the spec file somewhere, just
> finding the package.

Hmm, doesn't seem related to the sugarext library missing from the
expected directory.  Maybe I've misunderstood the problem.

-- 
James Cameron
http://quozl.linux.org.au/
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Sam P.]

Hi James,

On Thu, 3 Sep 2015 4:34 pm James Cameron  wrote:

On Thu, Sep 03, 2015 at 01:21:19AM -0500, Jerry Vonau wrote:
>
>
> > On September 3, 2015 at 12:38 AM James Cameron  wrote:
> >
> >
> > On Thu, Sep 03, 2015 at 12:11:38AM -0500, Jerry Vonau wrote:
> > > F22 upgrades clean to 106, feels pretty zippy in a VM. Back to
> > > figuring out what is now different in the dependencies given the
> > > logs[1] for the builds look more or less the same F21 <-> F23. There
> > > has to be a rpm package with a library that was split somewhere,
> > > sugar's toolkit wants to use something but is left unsatisfied.
> >
> > It might be interesting to see if the F23 build of 0.104.1 also has
> > the problem on F23.  As further proof that it isn't anything Sugar did.
> >
> I can get to a cmdline via the bootprompt using the F23 iso, so I'll do
the
> 104 on F23 test, just not tonight.
>
> It's not what sugar did, but didn't do yet, adapt to the new build
> environment for F23. Is the current recommended sugar development
> still sugar-build at EOL'd F20?

Good question.  Don't know.  Hopefully someone from Sugar Labs can
tell us.

I'm pretty sure it is.  But most of the maintainers disable broot and just
run in natively in Fedora.

Although sugar-build hides all these kinds of problems by building on
the target system instead of packaging.

sugar-build is not used when deploying, so it seems unwise to use it
for development.  ;-)

Works fine if you disable broot.  Maybe we should change defaults (no broot
for Fedora ever).

Otherwise there is really nothing wrong with a bunch of build scripts and
package install scripts IMHO :)

Thanks,
Sam

> > Comparing the logs of 0.104.1 build for x86_64 between F22 and F23,
> > the only interesting change that springs up is;
> >
> > -LDFLAGS='-Wl,-z,relro '
> > +LDFLAGS='-Wl,-z,relro -specs=/usr/lib/rpm/redhat/redhat-hardened-ld'
> >
> > Doesn't seem related to the reported problem, however.
>
> I tend to agree, but I'm not sure of the full effect of the flag. I
> think there is a package that was split and a dependence needed by
> sugar/toolkit moved to a new sub-package and is not in what is
> declared as (Build)Requires anymore. If this is true then the new
> sub-package should now be used in the spec file somewhere, just
> finding the package.

Hmm, doesn't seem related to the sugarext library missing from the
expected directory.  Maybe I've misunderstood the problem.

--
James Cameron
http://quozl.linux.org.au/
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Jerry Vonau]

> On September 3, 2015 at 1:34 AM James Cameron  wrote:
> 
> 
> On Thu, Sep 03, 2015 at 01:21:19AM -0500, Jerry Vonau wrote:
> > 
> > 
> > > On September 3, 2015 at 12:38 AM James Cameron 
> > > wrote:
> > > 
> > > 
> > > On Thu, Sep 03, 2015 at 12:11:38AM -0500, Jerry Vonau wrote:
> > > > F22 upgrades clean to 106, feels pretty zippy in a VM. Back to
> > > > figuring out what is now different in the dependencies given the
> > > > logs[1] for the builds look more or less the same F21 <-> F23.
> > > > There
> > > > has to be a rpm package with a library that was split somewhere,
> > > > sugar's toolkit wants to use something but is left unsatisfied.
> > > 
> > > It might be interesting to see if the F23 build of 0.104.1 also has
> > > the problem on F23.  As further proof that it isn't anything Sugar
> > > did.
> > > 
> > I can get to a cmdline via the bootprompt using the F23 iso, so I'll do
> > the
> > 104 on F23 test, just not tonight. 
> > 
> > It's not what sugar did, but didn't do yet, adapt to the new build
> > environment for F23. Is the current recommended sugar development
> > still sugar-build at EOL'd F20?
> 
> Good question.  Don't know.  Hopefully someone from Sugar Labs can
> tell us.
> 
> Although sugar-build hides all these kinds of problems by building on
> the target system instead of packaging.
> 

I know I agree, that is why I'm looking at the devel packages that are
installed there, make a change recompile and test, but a quick way to lose
track of a dependence... in which package or is that the lib package or...
You get the idea.

> sugar-build is not used when deploying, so it seems unwise to use it
> for development.  ;-)
>

Yea but some improvements got written while using it. 
 
> > > Comparing the logs of 0.104.1 build for x86_64 between F22 and F23,
> > > the only interesting change that springs up is;
> > > 
> > > -LDFLAGS='-Wl,-z,relro '
> > > +LDFLAGS='-Wl,-z,relro -specs=/usr/lib/rpm/redhat/redhat-hardened-ld'
> > > 
> > > Doesn't seem related to the reported problem, however.
> > 
> > I tend to agree, but I'm not sure of the full effect of the flag. I
> > think there is a package that was split and a dependence needed by
> > sugar/toolkit moved to a new sub-package and is not in what is
> > declared as (Build)Requires anymore. If this is true then the new
> > sub-package should now be used in the spec file somewhere, just
> > finding the package.
> 
> Hmm, doesn't seem related to the sugarext library missing from the
> expected directory.  Maybe I've misunderstood the problem.
> 

I wanted to poke around a bit because of Peter comment about s/b /usr/lib64
in bugzilla for info. That why I booted to the cmdline, filesystem looks
right /lib -> /usr/lib  /lib64 -> /usr/lib64 with libsugarext and
libsugar-eventcontrollor present.
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[James Cameron]

The errors complain of missing symbols;

- sugar_event_controller_get_state

- acme_volume_alsa_new

Which were both added after 0.98.

On my F18 test systems, the package sugar-toolkit-0.98 brings in a
_sugarext.so file in /usr/lib/python2.7/site-packages/sugar/ which
does not have these symbols.

If this file were being used instead, then the outcome may be as
reported.

Is the package present?

If so, does removing the .so file fix the problem?

Use of the file could also be verified with strace -e open.

On Thu, Sep 03, 2015 at 05:39:19PM -0400, Martin Abente wrote:
> I tried creating another VM with fc23, installed the fc23 sugar
> packages and sugar fails to start [1].  Note that the error is
> slightly different, but the is the same issue.
> [1] http://fpaste.org/263354/31582414/

** (main.py:1818): WARNING **: Failed to load shared library
   'libsugarext.so.0' referenced by the typelib:
   /lib64/libsugarext.so.0: undefined symbol:
   sugar_event_controller_get_state

Compare with previous report
http://paste.fedoraproject.org/260680/69152144/

** (main.py:2637): WARNING **: Failed to load shared library
   'libsugarext.so.0' referenced by the typelib:
   /lib/libsugarext.so.0: undefined symbol:
   sugar_event_controller_get_state

The line number is different.

@All, what is main.py and where are lines 1818 and 2637?

src/jarabe/main.py is much smaller, only 425 lines.

Some other diagnosis methods to suggest:

a.  in a fresh python process, try the relevant imports, e.g.

from jarabe.view import gesturehandler
from jarabe.model.sound import sound

b.  verify a symlink is present from /usr/lib/libsugarext.so.0 to
libsugarext.so.0.0.0 (or equivalent functionality, as this is a
guess from my experience with F18),

c.  use strings to check libsugarext.so.0, see if
sugar_event_controller_get_state is present,

strings /usr/lib/libsugarext.so.0 | \
grep sugar_event_controller_get_state

-- 
James Cameron
http://quozl.linux.org.au/
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Martin Abente]

On Thu, Sep 3, 2015 at 6:43 PM, James Cameron  wrote:

> The errors complain of missing symbols;
>
> - sugar_event_controller_get_state
>
> - acme_volume_alsa_new
>
> Which were both added after 0.98.
>
> On my F18 test systems, the package sugar-toolkit-0.98 brings in a
> _sugarext.so file in /usr/lib/python2.7/site-packages/sugar/ which
> does not have these symbols.
>
> If this file were being used instead, then the outcome may be as
> reported.
>
> Is the package present?
>

No, is not present. Also, that does not explain why the same 106 packages
work perfectly fine on fc23 when are built for fc22, while they fail when
built for fc23.

As you mentioned in a previous reply, the only difference in build time I
also noticed, looking at the logs, is the
-specs=/usr/lib/rpm/redhat/redhat-hardened-ld'
flag. which is passed to all gir-* (gobject introspection) calls.


>
> If so, does removing the .so file fix the problem?
>
> Use of the file could also be verified with strace -e open.
>
> On Thu, Sep 03, 2015 at 05:39:19PM -0400, Martin Abente wrote:
> > I tried creating another VM with fc23, installed the fc23 sugar
> > packages and sugar fails to start [1].  Note that the error is
> > slightly different, but the is the same issue.
> > [1] http://fpaste.org/263354/31582414/
>
> ** (main.py:1818): WARNING **: Failed to load shared library
>'libsugarext.so.0' referenced by the typelib:
>/lib64/libsugarext.so.0: undefined symbol:
>sugar_event_controller_get_state
>
> Compare with previous report
> http://paste.fedoraproject.org/260680/69152144/
>
> ** (main.py:2637): WARNING **: Failed to load shared library
>'libsugarext.so.0' referenced by the typelib:
>/lib/libsugarext.so.0: undefined symbol:
>sugar_event_controller_get_state
>
> The line number is different.
>
> @All, what is main.py and where are lines 1818 and 2637?
>
> src/jarabe/main.py is much smaller, only 425 lines.
>
> Some other diagnosis methods to suggest:
>
> a.  in a fresh python process, try the relevant imports, e.g.
>
> from jarabe.view import gesturehandler
> from jarabe.model.sound import sound
>
> b.  verify a symlink is present from /usr/lib/libsugarext.so.0 to
> libsugarext.so.0.0.0 (or equivalent functionality, as this is a
> guess from my experience with F18),
>
> c.  use strings to check libsugarext.so.0, see if
> sugar_event_controller_get_state is present,
>
> strings /usr/lib/libsugarext.so.0 | \
> grep sugar_event_controller_get_state
>
> --
> James Cameron
> http://quozl.linux.org.au/
>
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Re: [Sugar-devel] Bug 1240354 - SoaS live x86_64 20150706 does not login from live system user

[Jerry Vonau]

> On September 3, 2015 at 12:58 PM Martin Abente
>  wrote:
> 
> 
> On Thu, Sep 3, 2015 at 1:12 PM, Jerry Vonau  wrote:
> 
> >
> >
> > > On September 3, 2015 at 11:38 AM Martin Abente
> > >  wrote:
> > >
> > >
> > > Hello everyone,
> > >
> > > I rebuilt 0.106.0 packages from fc23 source rpms, for fc22 x86_64.
> >
> > So did I using my copr account, what a great tool to build rpms with,
> > perhaps SL should go back to using it.[1] Easy for testers to enable
> > with
> > dnf.
> >
> > > It works
> > > fine, so it definitely not something with Sugar. It makes sense since
> > > nothing changed in SugarExt.
> > >
> > > Now, this might be something with fc23, but is it on build time or
> > > run
> > > time? Ideas? Maybe something related to gobject instrospection?
> > >
> >
> > Runtime in F23, otherwise the F23 rpm would not been built.
> >
> 
> Not necessarily, e.g. there could be something wrong happening while
> generating the gobject introspection files. Something wrong in the sense
> that the content is not generated "properly", and does not necessarily
> triggers a "compilation error".
> 

Agreed, looking for what changed is a pain.

> In fact, I think this might be the case: I installed the 0.106.0 packages
> I
> built for fc22 on a fresh fc23 system and sugar works fine.
> 

Interesting, what is the "fresh system"? Are you selecting sugar at the
login manager or using sugar-runner?

There could be dependencies that are already met by the installed system,
that would mask an undeclared dependency in the sugar rpms. Kind of
re-enforces what I'm thinking, some Build|Requires moved to a sub-package
for a sugar dependency or there is a new unaccounted sugar dependency.

> Ideas? I suspect on the gobject introspection bindings generation, but
> can't think of what exactly it could be.
> 
Wouldn't the recent addition of importing configparser require the
python-configparser rpm to be declared as Required, not relying on to be
already present? Might explain your observations with a different desktop.

Jerry
___
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel