[gentoo-user] revdep-rebuild problem

[gentoo-user] revdep-rebuild problem

[John P. Burkett]

Doing revdep-rebuild on an amd64 machine, I received a response
the included the following lines:
* All prepared. Starting rebuild
emerge --oneshot  app-text/xpdf:0
gnome-base/gnome-panel:0
kde-base/kdegraphics:3.5
mail-client/-MERGING-evolution:2.0
media-plugins/gst-plugins-faad:0.8
media-plugins/xmms-alsa:0
media-plugins/xmms-vorbis:0
media-video/totem:0
..........
Calculating dependencies... done!
emerge: there are no ebuilds to satisfy
"mail-client/-MERGING-evolution:2.0".

After doing emerge -C evolution, I redid revdep-rebuild but got the same
response. After doing emerge evolution, I again redid revdep-rebuild,
with the same results.

Suggestions for how to successfully run revdep-rebuild would be most
welcome.  I'm willing to sacrifice evolution if that would help.

-- 
John P. Burkett
Department of Economics
University of Rhode Island
Kingston, RI 02881-0808
USA

phone (401) 874-9195

Re: [gentoo-user] revdep-rebuild problem

[Nick Fortino]

John P. Burkett wrote:
> Doing revdep-rebuild on an amd64 machine, I received a response
> the included the following lines:
> * All prepared. Starting rebuild
> emerge --oneshot  app-text/xpdf:0
> gnome-base/gnome-panel:0
> kde-base/kdegraphics:3.5
> mail-client/-MERGING-evolution:2.0
> media-plugins/gst-plugins-faad:0.8
> media-plugins/xmms-alsa:0
> media-plugins/xmms-vorbis:0
> media-video/totem:0
> ..........
> Calculating dependencies... done!
> emerge: there are no ebuilds to satisfy
> "mail-client/-MERGING-evolution:2.0".
>
> After doing emerge -C evolution, I redid revdep-rebuild but got the same
> response. After doing emerge evolution, I again redid revdep-rebuild,
> with the same results.
>
> Suggestions for how to successfully run revdep-rebuild would be most
> welcome.  I'm willing to sacrifice evolution if that would help.
>
>   
Hmm, somehow portage got the idea there is a package named
mail-client/-MERGING-evolution, which is false. If you are lucky:
emerge -avC mail-client/-MERGING-evolution

Although that could fail, as package names aren't supposed to start with
-, so who knows what portage does with it. Next shot, clear out anything
you don't need (make sure to check the output, although these days this
command is safe unless you have done something strange).
emerge -av --depclean

Final shot is to play around in /var/db/pkg. This is where portage keeps
track of what is installed, and how, so messing around here is
inherently unsafe. I have my suspicions as to the correct thing to do
here, but I won't post it since I'm not terribly sure. Someone with more
expertise could help here, but I suspect either method 1 or 2 will work.

Nick

Re: [gentoo-user] revdep-rebuild problem

[John P. Burkett]

Nick Fortino wrote:
> John P. Burkett wrote:
>> Doing revdep-rebuild on an amd64 machine, I received a response
>> the included the following lines:
>> * All prepared. Starting rebuild
>> emerge --oneshot  app-text/xpdf:0
>> gnome-base/gnome-panel:0
>> kde-base/kdegraphics:3.5
>> mail-client/-MERGING-evolution:2.0
>> media-plugins/gst-plugins-faad:0.8
>> media-plugins/xmms-alsa:0
>> media-plugins/xmms-vorbis:0
>> media-video/totem:0
>> ..........
>> Calculating dependencies... done!
>> emerge: there are no ebuilds to satisfy
>> "mail-client/-MERGING-evolution:2.0".
>>
>> After doing emerge -C evolution, I redid revdep-rebuild but got the same
>> response. After doing emerge evolution, I again redid revdep-rebuild,
>> with the same results.
>>
>> Suggestions for how to successfully run revdep-rebuild would be most
>> welcome.  I'm willing to sacrifice evolution if that would help.
>>
>>   
> Hmm, somehow portage got the idea there is a package named
> mail-client/-MERGING-evolution, which is false.
Nick, thank you for the diagnosis.
 If you are lucky:
> emerge -avC mail-client/-MERGING-evolution
The systems's response to this was
>>> These are the packages that would be unmerged:
--- Couldn't find 'mail-client/-MERGING-evolution' to unmerge.
>>> No packages selected for removal by unmerge


> 
> Although that could fail, as package names aren't supposed to start with
> -, so who knows what portage does with it. Next shot, clear out anything
> you don't need (make sure to check the output, although these days this
> command is safe unless you have done something strange).
> emerge -av --depclean
To see what this might do, I tried emerge --pretend -av --depclean.
The response identified 163 packages to be removed, none of which appear
to be related to mail-client/-MERGING-evolution:2.0.

> 
> Final shot is to play around in /var/db/pkg. This is where portage keeps
> track of what is installed, and how, so messing around here is
> inherently unsafe. I have my suspicions as to the correct thing to do
> here, but I won't post it since I'm not terribly sure. Someone with more
> expertise could help here, but I suspect either method 1 or 2 will work.
In /var/db/pkg I found a directory mail-client/-MERGING-evolution-2.12.1
containing the following files
CATEGORY  COUNTER      FEATURES   LICENSE  SLOT
repository
CBUILD    CXXFLAGS     HOMEPAGE   NEEDED   SRC_URI
CFLAGS    DEPEND       INHERITED  PDEPEND  USE
CHOST     DESCRIPTION  IUSE       PF       environment.bz2
CONTENTS  EAPI         KEYWORDS   RDEPEND  evolution-2.12.1.ebuild

Unsure what to do next, I'd be grateful for suggestions.

-John

> 
> Nick
> 
> 


-- 
John P. Burkett
Department of Economics
University of Rhode Island
Kingston, RI 02881-0808
USA

phone (401) 874-9195

[gentoo-user] revdep-rebuild problem

[Reinhard Kotucha]

Hi,
I'm using Gentoo for a couple of years and am quite amazed how good it
works.  So thanks to all involved in its develpoment.

However, after today's update, when I run revdep-rebuild, I get the
message

 * Checking dynamic linking consistency
 *   broken /usr/lib64/libogrove.la (requires -lstdc++)
 *   broken /usr/lib64/libospgrove.la (requires -lstdc++)
 *   broken /usr/lib64/libostyle.la (requires -lstdc++)

I would expect that revdep-rebuild re-builds the broken libraries, but
instead it only re-builds app-text/openjade.

Hence I can run revdep-rebuild again and again ad nauseam without any
effect.

Another, quite unrelated question:  Whenever I run 

   emerge --update --pretend

I'm warned in advance if a portage update is available.  It seems that
emerge is able to warn about critical updates.

On the other hand it happpened several times that *after* an update
I've been told that my system is completely broken and will not
re-boot unless I compile a new kernel.  It would be nice if I can be
warned *before* I run emerge without the --pretend option.  Then I
could postpone the update to the next weekend, when I have more time.

My propsal is to add a warning similar to that I get when portage
updates are available, so that users know in advance that a particular
update will break the system.  

Of course, this mailing list is not the proper place for such
suggestions.  Can anybody tell me whom I should ask?

Regards,
  Reinhard

-- 
----------------------------------------------------------------------------
Reinhard Kotucha                                      Phone: +49-511-3373112
Marschnerstr. 25
D-30167 Hannover                              mailto:reinhard.kotucha@web.de
----------------------------------------------------------------------------
Microsoft isn't the answer. Microsoft is the question, and the answer is NO.
----------------------------------------------------------------------------

Re: [gentoo-user] revdep-rebuild problem

[Volker Armin Hemmann]

Am Sonntag, 19. August 2012, 00:37:36 schrieb Reinhard Kotucha:
> Hi,
> I'm using Gentoo for a couple of years and am quite amazed how good it
> works.  So thanks to all involved in its develpoment.
> 
> However, after today's update, when I run revdep-rebuild, I get the
> message
> 
>  * Checking dynamic linking consistency
>  *   broken /usr/lib64/libogrove.la (requires -lstdc++)
>  *   broken /usr/lib64/libospgrove.la (requires -lstdc++)
>  *   broken /usr/lib64/libostyle.la (requires -lstdc++)

so, find out which package these three belong to - and remove them.
n
> 
>    emerge --update --pretend

why pretend?

> 
> I'm warned in advance if a portage update is available.  It seems that
> emerge is able to warn about critical updates.

only about portage updates.

> 
> On the other hand it happpened several times that *after* an update
> I've been told that my system is completely broken and will not
> re-boot unless I compile a new kernel. 

really? never saw that. Only with xorg-drivers after a xorg-server update.

> It would be nice if I can be
> warned *before* I run emerge without the --pretend option.  Then I
> could postpone the update to the next weekend, when I have more time.

so you want portage to read every single ebuild, making the operation A LOT 
longer? I am sorry but I am not willing to waste so much time.

> 
> My propsal is to add a warning similar to that I get when portage
> updates are available, so that users know in advance that a particular
> update will break the system.

please enlighten me which update breaks a system. Can't remember one. Hm, back 
with libss&co maybe?

> 
> Of course, this mailing list is not the proper place for such
> suggestions.  Can anybody tell me whom I should ask?

bugzilla. Feature request.

-- 
#163933

Re: [gentoo-user] revdep-rebuild problem

[Alan McKinnon]

A few extra inline comments to reinforce what you just said:


On Sun, 19 Aug 2012 01:02:24 +0200
Volker Armin Hemmann <volkerarmin@googlemail.com> wrote:

> Am Sonntag, 19. August 2012, 00:37:36 schrieb Reinhard Kotucha:
> > Hi,
> > I'm using Gentoo for a couple of years and am quite amazed how good
> > it works.  So thanks to all involved in its develpoment.
> > 
> > However, after today's update, when I run revdep-rebuild, I get the
> > message
> > 
> >  * Checking dynamic linking consistency
> >  *   broken /usr/lib64/libogrove.la (requires -lstdc++)
> >  *   broken /usr/lib64/libospgrove.la (requires -lstdc++)
> >  *   broken /usr/lib64/libostyle.la (requires -lstdc++)
> 
> so, find out which package these three belong to - and remove them.
> n
> > 
> >    emerge --update --pretend
> 
> why pretend?
> 
> > 
> > I'm warned in advance if a portage update is available.  It seems
> > that emerge is able to warn about critical updates.
> 
> only about portage updates.

and it's hard-coded:

if (update available)
  print alarming message

> 
> > 
> > On the other hand it happpened several times that *after* an update
> > I've been told that my system is completely broken and will not
> > re-boot unless I compile a new kernel. 
> 
> really? never saw that. Only with xorg-drivers after a xorg-server
> update.
> 
> > It would be nice if I can be
> > warned *before* I run emerge without the --pretend option.  Then I
> > could postpone the update to the next weekend, when I have more
> > time.
> 
> so you want portage to read every single ebuild, making the operation
> A LOT longer? I am sorry but I am not willing to waste so much time.

Not only that but it's impossible for portage to know this before hand
unless some dev puts the information in the ebuild. And the infra to
check that does not exist.

This is all right correct and proper. The only way to know an update
breaks something is to build and test it. That's how Ubuntu does it:
the dev builds it, installs it, finds it breaks stuff. So it gets
committed to a repo with a warning that *the*dev*already*knows*about*

The Gentoo dev DOES NOT know about it, and cannot either. These
breakages are usually dependant on the environment they are used in and
only the user knows that. As you and I well know, the compiling Gentoo
user is that analogy of the Ubuntu DEV

> 
> > 
> > My propsal is to add a warning similar to that I get when portage
> > updates are available, so that users know in advance that a
> > particular update will break the system.
> 
> please enlighten me which update breaks a system. Can't remember one.
> Hm, back with libss&co maybe?

jpeg-7, expat2 ....

Those were dealt with in the only sane manner possible:

~arch: tough. Keep both pieces.
 arch: news item in advance

> 
> > 
> > Of course, this mailing list is not the proper place for such
> > suggestions.  Can anybody tell me whom I should ask?
> 
> bugzilla. Feature request.
> 



-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] revdep-rebuild problem

[Volker Armin Hemmann]

Am Sonntag, 19. August 2012, 01:20:05 schrieb Alan McKinnon:
> A few extra inline comments to reinforce what you just said:

> > > My propsal is to add a warning similar to that I get when portage
> > > updates are available, so that users know in advance that a
> > > particular update will break the system.
> > 
> > please enlighten me which update breaks a system. Can't remember one.
> > Hm, back with libss&co maybe?
> 
> jpeg-7, expat2 ....

both broke world, not system. 

-- 
#163933

Re: [gentoo-user] revdep-rebuild problem

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 465 bytes --]

On Sun, 19 Aug 2012 03:02:38 +0200, Volker Armin Hemmann wrote:

> > > please enlighten me which update breaks a system. Can't remember
> > > one. Hm, back with libss&co maybe?  
> > 
> > jpeg-7, expat2 ....  
> 
> both broke world, not system. 

expat broke system here, but that is ancient history from which lessons
appear to have been learned.


-- 
Neil Bothwick

If bankers can count, how come they have eight windows and only four
tellers?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] revdep-rebuild problem

[Reinhard Kotucha]

On 2012-08-19 at 01:02:24 +0200, Volker Armin Hemmann wrote:

 > Am Sonntag, 19. August 2012, 00:37:36 schrieb Reinhard Kotucha:
 > > Hi,
 > > I'm using Gentoo for a couple of years and am quite amazed how good it
 > > works.  So thanks to all involved in its develpoment.
 > > 
 > > However, after today's update, when I run revdep-rebuild, I get the
 > > message
 > > 
 > >  * Checking dynamic linking consistency
 > >  *   broken /usr/lib64/libogrove.la (requires -lstdc++)
 > >  *   broken /usr/lib64/libospgrove.la (requires -lstdc++)
 > >  *   broken /usr/lib64/libostyle.la (requires -lstdc++)
 > 
 > so, find out which package these three belong to - and remove them.

Ok, will do.

 > >    emerge --update --pretend
 > 
 > why pretend?

Because whenever I see that there is an Xorg update, I nowadays make a
full backup before I do the actual update.

 > > On the other hand it happpened several times that *after* an update
 > > I've been told that my system is completely broken and will not
 > > re-boot unless I compile a new kernel. 
 > 
 > really? never saw that. Only with xorg-drivers after a xorg-server
 > update.

Presumably because you already were using a newer kernel.  It was the
kernel version number what mattered, not the configuration.

 > > It would be nice if I can be warned *before* I run emerge without
 > > the --pretend option.  Then I could postpone the update to the
 > > next weekend, when I have more time.
 > 
 > so you want portage to read every single ebuild, making the
 > operation A LOT longer? I am sorry but I am not willing to waste so
 > much time.

Maybe a news item would be sufficient.

 > > My propsal is to add a warning similar to that I get when portage
 > > updates are available, so that users know in advance that a
 > > particular update will break the system.
 > 
 > please enlighten me which update breaks a system. Can't remember
 > one. Hm, back with libss&co maybe?

Last time it happened after an udev update.  After the update I've
been told that my kernel was too old.  If I have to build a new kernel
the next day, I can only hope that there is no power failure
meanwhile and would rather postpone the update.

Regards,
  Reinhard

-- 
----------------------------------------------------------------------------
Reinhard Kotucha                                      Phone: +49-511-3373112
Marschnerstr. 25
D-30167 Hannover                              mailto:reinhard.kotucha@web.de
----------------------------------------------------------------------------
Microsoft isn't the answer. Microsoft is the question, and the answer is NO.
----------------------------------------------------------------------------

Re: [gentoo-user] revdep-rebuild problem

[Volker Armin Hemmann]

Am Montag, 20. August 2012, 01:00:34 schrieb Reinhard Kotucha:
> On 2012-08-19 at 01:02:24 +0200, Volker Armin Hemmann wrote:
>  > Am Sonntag, 19. August 2012, 00:37:36 schrieb Reinhard Kotucha:
>  > > Hi,
>  > > I'm using Gentoo for a couple of years and am quite amazed how good it
>  > > works.  So thanks to all involved in its develpoment.
>  > > 
>  > > However, after today's update, when I run revdep-rebuild, I get the
>  > > message
>  > > 
>  > >  * Checking dynamic linking consistency
>  > >  *   broken /usr/lib64/libogrove.la (requires -lstdc++)
>  > >  *   broken /usr/lib64/libospgrove.la (requires -lstdc++)
>  > >  *   broken /usr/lib64/libostyle.la (requires -lstdc++)
>  > 
>  > so, find out which package these three belong to - and remove them.
> 
> Ok, will do.
> 
>  > >    emerge --update --pretend
>  > 
>  > why pretend?
> 
> Because whenever I see that there is an Xorg update, I nowadays make a
> full backup before I do the actual update.
> 

that is what --ask is made for ;)

>  > > On the other hand it happpened several times that *after* an update
>  > > I've been told that my system is completely broken and will not
>  > > re-boot unless I compile a new kernel.
>  > 
>  > really? never saw that. Only with xorg-drivers after a xorg-server
>  > update.
> 
> Presumably because you already were using a newer kernel.  It was the
> kernel version number what mattered, not the configuration.

I have been sticking around with 3.0 for a long time. 

> 
>  > > It would be nice if I can be warned *before* I run emerge without
>  > > the --pretend option.  Then I could postpone the update to the
>  > > next weekend, when I have more time.
>  > 
>  > so you want portage to read every single ebuild, making the
>  > operation A LOT longer? I am sorry but I am not willing to waste so
>  > much time.
> 
> Maybe a news item would be sufficient.
> 
>  > > My propsal is to add a warning similar to that I get when portage
>  > > updates are available, so that users know in advance that a
>  > > particular update will break the system.
>  > 
>  > please enlighten me which update breaks a system. Can't remember
>  > one. Hm, back with libss&co maybe?
> 
> Last time it happened after an udev update.  After the update I've
> been told that my kernel was too old.  If I have to build a new kernel
> the next day, I can only hope that there is no power failure
> meanwhile and would rather postpone the update.

okay, there is a reason I masked udev updates a long time ago ;)

-- 
#163933

Re: [gentoo-user] revdep-rebuild problem

[Bryan Gardiner]

On August 20, 2012 15:27:54 Volker Armin Hemmann wrote:
> Am Montag, 20. August 2012, 01:00:34 schrieb Reinhard Kotucha:
> > On 2012-08-19 at 01:02:24 +0200, Volker Armin Hemmann wrote:
> >  > Am Sonntag, 19. August 2012, 00:37:36 schrieb Reinhard Kotucha:
> >  > >    emerge --update --pretend
> >  > 
> >  > why pretend?
> > 
> > Because whenever I see that there is an Xorg update, I nowadays make a
> > full backup before I do the actual update.
> 
> that is what --ask is made for ;)

Nothing wrong with --pretend.  I use it religiously for a couple of reasons.  
One is that when I want --quiet-build off, I can keep a complete list of the 
packages being merged in another screen tab, in case something fails.

The other is that I don't want "emerge --ask" littering my emerge.log with 
stuff like this while I'm messing with USE flags, keywords, etc. and not 
actually committing to any emerges:

1345531458: Started emerge on: Aug 20, 2012 23:44:18
1345531458:  *** emerge --ask --binpkg-respect-use=y --quiet-build=n --verbose 
--quiet-unmerge-warn extreme-tuxracer
1345531466:  *** exiting unsuccessfully with status '130'.
1345531466:  *** terminating.

Getting back to the actual topic :) I've noticed that more and more packages 
are getting pkg_pretend() checks so that ebuilds can abort before the build 
starts if something is missing, e.g. missing kernel modules or in the case of 
Xorg, gcc being too old.  This might be what you're asking for, in a way.

Cheers,
Bryan