Portage Output / End User Experience
 

On Mon, Jul 9, 2012 at 10:11 AM, Peter Stuge <peter@stuge.se> wrote:
> Rich Freeman wrote:
>> It also seems like the current portage output is giving the user some
>> contradictory and counterproductive advice. It seems like there are
>> really only two possible choices
>> 1. The user could choose to not install chromium.
>> 2. The user could enable icu for qt-webkit.
>
> Do you get the same suggestions from portage if you attempt emerge
> kde and chromium one at a time?

I'll test it out on a fresh install, but that will take a number of
hours (a total of 694 packages between them, which will take a while
even on 4 cores, and warm up the office as well...). If I try to
start out with chromium first the initial suggestion is to enable icu
on libxml2, and if I try to start out with kde-meta the initial
suggestion is to enable minizip on zlib (unrelated to this whole
issue).

I don't have a copy of the message but when I got the update to
qt-webkit the message was fairly cryptic when it added the !icu?
dependency on libxml2. If anything I think it was less clear.

Let me give that upgrade a try both ways, but it might take a few days
to have all the results of that.

Rich

Portage Output / End User Experience
 

On Mon, Jul 9, 2012 at 10:56 AM, Rich Freeman <rich0@gentoo.org> wrote:
> On Mon, Jul 9, 2012 at 10:11 AM, Peter Stuge <peter@stuge.se> wrote:
>> Rich Freeman wrote:
>>> It also seems like the current portage output is giving the user some
>>> contradictory and counterproductive advice. It seems like there are
>>> really only two possible choices
>>> 1. The user could choose to not install chromium.
>>> 2. The user could enable icu for qt-webkit.
>>
>> Do you get the same suggestions from portage if you attempt emerge
>> kde and chromium one at a time?
>
> I'll test it out on a fresh install, but that will take a number of
> hours (a total of 694 packages between them, which will take a while
> even on 4 cores, and warm up the office as well...). If I try to
> start out with chromium first the initial suggestion is to enable icu
> on libxml2, and if I try to start out with kde-meta the initial
> suggestion is to enable minizip on zlib (unrelated to this whole
> issue).
>
> I don't have a copy of the message but when I got the update to
> qt-webkit the message was fairly cryptic when it added the !icu?
> dependency on libxml2. If anything I think it was less clear.
>
> Let me give that upgrade a try both ways, but it might take a few days
> to have all the results of that.

When tracking down why glibc updates were killing my systems, I wrote
an install script to accelerate the install-update-fail-retry cycle.
You may find it useful:

https://github.com/mikemol/gentoo-install

It shouldn't require *too* much modification to automate what you're
trying to test. I intend to modify it to work in chroot environments,
as a prelude to some build-related bug reports I'm sitting on.

--
:wq

Portage Output / End User Experience
 

On Mon, Jul 9, 2012 at 11:04 AM, Michael Mol <mikemol@gmail.com> wrote:
> It shouldn't require *too* much modification to automate what you're
> trying to test. I intend to modify it to work in chroot environments,
> as a prelude to some build-related bug reports I'm sitting on.

Thanks - seems useful in general. Not sure I'll use it for this - a
chroot should be adequate to test this (I don't intend to actually run
KDE), and is much more efficient with CPU/RAM/etc.

One thing I did set up for doing VM installs of Gentoo is a PXE
enviornment that runs with an NFS root. I can stick stuff like
stage3s and portage snapshots and template config files on that NFS
root easily, and update it when necessary (though really the install
image really only needs to be updated to address hardware, which isn't
a big deal with VMs). It isn't anything too exciting, but maybe I
should write up a howto - there may already be some Gentoo PXE guides
already and any of those should basically work.

Rich

Portage Output / End User Experience
 

On Mon, Jul 9, 2012 at 6:41 PM, Rich Freeman <rich0@gentoo.org> wrote:
> Thanks - seems useful in general. Not sure I'll use it for this - a
> chroot should be adequate to test this (I don't intend to actually run
> KDE), and is much more efficient with CPU/RAM/etc.

Liberté Linux build scripts do a complete build-from-scratch in a chroot:

https://github.com/mkdesu/liberte/blob/master/mkroot
https://github.com/mkdesu/liberte/blob/master/src/root/setup

"build" is the top-level script, mastering a read-only deployment
image, but mkroot+setup suit the purpose of testing / deploying a
fresh Gentoo build.

--
Maxim Kammerer
Liberté Linux: http://dee.su/liberte

Portage Output / End User Experience
 

On 07/09/2012 07:02 AM, Rich Freeman wrote:
> Is there any way for portage to figure out that one of those is the
> eventual outcome, and then direct the user to the minimal changes to
> accomplish either?

You could try a larger emerge --backtrack setting. If it doesn't settle
on a sensible solution, then it would be interesting to look at --debug
output, to find out what has gone wrong.
--
Thanks,
Zac

Portage Output / End User Experience
 

On Mon, Jul 9, 2012 at 10:56 AM, Rich Freeman <rich0@gentoo.org> wrote:
> I'll test it out on a fresh install, but that will take a number of
> hours

If I install chromium first, I get the following messages when I try
to install kde-meta:
The following USE changes are necessary to proceed:
#required by dev-db/virtuoso-server-6.1.4-r1, required by
dev-libs/soprano-2.7.6[virtuoso], required by
app-office/akonadi-server-1.7.2,
required by kde-base/kdepim-runtime-4.8.3-r2, required by
kde-base/kdepim-meta-4.8.3, required by
kde-base/kde-meta-4.8.3[semantic-desktop], required by kde-meta (argument)
=sys-libs/zlib-1.2.5.1-r2 minizip
#required by x11-libs/qt-webkit-4.8.2[gstreamer], required by
kde-base/kdebase-menu-icons-4.8.3, required by
kde-base/kdebase-runtime-meta-4.8.3, required by
kde-base/kdebase-startkde-4.8.3, required by kde-base/kdebase-meta-4.8.3,
required by kde-base/kde-meta-4.8.3, required by kde-meta (argument)
=dev-libs/libxml2-2.8.0_rc1 -icu

You'll note that in this case there is nothing to suggest simply
enabling icu for qt-webkit.

If I emerge kde-meta first then I get the following when I try to
emerge chromium:
The following USE changes are necessary to proceed:
#required by www-client/chromium-20.0.1132.43, required by
chromium (argument)
=dev-libs/libxml2-2.8.0_rc1 icu

Then if I set the icu use flag on libxml2 it works. Apparently it
doesn't realize that I'm about to break qt-webkit. Portage doesn't
check use dependencies on existing packages when you go to rebuild
something?

I hope that helps...

Rich

Portage Output / End User Experience
 

On 07/09/2012 05:42 PM, Rich Freeman wrote:
> On Mon, Jul 9, 2012 at 10:56 AM, Rich Freeman <rich0@gentoo.org> wrote:
>> I'll test it out on a fresh install, but that will take a number of
>> hours
>
> If I install chromium first, I get the following messages when I try
> to install kde-meta:
> The following USE changes are necessary to proceed:
> #required by dev-db/virtuoso-server-6.1.4-r1, required by
> dev-libs/soprano-2.7.6[virtuoso], required by
> app-office/akonadi-server-1.7.2,
> required by kde-base/kdepim-runtime-4.8.3-r2, required by
> kde-base/kdepim-meta-4.8.3, required by
> kde-base/kde-meta-4.8.3[semantic-desktop], required by kde-meta (argument)
> =sys-libs/zlib-1.2.5.1-r2 minizip
> #required by x11-libs/qt-webkit-4.8.2[gstreamer], required by
> kde-base/kdebase-menu-icons-4.8.3, required by
> kde-base/kdebase-runtime-meta-4.8.3, required by
> kde-base/kdebase-startkde-4.8.3, required by kde-base/kdebase-meta-4.8.3,
> required by kde-base/kde-meta-4.8.3, required by kde-meta (argument)
> =dev-libs/libxml2-2.8.0_rc1 -icu
>
> You'll note that in this case there is nothing to suggest simply
> enabling icu for qt-webkit.
>
> If I emerge kde-meta first then I get the following when I try to
> emerge chromium:
> The following USE changes are necessary to proceed:
> #required by www-client/chromium-20.0.1132.43, required by
> chromium (argument)
> =dev-libs/libxml2-2.8.0_rc1 icu
>
> Then if I set the icu use flag on libxml2 it works. Apparently it
> doesn't realize that I'm about to break qt-webkit. Portage doesn't
> check use dependencies on existing packages when you go to rebuild
> something?

Not unless the --complete-graph option is enabled. What I'd like to do
is to automatically enable --complete-graph mode whenever the USE of an
installed package would change. It would be like that
--complete-graph-if-new-ver option which is already enabled by default,
but it would apply to USE instead of versions.
--
Thanks,
Zac

Portage Output / End User Experience
 

On 07/09/2012 05:58 PM, Zac Medico wrote:
> On 07/09/2012 05:42 PM, Rich Freeman wrote:
>> On Mon, Jul 9, 2012 at 10:56 AM, Rich Freeman <rich0@gentoo.org> wrote:
>>> I'll test it out on a fresh install, but that will take a number of
>>> hours
>>
>> If I install chromium first, I get the following messages when I try
>> to install kde-meta:
>> The following USE changes are necessary to proceed:
>> #required by dev-db/virtuoso-server-6.1.4-r1, required by
>> dev-libs/soprano-2.7.6[virtuoso], required by
>> app-office/akonadi-server-1.7.2,
>> required by kde-base/kdepim-runtime-4.8.3-r2, required by
>> kde-base/kdepim-meta-4.8.3, required by
>> kde-base/kde-meta-4.8.3[semantic-desktop], required by kde-meta (argument)
>> =sys-libs/zlib-1.2.5.1-r2 minizip
>> #required by x11-libs/qt-webkit-4.8.2[gstreamer], required by
>> kde-base/kdebase-menu-icons-4.8.3, required by
>> kde-base/kdebase-runtime-meta-4.8.3, required by
>> kde-base/kdebase-startkde-4.8.3, required by kde-base/kdebase-meta-4.8.3,
>> required by kde-base/kde-meta-4.8.3, required by kde-meta (argument)
>> =dev-libs/libxml2-2.8.0_rc1 -icu
>>
>> You'll note that in this case there is nothing to suggest simply
>> enabling icu for qt-webkit.
>>
>> If I emerge kde-meta first then I get the following when I try to
>> emerge chromium:
>> The following USE changes are necessary to proceed:
>> #required by www-client/chromium-20.0.1132.43, required by
>> chromium (argument)
>> =dev-libs/libxml2-2.8.0_rc1 icu
>>
>> Then if I set the icu use flag on libxml2 it works. Apparently it
>> doesn't realize that I'm about to break qt-webkit. Portage doesn't
>> check use dependencies on existing packages when you go to rebuild
>> something?
>
> Not unless the --complete-graph option is enabled. What I'd like to do
> is to automatically enable --complete-graph mode whenever the USE of an
> installed package would change. It would be like that
> --complete-graph-if-new-ver option which is already enabled by default,
> but it would apply to USE instead of versions.

Bug filed:

https://bugs.gentoo.org/show_bug.cgi?id=425558

--
Thanks,
Zac

Portage Output / End User Experience
 

On Mon, Jul 9, 2012 at 8:58 PM, Zac Medico <zmedico@gentoo.org> wrote:
>
> Not unless the --complete-graph option is enabled. What I'd like to do
> is to automatically enable --complete-graph mode whenever the USE of an
> installed package would change. It would be like that
> --complete-graph-if-new-ver option which is already enabled by default,
> but it would apply to USE instead of versions.

Ok, with this additional option, if I start with chromium and try to
install kde-meta it still suggests unsetting icu, which is going to
lead to future errors (or breaking chromium).

If I start with kde-meta and try to install chromium the conflict is
detected, but it does not suggest setting icu on qt-webkit:
!!! Multiple package instances within a single package slot have been pulled
!!! into the dependency graph, resulting in a slot conflict:

dev-libs/libxml2:2

(dev-libs/libxml2-2.8.0_rc1::gentoo, installed) pulled in by
dev-libs/libxml2:2[!icu?] required by (x11-libs/qt-webkit-4.8.2::gentoo,
installed)

(dev-libs/libxml2-2.8.0_rc1::gentoo, ebuild scheduled for merge)
pulled in by
dev-libs/libxml2[icu] required by
(www-client/chromium-20.0.1132.43::gentoo,
ebuild scheduled for merge)

So, seems like there is still room for improvement...

Rich

Portage Output / End User Experience
 

On 07/09/2012 06:11 PM, Rich Freeman wrote:
> On Mon, Jul 9, 2012 at 8:58 PM, Zac Medico <zmedico@gentoo.org> wrote:
>>
>> Not unless the --complete-graph option is enabled. What I'd like to do
>> is to automatically enable --complete-graph mode whenever the USE of an
>> installed package would change. It would be like that
>> --complete-graph-if-new-ver option which is already enabled by default,
>> but it would apply to USE instead of versions.
>
> Ok, with this additional option, if I start with chromium and try to
> install kde-meta it still suggests unsetting icu, which is going to
> lead to future errors (or breaking chromium).
>
> If I start with kde-meta and try to install chromium the conflict is
> detected, but it does not suggest setting icu on qt-webkit:
> !!! Multiple package instances within a single package slot have been pulled
> !!! into the dependency graph, resulting in a slot conflict:
>
> dev-libs/libxml2:2
>
> (dev-libs/libxml2-2.8.0_rc1::gentoo, installed) pulled in by
> dev-libs/libxml2:2[!icu?] required by (x11-libs/qt-webkit-4.8.2::gentoo,
> installed)
>
> (dev-libs/libxml2-2.8.0_rc1::gentoo, ebuild scheduled for merge)
> pulled in by
> dev-libs/libxml2[icu] required by
> (www-client/chromium-20.0.1132.43::gentoo,
> ebuild scheduled for merge)
>
> So, seems like there is still room for improvement...

Aside from the obvious need to improve the portage behavior, we might
also want to consider enabling USE=icu by default in the profile.
--
Thanks,
Zac