Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

[gentoo-dev] GTK und GTK2 Use Flag

78 views
Skip to first unread message

Fabian Zeindl

unread,
Mar 29, 2005, 4:20:08 AM3/29/05
to
Hello

There was a discussion on the gentoo-user-de list about this two
Useflags: gtk and gtk2. Because not everybody is sure what the mean, so
if you have -gtk +gtk2 some think that gtk2 should be installed and soon.

Wouldn't it be better if "gtk" meaned that the newest available gtk
version ist installed (gtk1 or gtk2) and a flag like oldgtk take the
older version gtk1.

Another question which occured: Is there a performancedrawback if a
program is compile with gtk1 AND gtk2 build in? Does this happen when
someone installs with +gtk +gtk2?


fabian


--
Musik ist niemals illegal: www.fairsharing.de

I prefer signed/encrypted Mail:
Fingerprint: CFE8 38A7 0BC4 3CB0 E454 FA8D 04F9 B3B6 E02D 25BA

signature.asc

Stefan Sperling

unread,
Mar 29, 2005, 6:10:10 AM3/29/05
to
On Tue, Mar 29, 2005 at 12:58:57PM +0200, Stefan Sperling wrote:
> There are lots of useflags that don't install anything.
> Take the new mp3 useflag for example, used by beep-media-player.
> would you expect it to install a program called 'mp3'? :)

What was I talking about?
Of course gtk1 and gtk2 get installed:

% qpkg -I -v gtk+
x11-libs/gtk+-2.6.4 *
x11-libs/gtk+-1.2.10-r11 *

I stand corrected.
--
stefan PGP Key: 0xF59D25F0
--
gento...@gentoo.org mailing list

Diego "Flameeyes" Pettenò

unread,
Mar 29, 2005, 6:10:10 AM3/29/05
to
On Tuesday 29 March 2005 12:58, Stefan Sperling wrote:
> Some programs use gtk1 only, some gtk2 only, and a few have
> support for both. As there is not much point in using both
> of them at runtime, if you use +gtk2, programs that support
> both will use gtk2.
Uhm AFAIK, you *need* to have both gtk and gtk2 to support gtk2 in apps which
has the three-way choice: nogtk, gtk, gtk2, as -gtk +gtk2 is a no-op.

That's probably the problem that Fabian was searching a solution for.

--
Diego "Flameeyes" Pettenò
http://wwwstud.dsi.unive.it/~dpetteno/

Stefan Sperling

unread,
Mar 29, 2005, 6:10:14 AM3/29/05
to
On Tue, Mar 29, 2005 at 12:15:17PM +0200, Fabian Zeindl wrote:
> Hello
>
> There was a discussion on the gentoo-user-de list about this two
> Useflags: gtk and gtk2. Because not everybody is sure what the mean, so
> if you have -gtk +gtk2 some think that gtk2 should be installed and soon.

There are lots of useflags that don't install anything.


Take the new mp3 useflag for example, used by beep-media-player.
would you expect it to install a program called 'mp3'? :)

> Wouldn't it be better if "gtk" meaned that the newest available gtk
> version ist installed (gtk1 or gtk2) and a flag like oldgtk take the
> older version gtk1.

As soon as gtk3 comes along, you scheme won't work anymore.

Some programs use gtk1 only, some gtk2 only, and a few have
support for both. As there is not much point in using both
of them at runtime, if you use +gtk2, programs that support
both will use gtk2.

I don't know what happens when you have both +gtk1 and +gtk2 set,
but I reckon that still evaluates to using gtk2. You can simply try
this out for yourself. The ldd command will tell you which libraries
a given binary is linked with (see the ldd manpage).

> Another question which occured: Is there a performancedrawback if a
> program is compile with gtk1 AND gtk2 build in? Does this happen when
> someone installs with +gtk +gtk2?

There is no performance penalty.
You can savely install both versions. gtk1 and gtk2 are not 'build' into
a program, they are shared libraries. What happens is that when a program
uses gtk{1,2}, a copy of the gtk{1,2} code is loaded from disk into memory
if it is not already in memory. So all programs compiled against the same
version of gtk use the very same copy of gtk code at runtime, which of
course saves a lot of memory.

> --
> Musik ist niemals illegal: www.fairsharing.de

My music is not illegal either (creative commons) :)
http://binarchy.net/werkstatt

Stefan Sperling

unread,
Mar 29, 2005, 6:20:13 AM3/29/05
to
On Tue, Mar 29, 2005 at 01:04:13PM +0200, Diego Flameeyes Pettenň wrote:
> -gtk +gtk2 is a no-op.

why?

Brian Harring

unread,
Mar 29, 2005, 6:20:11 AM3/29/05
to
On Tue, Mar 29, 2005 at 12:58:57PM +0200, Stefan Sperling wrote:
> On Tue, Mar 29, 2005 at 12:15:17PM +0200, Fabian Zeindl wrote:
> > Hello
> >
> > There was a discussion on the gentoo-user-de list about this two
> > Useflags: gtk and gtk2. Because not everybody is sure what the mean, so
> > if you have -gtk +gtk2 some think that gtk2 should be installed and soon.
>
> There are lots of useflags that don't install anything.
> Take the new mp3 useflag for example, used by beep-media-player.
> would you expect it to install a program called 'mp3'? :)
>
> > Wouldn't it be better if "gtk" meaned that the newest available gtk
> > version ist installed (gtk1 or gtk2) and a flag like oldgtk take the
> > older version gtk1.
>
> As soon as gtk3 comes along, you scheme won't work anymore.
>
> Some programs use gtk1 only, some gtk2 only, and a few have
> support for both. As there is not much point in using both
> of them at runtime, if you use +gtk2, programs that support
> both will use gtk2.
A saner (and previously proposed) approach would be
gtk ~= gtk support. *any* version.
gtk1 = gtk v1 support
gtk2 = gtk v2 support
gtkN = gtk vN support.

With the current approach, gtk2 is dependant on gtk; above approach
breaks the dependency between the two flags, and gives the user the
ability to control explicitly what versions are used, rather then the
partial crapshoot it is now. If they want *just* v2 support, above
allows it w/out resorting to a package.mask.
~brian
--
gento...@gentoo.org mailing list

Diego "Flameeyes" Pettenò

unread,
Mar 29, 2005, 6:30:15 AM3/29/05
to
On Tuesday 29 March 2005 13:11, Stefan Sperling wrote:

> On Tue, Mar 29, 2005 at 01:04:13PM +0200, Diego Flameeyes Pettenò wrote:
> > -gtk +gtk2 is a no-op.
> why?
As Brian said, gtk2 depends on gtk.
Just as a practical example, take ethereal, which could be built without gui
support, just using tethereal.
If you build it with -gtk +gtk2, it will build tethereal, not ethereal. I
submitted some time ago a patch to make this a more logical behaviour, but I
needed to change it to suit the same behaviour of other packages. See bug
#81055.

Take also amule which you can find on bugzilla as an example, which can be
built without gtk support, and on which gtk2 support depends on gtk.

Or wxGTK in which +wxnogtk flag is used to disable gtk1 support (there's gtk2
flag but no gtk flag).

I still think this is illogical, but I can't do much on this.

David Morgan

unread,
Mar 29, 2005, 6:40:10 AM3/29/05
to
On 13:24 Tue 29 Mar , Diego Flameeyes Petten? wrote:
> On Tuesday 29 March 2005 13:11, Stefan Sperling wrote:
> > On Tue, Mar 29, 2005 at 01:04:13PM +0200, Diego Flameeyes Petten? wrote:
> > > -gtk +gtk2 is a no-op.
> > why?
> As Brian said, gtk2 depends on gtk.
> Just as a practical example, take ethereal, which could be built without gui
> support, just using tethereal.
> If you build it with -gtk +gtk2, it will build tethereal, not ethereal. I
> submitted some time ago a patch to make this a more logical behaviour, but I
> needed to change it to suit the same behaviour of other packages. See bug
> #81055.
>
> Take also amule which you can find on bugzilla as an example, which can be
> built without gtk support, and on which gtk2 support depends on gtk.
>
> Or wxGTK in which +wxnogtk flag is used to disable gtk1 support (there's gtk2
> flag but no gtk flag).
>
> I still think this is illogical, but I can't do much on this.
>

What if I want things to be built without gtk where ever possible, but I
want things that have to use either gtk1 or gtk2 to use gtk2?

Under the current method this is what -gtk gtk2 should do, since
something that has to use gtk won't have a gtk flag, but if there's a
choice between gtk1 and gtk2, it should have a gtk2 flag.

I don't see how this would be possible under ferringb's scheme, so I
prefer the current way (though if ferringb's way of doing it could be
modified to do this without getting too ugly I'd be just as happy with
that).

--
djm

--
gento...@gentoo.org mailing list

Brian Harring

unread,
Mar 29, 2005, 7:00:13 AM3/29/05
to
On Tue, Mar 29, 2005 at 12:35:38PM +0100, David Morgan wrote:
> On 13:24 Tue 29 Mar , Diego Flameeyes Petten? wrote:
> > On Tuesday 29 March 2005 13:11, Stefan Sperling wrote:
> > > On Tue, Mar 29, 2005 at 01:04:13PM +0200, Diego Flameeyes Petten? wrote:
> > > > -gtk +gtk2 is a no-op.
> > > why?
> > As Brian said, gtk2 depends on gtk.
> > Just as a practical example, take ethereal, which could be built without gui
> > support, just using tethereal.
> > If you build it with -gtk +gtk2, it will build tethereal, not ethereal. I
> > submitted some time ago a patch to make this a more logical behaviour, but I
> > needed to change it to suit the same behaviour of other packages. See bug
> > #81055.
> >
> > Take also amule which you can find on bugzilla as an example, which can be
> > built without gtk support, and on which gtk2 support depends on gtk.
> >
> > Or wxGTK in which +wxnogtk flag is used to disable gtk1 support (there's gtk2
> > flag but no gtk flag).
> >
> > I still think this is illogical, but I can't do much on this.
> >
>
> What if I want things to be built without gtk where ever possible, but I
> want things that have to use either gtk1 or gtk2 to use gtk2?
>
> Under the current method this is what -gtk gtk2 should do,
Actually, no. And that is why the gtk flags need changing. :)
current setup.
gtk ~= gtk support AND fall back to gtk v1 support.
gtk2 ~= use gtk v2 support in preference to v1 if available.

There is no way to state, "I want v2, and _only_ v2" without resorting
to a package.mask'ing of gtk-1*.

> since
> something that has to use gtk won't have a gtk flag, but if there's a
> choice between gtk1 and gtk2, it should have a gtk2 flag.
>
> I don't see how this would be possible under ferringb's scheme,

my scheme- gtk == *any* version, the user doesn't care.
gtk{1,2,3} == use that version.

So... if you had USE="gtk2 and gtk3", ebuilds would use v3 if
possible, otherwise v2. No gtk v1.

If you had USE="gtk gtk2", you prefer gtk v2., but will use what is
available (any version).

> so I
> prefer the current way (though if ferringb's way of doing it could be
> modified to do this without getting too ugly I'd be just as happy with
> that).

The current scheme sucks for those who want a gtk v2 only system. How
do you specify gtk v2 only under the current system using _just_ use
flags? You don't. :)

David Morgan

unread,
Mar 29, 2005, 8:20:09 AM3/29/05
to

How do you specify gtk v2 only using use flags? You can't!

If something depends on gtk v1 regardless of use flags (by which I mean
there's no without any gtk or use v2 option), then if you want
to emerge it you have to have gtk v1. Unless you want to have some sort
of system where a package would be masked because of your use flags
(which would get annoying really quickly), but I don't think you are.

That being said, I agree that your way is better, assuming that -gtk
-gtk1 gtk2 means "disable optional gtk (any version) support, but if gtk
isn't optional but both versions are supported then use v2"

Brian Harring

unread,
Mar 29, 2005, 8:50:11 AM3/29/05
to
On Tue, Mar 29, 2005 at 02:09:51PM +0100, David Morgan wrote:
> How do you specify gtk v2 only using use flags? You can't!
>
> If something depends on gtk v1 regardless of use flags (by which I mean
> there's no without any gtk or use v2 option), then if you want
> to emerge it you have to have gtk v1. Unless you want to have some sort
> of system where a package would be masked because of your use flags
> (which would get annoying really quickly), but I don't think you are.
That's semantic quibbling. :)
Use flags are conditionals only; hard deps can't be sidestepped. :)
Finer grained control over conditional linkage is the goal, which the
current USE="gtk gtk2" doesn't engender.

> assuming that -gtk
> -gtk1 gtk2 means "disable optional gtk (any version) support, but if gtk
> isn't optional but both versions are supported then use v2"

E'yep.


>
> --
> djm
>
> --
> gento...@gentoo.org mailing list
>

--
gento...@gentoo.org mailing list

David Morgan

unread,
Mar 29, 2005, 9:30:18 AM3/29/05
to

Which (unless anyone has any objections) leaves the question of how is this
going to get done.

use.decs would need to be changed, which is easy enough, but then all
the ebuilds affected by this would need modifying.

So, either package maintainers would have to do this (which does seem
like something that's likely to happen) or 1 person/a few people
would have to change all the relevant ebuilds, which would require a lot
of work (and requires whoever's doing it to work out which options are
applicable to each package).

I think the second way is the only way that this'd actually get done,
but it requires one or more volunteers. (I'd happily do it, but it's not
really possible without being a dev).

Stefan Sperling

unread,
Mar 29, 2005, 10:40:11 AM3/29/05
to
On Tue, Mar 29, 2005 at 03:19:48PM +0100, David Morgan wrote:
> So, either package maintainers would have to do this (which does seem
> like something that's likely to happen) or 1 person/a few people
> would have to change all the relevant ebuilds, which would require a lot
> of work (and requires whoever's doing it to work out which options are
> applicable to each package).
>
> I think the second way is the only way that this'd actually get done,
> but it requires one or more volunteers. (I'd happily do it, but it's not
> really possible without being a dev).

I think the first way is definitely possible, but would probably
take much longer than the second. Anyway, I'd be willing to help out in
case people decide on way number two. But I'm not a dev either :)

Markus Nigbur

unread,
Mar 29, 2005, 12:30:20 PM3/29/05
to

The past has shown that the first way isn't a solution that should be
considered with the current information flow within the Gentoo
development team.
The second way is indeed hard to accomplish, but should be possible to
do within some days.
I would volunteer for the desktop-misc herded packages - which make up
quite a lot of the affected packages.

--
(o_ Markus Nigbur
//\ Gentoo GNU/Linux Developer
[ ]/_ http://www.gentoo.org
GPG Fingerprint: 9E31 187A 6C78 210C F248 EDBB F2FB A77F 707D F8C1

--
gento...@gentoo.org mailing list

Spider

unread,
Mar 29, 2005, 3:50:13 PM3/29/05
to
On Tue, 2005-03-29 at 12:15 +0200, Fabian Zeindl wrote:
> Hello
>
> There was a discussion on the gentoo-user-de list about this two
> Useflags: gtk and gtk2. Because not everybody is sure what the mean, so
> if you have -gtk +gtk2 some think that gtk2 should be installed and soon.
>
> Wouldn't it be better if "gtk" meaned that the newest available gtk
> version ist installed (gtk1 or gtk2) and a flag like oldgtk take the
> older version gtk1.
>
> Another question which occured: Is there a performancedrawback if a
> program is compile with gtk1 AND gtk2 build in? Does this happen when
> someone installs with +gtk +gtk2?


Common misconception, it tends to crop up once every 6 months or so
since I ever was foolish enough not to stomp a hard policy and beat the
people who want gtk+-1.2 support with a sledgehammer.


Okay. Here's the deal and the logic:

USE="gtk"

I want gtk support. Legacy. classic. Doesn't matter if the version
is gtk+-1.2 or 2.0. ( When it was introduced, there was only gtk+-1.2,
and that was it)

Then came a few, troublesome beasts and started to make our days
annoying. Mozilla was one of them, gFtp another. Enough of them
cropped up that used -both- gtk+-1.2 and gtk+-2 interfaces _AS A
TRANSITION PERIOD_

The gtk+-1.2 interfaces were planned to be deprecated when the new and
shiney(!) interfaces were usable, and have so been done.


The gtk2 USE flag came up for -only- theese packages. Nothing else.

Its a mistake, abusing the redundancy that boolean nested flags work
with, however, I was young, naive, and expected people to be able to
comprehend written instructions and nested logic blocks.

DEPEND="gtk? (
gtk2? ( >=x11-libs/gtk+-2.0 )
!gtk2? ( =x11-libs/gtk+-1.2* ))
!gtk? ( sys-libs/ncurses )"

is obvious and fairly logical in my mind, and is the code that describes
what the gtk2 USE flag is for.

Along with the following descriptions:

gtk - Adds support for x11-libs/gtk+ (The GIMP Toolkit)
gtk2 - Use gtk+-2.0.0 over gtk+-1.2 in cases where a program supports
both.


If someone can suggest a better formulation for this, please go ahead.


Lets face it, gtk+-1.2 is _deprecated_ and all packages that use it
should either be taken out and shot in the head, ( for mercy's sake,
that and utf8 ) or updated to track the new and maintained library
API.

We all know this isn't feasible, most projects are just recently wiping
stale code for interfaces ( Hi mozilla! ) and as such, the mistake is
left to stay.

-Changing- existing logic is -not- going to solve this solution
cleanly. We suggested to remove the gtk2 flag, (make it default) and
implement :
USE="shoot-me-in-the-head-with-deprecated-libraries-please" instead,
however people considered it as offending our users for some reason.

As for the discussion of USE="-gtk +gtk2" :

there is one case where this could be a acceptable and will produce
expected results. That is when there is only a gtk+ interface, no
alternative ( ergo, the -gtk is ignored ) and there is a choice between
gtk+-1.2 or 2.0 (the gtk2 selection flag) However, this is such a
corner case only because the "gtk" flag is completely ignored.


//Spider

--
begin .signature
Tortured users / Laughing in pain
See Microsoft KB Article Q265230 for more information.
end

signature.asc

foser

unread,
Mar 29, 2005, 4:40:13 PM3/29/05
to
On Tue, 2005-03-29 at 19:20 +0200, Markus Nigbur wrote:
<snip much gtk2 use flag blah and old ideas>

The gtk2 flag is meant to be removed in time, not to be promoted even
more -than it was ever meant to be- as it is.

Do it right this time this time, don't try to redo it in as much a
crappy way as it is now. The proposals here are as much time-context
dependant as the current solution is.

- foser

signature.asc

Ciaran McCreesh

unread,
Mar 29, 2005, 4:50:12 PM3/29/05
to
On Tue, 29 Mar 2005 23:35:50 +0200 foser <fo...@gentoo.org> wrote:
| Do it right this time this time, don't try to redo it in as much a
| crappy way as it is now. The proposals here are as much time-context
| dependant as the current solution is.

I think it's been rather well illustrated that GTK+-1 support is not
time dependent, and that GTK+-1 will be around for a very long time...

--
Ciaran McCreesh : Gentoo Developer (Vim, Fluxbox, shell tools)
Mail : ciaranm at gentoo.org
Web : http://dev.gentoo.org/~ciaranm

Brian Harring

unread,
Mar 29, 2005, 5:40:13 PM3/29/05
to
*cough* quicky question then...

What are you going to do when gtk v3 hits? Deprecate gtk v2? Expect
upstream to quickly migrate all projects/code to v3? Cause projects
sure moved off of v1 in a hurry :P

Continuing, when v3 hits, abuse this same interdependent use flag
trickery interdependent, or move to sane versioned use flags?
Personally at this rate I'm expecting xmms to be gtk1 dependant for as
long as gtk-1* compiles, probably right up through when gtk v3 some
day hits :)

If you use the approach I've laid out (yes, not new, I laid it out in
24439) you wouldn't have to dick around with deprecating a version,
nor essentially mandating what version is default. You'd leave the
total control over what versions the user wants to deal with in the
hands of the _users_, and what versions the package supports would
be represented properly/clearly in the IUSE.

So far... I've not really heard a good reason aside from "it's in
place, we'll just deprecate gtk v1 instead of clean it up" for why
this cannot be corrected _now_, or really in the past.

hell, you even have volunteers. :)
~brian


Brian Harring

unread,
Mar 29, 2005, 5:50:10 PM3/29/05
to
On Tue, Mar 29, 2005 at 10:49:00PM +0200, Spider wrote:
> On Tue, 2005-03-29 at 12:15 +0200, Fabian Zeindl wrote:
> > Hello
> >
> > There was a discussion on the gentoo-user-de list about this two
> > Useflags: gtk and gtk2. Because not everybody is sure what the mean, so
> > if you have -gtk +gtk2 some think that gtk2 should be installed and soon.
> >
> > Wouldn't it be better if "gtk" meaned that the newest available gtk
> > version ist installed (gtk1 or gtk2) and a flag like oldgtk take the
> > older version gtk1.
> >
> > Another question which occured: Is there a performancedrawback if a
> > program is compile with gtk1 AND gtk2 build in? Does this happen when
> > someone installs with +gtk +gtk2?
>
>
> Common misconception, it tends to crop up once every 6 months or so
> since I ever was foolish enough not to stomp a hard policy and beat the
> people who want gtk+-1.2 support with a sledgehammer.
while I'm certainly not advocating that the user is always
right... user requirements/desires for the tree should be reflected,
not have mandates handed down. (realize you didn't quite mean it that
way, but it plays into my point so I took a shot at it :)

> If someone can suggest a better formulation for this, please go ahead.

Start the process of adjusting ebuilds so that the use flags reflect
what everybody thinks they should.

I've seen countless users (despite use.desc) flip on *just* gtk2,
because they want *just* gtk v2 support linked in if available.

> Lets face it, gtk+-1.2 is _deprecated_ and all packages that use it
> should either be taken out and shot in the head, ( for mercy's sake,
> that and utf8 ) or updated to track the new and maintained library
> API.
> We all know this isn't feasible, most projects are just recently wiping
> stale code for interfaces ( Hi mozilla! ) and as such, the mistake is
> left to stay.
>
> -Changing- existing logic is -not- going to solve this solution
> cleanly. We suggested to remove the gtk2 flag, (make it default) and
> implement :
> USE="shoot-me-in-the-head-with-deprecated-libraries-please" instead,
> however people considered it as offending our users for some reason.

How is this any different (aside from naming) gtk1 vs gtk2?

Only actual difference I see is sticking 'deprecated' or something equivalent
into the use flag name, and mandating the default gtk+- version that
is used, rather then just exposing the options via use flags, and
letting users decide for themselves.

Aside from that, as I stated in another email, such an approach is
going to bite you in the ass on the next major gtk release,
wash rinse repeat (lesson learned being?)
~brian

Ciaran McCreesh

unread,
Mar 29, 2005, 5:50:05 PM3/29/05
to
On Tue, 29 Mar 2005 16:29:36 -0600 Brian Harring <ferr...@gentoo.org>
wrote:

| What are you going to do when gtk v3 hits?

Will v3 be compatible with v2? Or do you think they haven't learned that
lesson?

Olivier Crête

unread,
Mar 29, 2005, 6:20:07 PM3/29/05
to
On Tue, 2005-29-03 at 23:46 +0100, Ciaran McCreesh wrote:
> On Tue, 29 Mar 2005 16:29:36 -0600 Brian Harring <ferr...@gentoo.org>
> wrote:
> | What are you going to do when gtk v3 hits?
>
> Will v3 be compatible with v2? Or do you think they haven't learned that
> lesson?

The current gnome policy is that it will stay v2 as long as its
compatible.. so no v3 will not be compatible..

--
Olivier Crête
tes...@gentoo.org
x86 Security Liaison

signature.asc

foser

unread,
Mar 29, 2005, 6:20:10 PM3/29/05
to
On Tue, 2005-03-29 at 16:29 -0600, Brian Harring wrote:
> If you use the approach I've laid out (yes, not new, I laid it out in
> 24439) you wouldn't have to dick around with deprecating a version,
> nor essentially mandating what version is default. You'd leave the
> total control over what versions the user wants to deal with in the
> hands of the _users_, and what versions the package supports would
> be represented properly/clearly in the IUSE.

I know the proposed system obviously and it's still flawed in that
really doesn't deal with USE flag versioning in a consistent,
predictable way.

> So far... I've not really heard a good reason aside from "it's in
> place, we'll just deprecate gtk v1 instead of clean it up" for why
> this cannot be corrected _now_, or really in the past.

That is all the reason needed here. I'm no biggie on the current
situation (altough if people actually read the USE flag descriptions it
wouldn't be half the issue), but interchanging to be deprecated
behaviour with something just as bad is a regression. I'll post more on
some changes that I'd like to see implemented for this later.

And the issue never has been about changing a few occurances in the
tree.

- foser

signature.asc

Brian Harring

unread,
Mar 29, 2005, 6:40:13 PM3/29/05
to
On Wed, Mar 30, 2005 at 01:13:12AM +0200, foser wrote:
> On Tue, 2005-03-29 at 16:29 -0600, Brian Harring wrote:
> > If you use the approach I've laid out (yes, not new, I laid it out in
> > 24439) you wouldn't have to dick around with deprecating a version,
> > nor essentially mandating what version is default. You'd leave the
> > total control over what versions the user wants to deal with in the
> > hands of the _users_, and what versions the package supports would
> > be represented properly/clearly in the IUSE.
>
> I know the proposed system obviously and it's still flawed in that
> really doesn't deal with USE flag versioning in a consistent,
> predictable way.
State how it's flawed please. If you're going to argue this is a
repeat of the gtk2/gtk fiasco that's in the tree currently, please
validate how/why it is. The approach detailed above, and in aug '04
*is* the most flexible approach with use flags that addresses user
needs, and is extensible (gtk3).

> > So far... I've not really heard a good reason aside from "it's in
> > place, we'll just deprecate gtk v1 instead of clean it up" for why
> > this cannot be corrected _now_, or really in the past.
>
> That is all the reason needed here. I'm no biggie on the current
> situation (altough if people actually read the USE flag descriptions it
> wouldn't be half the issue), but interchanging to be deprecated
> behaviour with something just as bad is a regression. I'll post more on
> some changes that I'd like to see implemented for this later.
>
> And the issue never has been about changing a few occurances in the
> tree.

Well, the issue is what then? It's an often requested change. While
people *should* know wth they're flipping on via looking in use.desc,
use flags *should* be relatively sane/clear in their implications, at
least on a general scale.
~brian

Dulmandakh Sukhbaatar

unread,
Mar 29, 2005, 11:40:08 PM3/29/05
to
I think that latest version is better than older ones. My opinion is
don't use gtk flag. Becouse when bootstrapping from stage1, gcc has gtk
USE flag and depends on gtk. And gtk needs g++ and we haven't g++. That
was big problem for me some time before. Almost all software may be
compiled against GTK2, there is no need of gtk. If you want to build
gtk, there is some dependencies and some disc space penalty.

--
gento...@gentoo.org mailing list

Diego "Flameeyes" Pettenò

unread,
Mar 30, 2005, 3:50:09 AM3/30/05
to
On Wednesday 30 March 2005 01:13, foser wrote:
> (altough if people actually read the USE flag descriptions it
> wouldn't be half the issue)
Actually, I had troubles with it also if I had read the descrition:

| gtk2 - Use gtk+-2.0.0 over gtk+-1.2 in cases where a program supports both.

It never says "Needs gtk useflag enabled if present.", and it made me think
"well this enable gtk2 if an app has both... if one has only gtk2 it will
probably be enabled by gtk"... but... packages with both gtk AND gtk2 have
needs both useflag (and doesn't throw a warn line about it, either).

From an user point of view, this is complicated...

Also, as Brian already said, there's no plans in the future to be able to
build everything with gtk2.. also wxGTK is still experimental with gtk2
support, and wxGTK will add gtk/gtk2 useflags to everything uses wxWindows
for the future.. and there are quite many packages which uses it.

Maybe a three-way selection could be used?

gtk? ( gtk1? ( )
gtk2? ( )
)

This way one could disable all gtk, only gtk1, only gtk2, or enable all.

Paul de Vrieze

unread,
Mar 30, 2005, 4:10:16 AM3/30/05
to
On Tuesday 29 March 2005 23:42, Ciaran McCreesh wrote:
> I think it's been rather well illustrated that GTK+-1 support is not
> time dependent, and that GTK+-1 will be around for a very long time...

Just remember there are much older and much much crappier toolkits still
around and used. Taking those as example gtk+-1 will probably stay indeed
until no-one can make it to compile anymore.

Paul

--
Paul de Vrieze
Gentoo Developer
Mail: pau...@gentoo.org
Homepage: http://www.devrieze.net

0 new messages