x11-wm/twm x11-apps/xclock x11-terms/xterm Remove them please. Reproducible: Always Steps to Reproduce: 1. 2. 3.
Have you looked at the /etc/X11/xinit/xinitrc provided? # Failsafe else # start some nice programs twm & xclock -geometry 50x50-1+1 & xterm -geometry 80x50+494+51 & xterm -geometry 80x20+494-0 & exec xterm -geometry 80x66+0+0 -name login fi Those deps can't be dropped unless the failsafe state changes. I'm certainly open for suggestions for a failsafe state that doesn't require these apps. I see no reason why we can't drop xclock, for example. Donnie, Robert, ideas/rationale?
Rationale being that this has been the traditional failsafe state for ages, so people expect to see it.
No I haven't, but I did now. So just because the apps are used in failsafemode that means that everyone needs to install them? I have never used a failsafemode when it comes to X and I can't see why I would want to. I just want to keep my modular X installation clean and not bloat it with several things that I don't need. Hopefully this situation will be taken care of my someone with better insight on the matter than me, and also that changes are made that pleases both sides. Failsafers and Nosafers alike :)
i can understand the reasoning behind having them, it would still be nice to have the ability to not pull them in.
However, we cannot install a script that relies on things that are optional. The script itself would have to drop the failsafe state when the failsafe apps were not installed. Maybe this would fall under a "minimal" USE-flag?
(In reply to comment #5) > However, we cannot install a script that relies on things that are optional. > The script itself would have to drop the failsafe state when the failsafe apps > were not installed. > > Maybe this would fall under a "minimal" USE-flag? I don't think it's worth it, unless there's some sort of actual benefit besides "the cruft," because even added together, they install less than 1MB worth of files.
(In reply to comment #6) > (In reply to comment #5) > > However, we cannot install a script that relies on things that are optional. > > The script itself would have to drop the failsafe state when the failsafe apps > > were not installed. > > > > Maybe this would fall under a "minimal" USE-flag? That would be ok by me. > > I don't think it's worth it, unless there's some sort of actual benefit besides > "the cruft," because even added together, they install less than 1MB worth of files. It is not that much about the amount of space. It is more about the feeling that you don't have anything you don't need installed. And some people like to look through their list of updates to see what has changed and so on, so when there are less packages it takes less time.
Please remove the depends or make them conditional. That is the point of modular X .. not having to insstall apps you do not need. I would prefer some sort of "fallback"-USE-flag so that the depends are default off .. with "minimal" they would be default on and users would have to add xinit to package.use.
It might be worth comparing our xinitrc with upstream's to see whether there are differences in the runtime dependencies.
Can we please have something like that? What is the problem with it? # diff -u xinitrc.orig xinitrc --- xinitrc.orig 2005-11-03 14:19:08.000000000 +0100 +++ xinitrc 2005-11-03 14:21:40.000000000 +0100 @@ -34,10 +34,5 @@ exec "`/etc/X11/chooser.sh`" # Failsafe else - # start some nice programs - twm & - xclock -geometry 50x50-1+1 & - xterm -geometry 80x50+494+51 & - xterm -geometry 80x20+494-0 & - exec xterm -geometry 80x66+0+0 -name login + echo "No session found in $HOME/.xinitrc and /etc/X11/chooser.sh" >&2 fi
What is wrong with it is that there is no fallback. It's this thing about usability where having working X is important, even if it's using twm. And it's a lot more important to have working X than to not install 1MB worth of apps.
(In reply to comment #11) > What is wrong with it is that there is no fallback. It's this thing about > usability where having working X is important, even if it's using twm. And it's > a lot more important to have working X than to not install 1MB worth of apps. Well then we could probably disable the python use flag. It just installs language bindings that are probably useful in some cases and don't take so much space. What is wrong with having a for example a xsession or something like that use flag pulling these in? After something is implemented for bug 61732 it can be on by default.
(In reply to comment #9) > It might be worth comparing our xinitrc with upstream's to see whether there are > differences in the runtime dependencies. Nope, our fallback is exactly the same.
OK, then this is not going to change in Gentoo without changing upstream. For that reason, I'm marking it upstream. Feel free to file an upstream bug, although I don't expect they'll be willing to change it either, because a sane fallback just makes sense.
*** Bug 137303 has been marked as a duplicate of this bug. ***
*** Bug 149011 has been marked as a duplicate of this bug. ***
*** Bug 168698 has been marked as a duplicate of this bug. ***
*** Bug 263192 has been marked as a duplicate of this bug. ***
*** Bug 424205 has been marked as a duplicate of this bug. ***
The last human response to this bug, in 2005, was comment #14, which concluded, based on discussion in the previous comments: > OK, then this is not going to change in Gentoo without changing upstream. Sometime between then and now, the package dependencies /were/ changed in Gentoo without the xinitrc script changing upstream: the broken configuration that comment 5 and comment 11 advise against is now in the x11-apps/xinit-1.3.1-r1 package. I do not know when or why xinit's dependencies were changed. I don't have a strong opinion on whether this should be solved by changing the fallback configuration, or restoring the x11-apps/xinit dependencies, or doing something else entirely, but xinit should not have a default configuration which simply fails to work, as it does now.
What the user has to do to get a working startx is properly documented in the x11 documentation: http://www.gentoo.org/doc/en/xorg-config.xml#doc_chap3_sect3 xinit works perfectly fine without twm and xterm. If the user doesn't want them, there is no reason to force an install.
(In reply to comment #23) > xinit works perfectly fine without twm and xterm. If the user doesn't want > them, there is no reason to force an install. A quite valid reason was given in comment #11. It's fine if you don't agree with this reason, but it is, in fact, a reason. So saying "there is no reason" is disingenuous. If Donnie Berkholz's reason is deemed no longer important, then a patch like the one proposed in comment #10 would be very simple to make, and would prevent a broken default xinitrc script. Because it sure seems to me that there is no reason to ship xinit with a default configuration that simply does not work.
If it is properly documented how to make it work (and I think it is, if you disagree please point out deficiencies), then "doesn't work" type problems with the default configuration are secondary IMO. Worst thing that happens is that startx fails, complaining about missing twm/xterm.
(In reply to comment #25) > If it is properly documented how to make it work (and I think it is, if you > disagree please point out deficiencies), It's not documented in the package at all. The document you pointed to is an external page. If this were still 2005, when there was no reasonable way to get an X installation running without a lot of coaxing and hand-holding, requiring an outside install guide would be a reasonable requirement. Today's X can nearly be emerged and run without needing an install guide. In general, over time it's only moving closer to this ideal; including a broken default configuration moves it further away. > then "doesn't work" type problems > with the default configuration are secondary IMO. What has changed since 2005 about the principles of usability to render Donnie's comment #11 obsolete? > Worst thing that happens > is that startx fails, complaining about missing twm/xterm. Since the fallback configuration starts twm (and xclock) in the background, no error message at all is generated. When the foregrounded xterm tries to start, it gives the error "xterm: xterm: cannot execute binary file", a misleading error message, as it implies there is an xterm binary on the system that for some unspecified reason cannot be executed. At a bare minimum, if the packages necessary for the fallback configuration aren't installed, the fallback portion of xinitrc should detect this and generate a meaningful error message, preferably something that gives specific steps to solve the problem. This would be a pretty simple change to xinitrc, and would help preserve usability without forcing installation of extra packages.