| View previous topic :: View next topic |
| Author |
Message |
charles17
Advocate
Joined: 02 Mar 2008
Posts: 3664
|
 Posted: Sun Jan 21, 2018 3:36 pm Post subject: |
 |
|
| bunder wrote: | | yes that is true, but they are needed by other packages on your system. |
Wrong. If they were needed by other packages, there was no reason to have them in the world file. They are in the world file - as everything there - because I installed them.
| bunder wrote: | | you can choose whether to remove them from world or not. |
That's why we have the script shown at https://wiki.gentoo.org/wiki/World_set_%28Portage%29#Keep_it_clean |
|
| Back to top |
|
 |
charles17
Advocate
Joined: 02 Mar 2008
Posts: 3664
|
| Posted: Sun Jan 21, 2018 3:54 pm Post subject: |
|
|
| helecho wrote: | charles17, I do not understand why you want to add a second step. Your algorithm could be done in one step.
| Code: | #!/bin/bash
while read -r i ; do
if equery depends "$i" >> /dev/null ; then
echo $i' needs to stay in @world'
else
echo "$i can be deselected"
echo $i >> /tmp/deselect
fi
done < /var/lib/portage/world |
However, this is the opposite result to the expected one. |
In case you really want it one-step, try | Code: | #!/bin/bash
while read i ; do \
if [ -n "$(emerge -p --quiet --depclean $i)" ]; then \
echo $i' needs to stay in @world, otherwise next depclean would unmerge it.'
else
echo $i' can be deselected'
echo $i >> /tmp/deselect
fi
done < /var/lib/portage/world
|
|
|
| Back to top |
|
|
saboya
Guru
Joined: 28 Nov 2006
Posts: 552
Location: Brazil
|
| Posted: Sun Jan 21, 2018 3:55 pm Post subject: |
|
|
The concept that packages that are depended upon by others shouldn't be in the world file is completely wrong. It's a very real possibility that you have a package A that is a dependency for package B but you actually want A installed regardless of B.
That script is just a bad practice and the real issue is managing the packages you want installed correctly. |
|
| Back to top |
|
|
charles17
Advocate
Joined: 02 Mar 2008
Posts: 3664
|
| Posted: Sun Jan 21, 2018 3:58 pm Post subject: |
|
|
| saboya wrote: | | The concept that packages that are depended upon by others shouldn't be in the world file is completely wrong. It's a very real possibility that you have a package A that is a dependency for package B but you actually want A installed regardless of B. |
As long B is installed there is no reason for A being listed in the world file. Except you intentionally put it there for some strange reason. |
|
| Back to top |
|
|
Mr. T.
Guru
Joined: 26 Dec 2016
Posts: 477
|
| Posted: Sun Jan 21, 2018 4:40 pm Post subject: |
|
|
| charles17 wrote: | | saboya wrote: | | The concept that packages that are depended upon by others shouldn't be in the world file is completely wrong. It's a very real possibility that you have a package A that is a dependency for package B but you actually want A installed regardless of B. |
As long B is installed there is no reason for A being listed in the world file. Except you intentionally put it there for some strange reason. |
| charles17 wrote: | In case you really want it one-step, try
| Code: | #!/bin/bash
while read i ; do \
if [ -n "$(emerge -p --quiet --depclean $i)" ]; then \
echo $i' needs to stay in @world, otherwise next depclean would unmerge it.'
else
echo $i' can be deselected'
echo $i >> /tmp/deselect
fi
done < /var/lib/portage/world |
|
charles17, you say something but you do the contrary!!
Edition: The confusion between the kind of dependencies seems unlikely.
| Code: | -n string
True if the length of string is non-zero.
|
helecho |
|
| Back to top |
|
|
saboya
Guru
Joined: 28 Nov 2006
Posts: 552
Location: Brazil
|
| Posted: Sun Jan 21, 2018 5:44 pm Post subject: |
|
|
| charles17 wrote: | | As long B is installed there is no reason for A being listed in the world file. Except you intentionally put it there for some strange reason. |
That's a misinterpretation of the world file concept.
The world file should have packages that you want to explicitly have installed, and let portage figure out the dependencies of those packages. There's no reason to assume that the world file and it's dependencies should not intersect with each other.
The world file's purpose is not "having the least amount of lines to pull the packages you want", it's about "having the necessary lines to express the packages you want installed". Because if package A depends on package B but you want B installed regardless of A, package B should be in the world file. |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6051
Location: Removed by Neddy
|
| Posted: Sun Jan 21, 2018 5:50 pm Post subject: |
|
|
| saboya wrote: | | charles17 wrote: | | As long B is installed there is no reason for A being listed in the world file. Except you intentionally put it there for some strange reason. |
That's a misinterpretation of the world file concept.
The world file should have packages that you want to explicitly have installed, and let portage figure out the dependencies of those packages. There's no reason to assume that the world file and it's dependencies should not intersect with each other.
The world file's purpose is not "having the least amount of lines to pull the packages you want", it's about "having the necessary lines to express the packages you want installed". Because if package A depends on package B but you want B installed regardless of A, package B should be in the world file. |
which brings us back to the perceived problem with the "automanagement of the world file"
Take my example... gcc
GCC does not need to be listed in the world file because
1) it is part of @system
2) other packages depend on GCC
| Code: | equery depends gcc
* These packages depend on gcc:
dev-java/icedtea-bin-3.6.0 (>=sys-devel/gcc-5.4.0[multilib?])
media-libs/libmypaint-1.3.0 (openmp ? sys-devel/gcc[openmp])
media-libs/mesa-17.3.1 (opencl ? >=sys-devel/gcc-4.6)
net-libs/webkit-gtk-2.18.5 (>=sys-devel/gcc-4.9)
sys-devel/llvm-5.0.1 (>=sys-devel/gcc-3.0)
sys-libs/glibc-2.26-r5 (>=sys-devel/gcc-4.9)
sys-libs/libcxx-5.0.1 (!libcxxrt ? >=sys-devel/gcc-4.7[cxx])
virtual/fortran-0 (sys-devel/gcc[fortran,openmp?]) |
If I have libcxx install why do I need to gcc in the world as portage will sort it out. HOWEVER while my main gcc is 7.2 I have 6.4 explicitly installed for other reasons. Any automanagement of world shortlisting will result in this being removed.
So far the wiki variant produces the most conservative review of the world file that permit additional steps to ensure the file is not scrubbed to much
_________________
| Quote: | | Removed by Chiitoo |
|
|
| Back to top |
|
|
krinn
Watchman
Joined: 02 May 2003
Posts: 7470
|
| Posted: Sun Jan 21, 2018 8:14 pm Post subject: |
|
|
bunder's script is faster, but use an even rawer logic than charles's one
the added value from charles's script is that it doesn't make any assumption of result, but do test the real result of depclean.
the difference is that bunder's script assume "if a package is in world, and it have a dependency -> the package is protect from depclean : remove it from world"
but the script didn't check if that dependency itself have any protection, because the package is protect by the dependency ONLY if that dependency itself is protect.
Which mean (as a real example in my own system) "vlc" in world, and "qdepends result is media-plugins/npapi-vlc-2.0.0 (>=media-video/vlc-1.1)", bunder script will hint me to remove "vlc" from world.
But i don't have media-plugins/npapi-vlc in world, because i have install it to test it and i don't care about it ; while "vlc" is something i want keep.
You could update bunder's script with the depclean test and get a 100% functionality of charles's script (i didn't test, i just assume it should).
The difference of both script will came from first test: Is qdepends -Q X faster than equery d X?
On my system, looks like qdepends is the winner
| Code: | time qdepends -Q vlc
media-plugins/npapi-vlc-2.0.0
real 0m0.014s
user 0m0.007s
sys 0m0.006s
time equery d vlc
media-plugins/npapi-vlc-2.0.0 (>=media-video/vlc-1.1)
real 0m2.921s
user 0m2.845s
sys 0m0.076s
|
It might still be worth to update charles's script to add the "else echo keep it in world" over the qdepends test, or remove any "keep it in world" phrase from it to only keep the "remove it from world" ; for output clarity. |
|
| Back to top |
|
|
Mr. T.
Guru
Joined: 26 Dec 2016
Posts: 477
|
| Posted: Mon Jan 22, 2018 7:51 am Post subject: |
|
|
| The thread progresses in spiral with frequent unnecessary and abusive repetitions. |
|
| Back to top |
|
|
|
Other Things Gentoo
