[gentoo-user] world favorites: pros and cons

[gentoo-user] world favorites: pros and cons

[Daniel]

Good afternoon,


I would like to ask what advantages does one gain from (not) putting
packages in the world file?

I know the use of "emerge --oneshot <some-packages>" emerges packages
without recording them in the world set. I also know that all the
packages installed as dependencies don't get recorded in the world set
either.

I see only one advantage in this - the next time I do "emerge --update
world" the checking for available updates would be faster because the
world file doesn't contain all the packages that are actually emerged.

BUT...What happens if there are "critical" updates for packages not
listed in the world?
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1039 bytes --]

On Wed, 05 Jul 2006 12:48:31 +0300, Daniel wrote:

> I know the use of "emerge --oneshot <some-packages>" emerges packages
> without recording them in the world set. I also know that all the
> packages installed as dependencies don't get recorded in the world set
> either.
> 
> I see only one advantage in this - the next time I do "emerge --update
> world" the checking for available updates would be faster because the
> world file doesn't contain all the packages that are actually emerged.
> 
> BUT...What happens if there are "critical" updates for packages not
> listed in the world?

You won't see them, nor any updates to packages that are only dependencies
of package not in world. In short, you break your system.

The only time I use --oneshot for new installs is when trying a package
to see if I want it. If I do, I add it to world with --noreplace. If I
don't find it useful, my next emerge --depclean reminds me to remove it.


-- 
Neil Bothwick

Foolproof operation: No provision for adjustment.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] world favorites: pros and cons

[Alexander Skwar]

Daniel wrote:

> BUT...What happens if there are "critical" updates for packages not
> listed in the world?

They won't get installed. That's why I always do "emerge --deep --update"
(or rather: "emerge -Duvat"), as then packages which are installed to
meet dependencies, will also get updated.

But you'll still miss some packages this way - packages which aren't
in the world file and which are also no dependency of *CURRENTLY* installed
packages. Those are normally packages, which aren't used anymore and
could be removed. I forgot how to find out, which packages that are.

Alexander Skwar
-- 
The more laws and order are made prominent, the more thieves and
robbers there will be.
		-- Lao Tsu
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Alexander Skwar]

Neil Bothwick wrote:

> The only time I use --oneshot for new installs is when trying a package
> to see if I want it. If I do, I add it to world with --noreplace. If I
> don't find it useful, my next emerge --depclean reminds me to remove it.

I use --oneshot, when the compilation of a package breks, which is
a dependency of a package, that I want.

Suppose, I want "a" and "a" needs "b". Now "b" breaks. I fix it, so that
"b" can be compiled. Then I'd do "emerge -1 b".

Another case, which hit me just recently: "a" needs "b", but it needs "b"
to be compiled with a specific flag. Now "b" is already installed, but "a"
can't get installed. In this case, I'd modify my package.use for package
"b" and again do a "emerge -1 b".

Alexander Skwar
-- 
The more laws and order are made prominent, the more thieves and
robbers there will be.
		-- Lao Tsu
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 479 bytes --]

On Wed, 05 Jul 2006 12:18:20 +0200, Alexander Skwar wrote:

> But you'll still miss some packages this way - packages which aren't
> in the world file and which are also no dependency of *CURRENTLY*
> installed packages. Those are normally packages, which aren't used
> anymore and could be removed. I forgot how to find out, which packages
> that are.

emerge --depclean --pretend


-- 
Neil Bothwick

If Barbie is so popular, why do you have to buy her friends?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] world favorites: pros and cons

[Daniel]

Neil Bothwick wrote:
> On Wed, 05 Jul 2006 12:18:20 +0200, Alexander Skwar wrote:
> 
>> But you'll still miss some packages this way - packages which aren't
>> in the world file and which are also no dependency of *CURRENTLY*
>> installed packages. Those are normally packages, which aren't used
>> anymore and could be removed. I forgot how to find out, which packages
>> that are.
> 
> emerge --depclean --pretend
> 
> 

Your replies make me feel I haven't done wrong trying to put every
single package in the world set. Actually in my fear not to miss some
updates I use this script:
---
emerge -DuNpv <package-name> | cut -sf2 -d '/' |\
cut -f1 -d ' '|\
while read pkg;
  do find /usr/portage/ -name ${pkg}.ebuild;
done | sed 's/\/usr\/portage\///g' |\
while read a;
  do echo ${a%/*}; done |\
xargs -n1 emerge
---
This way all dependencies get individually emerged and therefore
recorded in the world file. Of course excluding some particular cases.
For example:
"emerge xmms" -> pulls-in gtk+-1.2", while
"emerge mozilla-firefox" -> pulls-in gtk+2.8.

So in this case the aforementioned script used with "emerge xmms
mozilla-firefox" will individually emerge only gtk+-2.8 and gtk+-1.2
would be emerged as dependency of xmms and won't get recorded in the
world set.



--
Best regards
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Rumen Yotov]

Daniel wrote:

>Neil Bothwick wrote:
>  
>
>>On Wed, 05 Jul 2006 12:18:20 +0200, Alexander Skwar wrote:
>>
>>    
>>
>>>But you'll still miss some packages this way - packages which aren't
>>>in the world file and which are also no dependency of *CURRENTLY*
>>>installed packages. Those are normally packages, which aren't used
>>>anymore and could be removed. I forgot how to find out, which packages
>>>that are.
>>>      
>>>
>>emerge --depclean --pretend
>>
>>
>>    
>>
>
>Your replies make me feel I haven't done wrong trying to put every
>single package in the world set. Actually in my fear not to miss some
>updates I use this script:
>---
>emerge -DuNpv <package-name> | cut -sf2 -d '/' |\
>cut -f1 -d ' '|\
>while read pkg;
>  do find /usr/portage/ -name ${pkg}.ebuild;
>done | sed 's/\/usr\/portage\///g' |\
>while read a;
>  do echo ${a%/*}; done |\
>xargs -n1 emerge
>---
>This way all dependencies get individually emerged and therefore
>recorded in the world file. Of course excluding some particular cases.
>For example:
>"emerge xmms" -> pulls-in gtk+-1.2", while
>"emerge mozilla-firefox" -> pulls-in gtk+2.8.
>
>So in this case the aforementioned script used with "emerge xmms
>mozilla-firefox" will individually emerge only gtk+-2.8 and gtk+-1.2
>would be emerged as dependency of xmms and won't get recorded in the
>world set.
>
>
>
>--
>Best regards
>Daniel
>
>  
>
Hi,
In the contrary, i (at least) put in 'world' only things i emerge.
The reason - the world-file is smaller and eventually is scanned more 
quickly.
Unless you also use "-D|--deep" option, which also scans the deps.
HTH.Rumen
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Rumen Yotov wrote:

>>
> Hi,
> In the contrary, i (at least) put in 'world' only things i emerge.
> The reason - the world-file is smaller and eventually is scanned more
> quickly.
> Unless you also use "-D|--deep" option, which also scans the deps.
> HTH.Rumen

That is clear. Scans for updates take less time but you don't get
completely updated system. Why would one sacrifice updates for shorter
scan timings? That's my real question indeed.


-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 774 bytes --]

On Wed, 05 Jul 2006 14:11:44 +0300, Daniel wrote:

> Your replies make me feel I haven't done wrong trying to put every
> single package in the world set.

In fact you've completely broken the concept of a world set of packages
on your Gentoo system. World should contain only those packages you want
to use, not their dependencies. Now portage has no idea of which packages
are there because you want them, which are there because they are
dependencies of something you want and which are redundant cruft installed
as a dependency of a package you no longer have installed.

On your system, your packages, their dependencies and the cruft are all
considered part of world.


-- 
Neil Bothwick

Found my .sig, it was in behind the cushion on the settee.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Neil Bothwick wrote:
> 
> Now portage has no idea of which packages
> are there because you want them, which are there because they are
> dependencies of something you want and which are redundant cruft installed
> as a dependency of a package you no longer have installed.
> 
> On your system, your packages, their dependencies and the cruft are all
> considered part of world.
> 
> 

That is correct. What are the disadvantages besides the longer seeks for
updates?

I have no problem with the redundant cruft - when I want just to try
some package I do "emerge --pretend" and record the list of dependencies
it wants to pull-in. If I decide the package is not useful to me, I
"un-emerge" not only the package, but also the dependencies it had
pulled-in during its installation.


-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Alexander Skwar]

Daniel Iliev wrote:

> I have no problem with the redundant cruft - when I want just to try
> some package I do "emerge --pretend" and record the list of dependencies
> it wants to pull-in. If I decide the package is not useful to me, I
> "un-emerge" not only the package, but also the dependencies it had
> pulled-in during its installation.

That's risky!

Suppose, you want to install "a". "a" needs "b". You keep "a" & "b"
installed.

Later on, you decide to try "c". "c" needs "b" as well. But as "b"
is already installed, "emerge -p c" won't show "b". You install "c"
and do *NOT* write down, that "c" needs "b", as you don't know that.

Even more later on, you decide to deinstall "a". According to what
you wrote above and according to your documentation, you'll see that
"b" got installed because of "a" and you'll remove "b" as well.

Yet more later on, you find out, that "c" is broken and wonder why.

The basic problem here is, that there's no way to see, which packages
depend on a given package - at least I don't know how to find that out.
What's required, is a way to be told, that packages "a" and "c" depend
on "b".

Now, if you'd use the world file as it was supposed to be used, you'd
remove "a" and could do a "emerge --depclean --pretend". Doing so, the
system would *NOT* show you package "b", as it's still a dependency
of "c". Only after you remove "c" as well, "b" would show up in a
depclean run.

Alexander Skwar
-- 
The more laws and order are made prominent, the more thieves and
robbers there will be.
		-- Lao Tsu
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1078 bytes --]

On Wed, 05 Jul 2006 15:43:53 +0300, Daniel Iliev wrote:

> That is correct. What are the disadvantages besides the longer seeks for
> updates?

What longer seeks? --update only check one level of dependencies for
updates, a few seconds at most. That's nothing compared with the time you
could spend trying to fix a broken system.

> I have no problem with the redundant cruft - when I want just to try
> some package I do "emerge --pretend" and record the list of dependencies
> it wants to pull-in. If I decide the package is not useful to me, I
> "un-emerge" not only the package, but also the dependencies it had
> pulled-in during its installation.

What if you installed something else with overlapping dependencies
between merging and unmerging? You'll break it because you have removed
its dependencies.

The world file is part of how portage manages dependencies, pollute it
with packages that should not be there and portage will not work as it
should.


-- 
Neil Bothwick

Someone who thinks logically is a nice contrast to the real world.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Alexander, Neil thank you for pointing me out this problem.

I think both of you refer to the same scenario and Alexander illustrated
it with an example. For clarity I'll use the same letters to substitute
package names in my next question.


1) I install "a" which pulls-in "c"
2) I *manually* install "c". I install "a"
3) I Install "b". "b" depends on "c". "b" doesn't pull-in "c" because
"c" is already *manually* installed along with "a"
4) I uninstall "a"
5) I *manually* uninstall "c"
6) "b" becomes broken because "c" is no longer in the system


Lets investigate further:


"emerge --deep --update world" will install "c", won't it?

"emerge b" or "emerge c" will solve the problem, won't it?


It appears removing "c" is not as dangerous as it seems at first glance
or I'm wrong?


-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Alan McKinnon]

On Wed, 2006-07-05 at 15:16 +0200, Alexander Skwar wrote:
> The basic problem here is, that there's no way to see, which packages
> depend on a given package - at least I don't know how to find that
> out.

equery depends <given package name>

Not always 100% accurate though, as someone politely pointed out
yesterday

alan


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1277 bytes --]

On Wed, 05 Jul 2006 17:29:16 +0300, Daniel Iliev wrote:

> 1) I install "a" which pulls-in "c"
> 2) I *manually* install "c". I install "a"
> 3) I Install "b". "b" depends on "c". "b" doesn't pull-in "c" because
> "c" is already *manually* installed along with "a"
> 4) I uninstall "a"
> 5) I *manually* uninstall "c"
> 6) "b" becomes broken because "c" is no longer in the system
> 
> 
> Lets investigate further:
> 
> 
> "emerge --deep --update world" will install "c", won't it?
> 
> "emerge b" or "emerge c" will solve the problem, won't it?

so you go to a lot of trouble to circumvent portage's dependency
handling, then you rely on portage to fix things up after your break
them. You need to keep lists of what you have merged and unmerged simply
to compensate for having broken portage's own list for no good reason.

What happens if you reboot after unmerging "c", and its absence causes
the system to fail to boot? What if you remove something that stops
emerge working?

Gentoo is all about choice, so you are free to choose to use it like
this, just as you are free to do "rm -fr /*". But don't expect someone to
come up with a magic fix when things get screwed up.


-- 
Neil Bothwick

Isn't 'Criminal Lawyer' rather redundant?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] world favorites: pros and cons

[Richard Fish]

On 7/5/06, Daniel Iliev <danny@ilievnet.com> wrote:
> That is correct. What are the disadvantages besides the longer seeks for
> updates?

Another disadvantage is that you defeat a big reason for having USE
flags.  For example, if you merge pkg A that USEs X to depend on pkg
B, and you have X in your USE flag, the A will depend on B and pull it
in as a dependancy.

If you later take X out of your use flags, and do an emerge -DNuv
world, the A no longer depends on B.  But since it is still in your
world file, portage will assume you want this package, and continue to
compile updates for it with each new version.  That can be a pretty
huge waste of time.

> I have no problem with the redundant cruft - when I want just to try
> some package I do "emerge --pretend" and record the list of dependencies
> it wants to pull-in. If I decide the package is not useful to me, I
> "un-emerge" not only the package, but also the dependencies it had
> pulled-in during its installation.

You're going through a lot of work to circumvent the dependancy
tracking that is already built into portage.  Why not just merge the
top-level package, and if you don't like it, unmerge and use
--depclean --pretend to figure out what can safely be removed?

And I don't necessarily believe that having everything in world
results in a significantly faster scan time than having only top-level
packages there.  I would like to see actual proof of this assertion.

-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Neil Bothwick wrote:

> 
> so you go to a lot of trouble to circumvent portage's dependency
> handling, then you rely on portage to fix things up after your break
> them. You need to keep lists of what you have merged and unmerged simply
> to compensate for having broken portage's own list for no good reason.

Well I don't have the feeling I go to a lot of trouble and I *absolutely
don't circumvent portage's dependency handling* and I don't see anything
broken in my system even it is about 2 years old.
Keeping lists happens in very rare occasions. Testing a package means I
install, look around and uninstall it. I'm not randomly emerging other
stuff in the mean time.


> What happens if you reboot after unmerging "c", and its absence causes
> the system to fail to boot? What if you remove something that stops
> emerge working?
> 

Highly unlikely. For two reasons:

1) How come that I was able to boot w/o the package in question in first
place? :)
2) The kind of package you're talking about is listed in the system
profile. If you try to remove such a package portage yells out a big fat
warning.


> Gentoo is all about choice, so you are free to choose to use it like
> this, just as you are free to do "rm -fr /*". But don't expect someone to
> come up with a magic fix when things get screwed up.
> 
> 

Correct. And I triggered this discussion here about a different way of
handling packages. A way that is not forbidden neither mentioned as
inappropriate in the official documentation. So there shouldn't be
anything wrong with it, right?
I find your comparison involving "rm -rf /*" to be irrelevant. Using a
system one way or another is not the same as making a "human error".

So far I haven't made the choice of doing "rm -rf /" but actually once I
did "cat /dev/zero > /dev/hda" instead of "cat /dev/zero > /dev/hda2" by
mistake. In cases like this there's no package management system that
could help, no matter if it is portage, apt, yast, swaret or whatever.
Long live the...backups! :)


Last but not least. When it comes to redundant packages in the system.
What happens when you do (the right way?):

1) emerge a
2) "a" pulls-in "b" and "c" as dependencies
3) emerge -C a
4) "a" goes out but "b" and "c" stay there just to take place
5) emerge --depclean

Well...The first thing one can see reads:
" *** WARNING ***  --depclean is known to be broken."

So you prefer to clean the system up using procedure that is "known to
be broken" or you just leave useless packages to take space on your HDDs?

It is my opinion that Gentoo's documentation and portage's behavior
suggest leaving junk packages on your system.
Which indeed is "the right way"?


-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel da Veiga]

On 7/5/06, Daniel Iliev <danny@ilievnet.com> wrote:
> Neil Bothwick wrote:
> >
> > Now portage has no idea of which packages
> > are there because you want them, which are there because they are
> > dependencies of something you want and which are redundant cruft installed
> > as a dependency of a package you no longer have installed.
> >
> > On your system, your packages, their dependencies and the cruft are all
> > considered part of world.
> >
> >
>
> That is correct. What are the disadvantages besides the longer seeks for
> updates?
>
> I have no problem with the redundant cruft - when I want just to try
> some package I do "emerge --pretend" and record the list of dependencies
> it wants to pull-in. If I decide the package is not useful to me, I
> "un-emerge" not only the package, but also the dependencies it had
> pulled-in during its installation.
>
>

You're manually doying stuff that portage should do. This breaks
portage system, gives you more trouble (because you have to manually
"undo" stuff in order to not break your dependency list) and have
turned the whole dependency check lists and ebuils dependency check
useless. A "emerge --update --deep world" for you is a "emerge world".
You put some of the work of portage on your own hands, don't be
surprised if that breaks something.

-- 
Daniel da Veiga
Computer Operator - RS - Brazil
-----BEGIN GEEK CODE BLOCK-----
Version: 3.1
GCM/IT/P/O d-? s:- a? C++$ UBLA++ P+ L++ E--- W+++$ N o+ K- w O M- V-
PS PE Y PGP- t+ 5 X+++ R+* tv b+ DI+++ D+ G+ e h+ r+ y++
------END GEEK CODE BLOCK------
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Richard Fish wrote:

> If you later take X out of your use flags, and do an emerge -DNuv
> world, the A no longer depends on B.  But since it is still in your
> world file, portage will assume you want this package, and continue to
> compile updates for it with each new version.  That can be a pretty
> huge waste of time.

Thanks! Good point!


--snip
>  Why not just merge the
> top-level package, and if you don't like it, unmerge and use
> --depclean --pretend to figure out what can safely be removed?
> 

Because if I decide to keep it, all dependencies it pulls-in don't get
updated until the top-level package starts depending on a different
version of those packages. Actually this is the main reason I started
this practice.
"emerge --depclean" yells a big warning that it is broken.

> And I don't necessarily believe that having everything in world
> results in a significantly faster scan time than having only top-level
> packages there.  I would like to see actual proof of this assertion.
> 
> -Richard

No, no! I'm saying just the opposite - the more packages you have
recorded in the world list, the slower scanning you get.


-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Richard Fish]

On 7/5/06, Daniel Iliev <danny@ilievnet.com> wrote:
> >  Why not just merge the
> > top-level package, and if you don't like it, unmerge and use
> > --depclean --pretend to figure out what can safely be removed?
> >
>
> Because if I decide to keep it, all dependencies it pulls-in don't get
> updated until the top-level package starts depending on a different
> version of those packages. Actually this is the main reason I started
> this practice.

Not if you use --deep on your updates.  Then dependancies are also
considered for updates.  Some people here will tell you that --deep is
troublesome, but I am not one of them, and it seems like what you want
to do.

> "emerge --depclean" yells a big warning that it is broken.

There are 2 "problems" with --depclean:

1. it takes your current use flags into account, rather than those
that were in effect at the time a package was merged.  So if you
modify USE flags, it can report things can be removed, when in reality
that would break something.  But if you do an "emerge -DNvp world",
and it doesn't report anything needing to be [re]merged, then this
doesn't apply.

2. it can remove packages that you really do want.  As an example,
let's say you are programming something that uses the "boost" c++
library.  If you were to remove everything in portage that depended on
boost, and it wasn't in your world file, then depclean would want to
remove it.  The solution here is to add boost to your world file,
since you want that no matter what else is installed.

IMO neither of the above 'problems' are particularly serious, or a
good reason to add every dependancy to world.

> > And I don't necessarily believe that having everything in world
> > results in a significantly faster scan time than having only top-level
> > packages there.  I would like to see actual proof of this assertion.
>
> No, no! I'm saying just the opposite - the more packages you have
> recorded in the world list, the slower scanning you get.

Yeah, well, I don't necessarily believe the reverse either! :-)

Regards,
-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Daniel da Veiga wrote:

> You're manually doying stuff that portage should do. This breaks
> portage system, gives you more trouble (because you have to manually
> "undo" stuff in order to not break your dependency list) and have
> turned the whole dependency check lists and ebuils dependency check
> useless. A "emerge --update --deep world" for you is a "emerge world".
> You put some of the work of portage on your own hands, don't be
> surprised if that breaks something.
> 

OK. I agree that "my way" makes "emerge --update --deep world" equal to
"emerge --update world". Then what is the original purpose of "emerge
--update world"?


-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Richard Fish wrote:

> 
> Not if you use --deep on your updates.  Then dependancies are also
> considered for updates.  Some people here will tell you that --deep is
> troublesome, but I am not one of them, and it seems like what you want
> to do.

Then what is the purpose of:
"emerge --update world" w/o "--deep"?


> 
> There are 2 "problems" with --depclean:
> 
--snip
> IMO neither of the above 'problems' are particularly serious, or a
> good reason to add every dependancy to world.

Well, this means that one has to manually handle things as well as in
the way I deal with packages, right? ;-)


>> No, no! I'm saying just the opposite - the more packages you have
>> recorded in the world list, the slower scanning you get.
> 
> Yeah, well, I don't necessarily believe the reverse either! :-)
> 

Well, I have a Pentium 2 @ 400MHz with 128MB RAM. I use it as a router
and prefer not to even remember of its existence. :)
Let's say once a week I update it, but it has only the base system plus
iptables qmail and squid installed.

My desktop is an Athlnon XP 1700+ (working at 1.9GHz), 512MB RAM.

Compared to it, the router checks for updates about 2 times faster.
I can't be precise, but if you insist I could do a "time emerge -pvuDN
world" on both of them and send the results.

The router world file has 90 lines, the desktop world file has 751
lines. ;-)



-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel da Veiga]

On 7/5/06, Daniel Iliev <danny@ilievnet.com> wrote:
> Daniel da Veiga wrote:
>
> > You're manually doying stuff that portage should do. This breaks
> > portage system, gives you more trouble (because you have to manually
> > "undo" stuff in order to not break your dependency list) and have
> > turned the whole dependency check lists and ebuils dependency check
> > useless. A "emerge --update --deep world" for you is a "emerge world".
> > You put some of the work of portage on your own hands, don't be
> > surprised if that breaks something.
> >
>
> OK. I agree that "my way" makes "emerge --update --deep world" equal to
> "emerge --update world". Then what is the original purpose of "emerge
> --update world"?
>

I'll just quote the "emerge" man page, that is pretty clear there:

--update (-u)
              Updates  packages  to  the best version available, which may not
              always be the highest version number due to masking for  testing
              and  development.   This  will  also  update direct dependencies
              which may not be what you want.  In  general,  use  this  option
              only in combination with the world or system target.

Note the words "DIRECT dependencies". So, your command "emerge
--update --deep world" is in fact just "emerge world", because every
direct/indirect dependency is part of your world file. Your "way" made
"--update" useless, because a simple "emerge <package>" would update
the package.

--deep (-D)
              When used in conjunction with --update, this flag forces  emerge
              to  consider  the entire dependency tree of packages, instead of
              checking only the immediate dependencies of the packages.  As an
              example, this catches updates in libraries that are not directly
              listed in the dependencies of a package.

So, you "way" also made "--deep" useless.
This flags are there because they mantain portage in a way that you
can't easily break consistency by accident, and with that I mean
libraries and indirect dependencies.

I'm not arguing that your system WILL break by putting every single
atom of package installed in world, I just say that you are going
against portage evolution by doying its work, and that MAY cause
problems.

Also, the world file is a simple way to keep a package version (by
removing it from world), for instance, I don't wanna upgrade mysql
with my nightly "emerge -uDN world", so, its not in my world file.

Also note that "indirect" dependencies can be a pain, and packages may
depend on a LOT of other packages, if you want an example, check
"emerge -euDt links -pv". You can check indirect dependencies! I just
say there are quite a few, and portage knows how to deal with all this
stuff (at least never proved me wrong).

-- 
Daniel da Veiga
Computer Operator - RS - Brazil
-----BEGIN GEEK CODE BLOCK-----
Version: 3.1
GCM/IT/P/O d-? s:- a? C++$ UBLA++ P+ L++ E--- W+++$ N o+ K- w O M- V-
PS PE Y PGP- t+ 5 X+++ R+* tv b+ DI+++ D+ G+ e h+ r+ y++
------END GEEK CODE BLOCK------
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Richard Fish]

On 7/5/06, Daniel Iliev <danny@ilievnet.com> wrote:
> Then what is the purpose of:
> "emerge --update world" w/o "--deep"?

To update only the packages in world, without updating dependancies.
As I think I mentioned, some people do not like using --deep, because
they don't necessarily want to update all libraries to the latest
available version for fear of introducing instability/bugs into their
systems.  So they *may* want to update to the latest firefox, but that
doesn't mean they want the latest gtk+ libraries as well.  Presumably
they also monitor the GLSA channels to make sure they don't miss
important security updates...

> Well, this means that one has to manually handle things as well as in
> the way I deal with packages, right? ;-)

Well, yes, but only for the few things that you really care about, not
the entire system.  And why --depclean should always be run with
--pretend first.

> Compared to it, the router checks for updates about 2 times faster.
> I can't be precise, but if you insist I could do a "time emerge -pvuDN
> world" on both of them and send the results.

Ok, but that is for two completely different systems with different
sets of packages installed.  It doesn't tell us whether the time is a
function of the total number of packages that are installed, or the
number of things listed in world.  The question is, if your athlon
didn't have any dependancies in world, would the update check run
faster or slower?  I don't _actually_ care about the answer, I'm just
pointing out that comparing the performance of systems with different
sets of packages installed isn't a good way to test how the
performance of portage relates to the size of the world file.

-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Daniel da Veiga wrote:
> 
> I'll just quote the "emerge" man page, that is pretty clear there:
> 
--snip
> 
> Note the words "DIRECT dependencies". So, your command "emerge
> --update --deep world" is in fact just "emerge world", because every
> direct/indirect dependency is part of your world file. Your "way" made
> "--update" useless, because a simple "emerge <package>" would update
> the package.
> 

--snip

> So, you "way" also made "--deep" useless.
> This flags are there because they mantain portage in a way that you
> can't easily break consistency by accident, and with that I mean
> libraries and indirect dependencies.

Yes, and I'll ask again what's the point of doing:

"emerge world" or "emerge --update world"?

If one doesn't use "--deep" not all the packages get updated. Thats what
bothers me. Later on this mail you say that even you make "emerge -iDN
world" on a daily basis.


--snip

> Also, the world file is a simple way to keep a package version (by
> removing it from world), for instance, I don't wanna upgrade mysql
> with my nightly "emerge -uDN world", so, its not in my world file.
> 

Nothing prevents me of doing the same thing, right? ;-)

> Also note that "indirect" dependencies can be a pain, and packages may
> depend on a LOT of other packages, if you want an example, check
> "emerge -euDt links -pv". You can check indirect dependencies! I just
> say there are quite a few, and portage knows how to deal with all this
> stuff (at least never proved me wrong).
> 

Yes, and putting almost all of the packages in the world list does not
prevent portage of doing its job.

So who and why would use "emerge world" and "emerge --update world" ?

-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1998 bytes --]

On Wed, 05 Jul 2006 19:53:42 +0300, Daniel Iliev wrote:

> Neil Bothwick wrote:
> > What happens if you reboot after unmerging "c", and its absence causes
> > the system to fail to boot? What if you remove something that stops
> > emerge working?
> > 
> 
> Highly unlikely. For two reasons:
> 
> 1) How come that I was able to boot w/o the package in question in first
> place? :)

You did have the package. ??/i mentioned rebooting after removing it, so
it was there before.

> 2) The kind of package you're talking about is listed in the system
> profile. If you try to remove such a package portage yells out a big fat
> warning.

Not necessarily, it is possible to break things with non-system packages.

> Last but not least. When it comes to redundant packages in the system.
> What happens when you do (the right way?):
> 
> 1) emerge a
> 2) "a" pulls-in "b" and "c" as dependencies
> 3) emerge -C a
> 4) "a" goes out but "b" and "c" stay there just to take place
> 5) emerge --depclean
> 
> Well...The first thing one can see reads:
> " *** WARNING ***  --depclean is known to be broken."
> 
> So you prefer to clean the system up using procedure that is "known to
> be broken" or you just leave useless packages to take space on your
> HDDs?

That text is fairly old and hardly applies any more, at least in my
experience. As Richard mentioned, it can fall over when USE flags have
changed, but the rest of the earning, that you didn't quote, tells you to
run emerge --update --newuse --deep before using it. If you do so, your
USE flags will be consistent and it won't break things. I always use it
with --ask anyway.

> It is my opinion that Gentoo's documentation and portage's behavior
> suggest leaving junk packages on your system.
> Which indeed is "the right way"?

Only if you break the file it uses to determine which packages are junk.


-- 
Neil Bothwick

Good fortune will find you provided you left clear instructions.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] world favorites: pros and cons

[Daniel Iliev]

Daniel wrote:
> Good afternoon,
> 
> 
> I would like to ask what advantages does one gain from (not) putting
> packages in the world file?
> 
> I know the use of "emerge --oneshot <some-packages>" emerges packages
> without recording them in the world set. I also know that all the
> packages installed as dependencies don't get recorded in the world set
> either.
> 
> I see only one advantage in this - the next time I do "emerge --update
> world" the checking for available updates would be faster because the
> world file doesn't contain all the packages that are actually emerged.
> 
> BUT...What happens if there are "critical" updates for packages not
> listed in the world?


I would like to thank everyone who took part in this "low priority" thread.
I think its enough what we exchanged as thoughts, ideas and arguments so
far. My suggestion is that if everyone agrees we should consider this
topic closed. Of course its only my opinion and if somebody feels that
he/she has to add something important the list is still open :)

I would try to draw a fair general the conclusions from the thread:

1) Putting packages in the world file is unlikely to corrupt the system
2) Putting package dependencies in the world set leads to manual work
without providing any advantages and may lead to problems
3) The best way to handle packages is to let portage do its job without
external tweaking (wise!) :)
4) The best reason to put manually individual packages in world set is
to protect them against removing with "emerge --depclean"
5) The best reason for manual removing individual packages from the
world set is to prevent them from upgrading.

I hope it's a fear conclusion.


Thanks, guys.


-- 
Best regards,
Daniel

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Ryan Tandy]

Daniel Iliev wrote:
> 5) The best reason for manual removing individual packages from the
> world set is to prevent them from upgrading.

I wouldn't call that a good reason.  /etc/portage is there for that kind 
of thing.  If you remove a package from world, and nothing depends on 
it, then it'll get swept up next time you --depclean without remembering 
to put it back first.  If you remove a package from world, and something 
depends on it, it'll get upgraded anyway next time you -u or -uD.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] world favorites: pros and cons

[Enrico Weigelt]

* Daniel da Veiga <danieldaveiga@gmail.com> wrote:

<snip>

> Also, the world file is a simple way to keep a package version (by
> removing it from world), for instance, I don't wanna upgrade mysql
> with my nightly "emerge -uDN world", so, its not in my world file.

Isn't there any way for nailing it down to some specific version ?


cu
-- 
---------------------------------------------------------------------
 Enrico Weigelt    ==   metux IT service - http://www.metux.de/
---------------------------------------------------------------------
 Please visit the OpenSource QM Taskforce:
 	http://wiki.metux.de/public/OpenSource_QM_Taskforce
 Patches / Fixes for a lot dozens of packages in dozens of versions:
	http://patches.metux.de/
---------------------------------------------------------------------
-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: world favorites: pros and cons

[Jim Ramsay]

Enrico Weigelt wrote:
> * Daniel da Veiga <danieldaveiga@gmail.com> wrote:
> > Also, the world file is a simple way to keep a package version (by
> > removing it from world), for instance, I don't wanna upgrade mysql
> > with my nightly "emerge -uDN world", so, its not in my world file.
> 
> Isn't there any way for nailing it down to some specific version ?

Sure, mask anything higher than the version you want in
your /etc/portage/package.mask file:

 >dev-db/mysql-4.1.20

-- 
Jim Ramsay
"Me fail English?  That's unpossible!"

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: world favorites: pros and cons

[Alexander Skwar]

Jim Ramsay schrieb:
> Enrico Weigelt wrote:
>> * Daniel da Veiga <danieldaveiga@gmail.com> wrote:
>> > Also, the world file is a simple way to keep a package version (by
>> > removing it from world), for instance, I don't wanna upgrade mysql
>> > with my nightly "emerge -uDN world", so, its not in my world file.
>> 
>> Isn't there any way for nailing it down to some specific version ?
> 
> Sure, mask anything higher than the version you want in
> your /etc/portage/package.mask file:
> 
>  >dev-db/mysql-4.1.20

And then put the ebuild you're using into a local overlay. Reason:
Sooner or later, the package might/will be dropped (as it's too
old) but you'll still need the ebuild.

Alexander Skwar
-- 
"I'm not a god, I was misquoted."
		-- Lister, Red Dwarf
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: world favorites: pros and cons

[Enrico Weigelt]

* Alexander Skwar <listen@alexander.skwar.name> wrote:

<snip>

> >Sure, mask anything higher than the version you want in
> >your /etc/portage/package.mask file:
> >
> > >dev-db/mysql-4.1.20
> 
> And then put the ebuild you're using into a local overlay. Reason:
> Sooner or later, the package might/will be dropped (as it's too
> old) but you'll still need the ebuild.

Yeah, but this would also prevent fixes in this ebuild. 
(ie. if there's a typo or broken dependency)

cu
-- 
---------------------------------------------------------------------
 Enrico Weigelt    ==   metux IT service - http://www.metux.de/
---------------------------------------------------------------------
 Please visit the OpenSource QM Taskforce:
 	http://wiki.metux.de/public/OpenSource_QM_Taskforce
 Patches / Fixes for a lot dozens of packages in dozens of versions:
	http://patches.metux.de/
---------------------------------------------------------------------
-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: world favorites: pros and cons

[Jim Ramsay]

Enrico Weigelt wrote:
> Yeah, but this would also prevent fixes in this ebuild. 
> (ie. if there's a typo or broken dependency)

Then you could do:

 >=dev-db/mysql-4.2

This would allow any mysql-4.1.*, including 4.1.20-r1 and 4.1.99

Though you do raise a good point... it would be nice to specify a "real
upstream version number" differently from the Gentoo "interim release
number" such as -r2.  Can you think of a sane way to do that?

My suggestion would be to use a '+' character to represent "all
possible gentoo interim releases", as follows:

Masking
 >dev-db/mysql-4.1.20+
should allow mysql-4.1.20-r1 but not mysql-4.1.21

-- 
Jim Ramsay
"Me fail English?  That's unpossible!"

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: world favorites: pros and cons

[Daniel da Veiga]

On 7/28/06, Jim Ramsay <i.am@jimramsay.com> wrote:
> Enrico Weigelt wrote:
> > Yeah, but this would also prevent fixes in this ebuild.
> > (ie. if there's a typo or broken dependency)
>
> Then you could do:
>
>  >=dev-db/mysql-4.2
>
> This would allow any mysql-4.1.*, including 4.1.20-r1 and 4.1.99
>
> Though you do raise a good point... it would be nice to specify a "real
> upstream version number" differently from the Gentoo "interim release
> number" such as -r2.  Can you think of a sane way to do that?
>
> My suggestion would be to use a '+' character to represent "all
> possible gentoo interim releases", as follows:
>
> Masking
>  >dev-db/mysql-4.1.20+
> should allow mysql-4.1.20-r1 but not mysql-4.1.21
>

Take a look at the "~" operator...

-- 
Daniel da Veiga
Computer Operator - RS - Brazil
-----BEGIN GEEK CODE BLOCK-----
Version: 3.1
GCM/IT/P/O d-? s:- a? C++$ UBLA++ P+ L++ E--- W+++$ N o+ K- w O M- V-
PS PE Y PGP- t+ 5 X+++ R+* tv b+ DI+++ D+ G+ e h+ r+ y++
------END GEEK CODE BLOCK------
-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: world favorites: pros and cons

[Jim Ramsay]

Daniel da Veiga wrote:
> Take a look at the "~" operator...

That's what I get for not RTFM first!

>From 'man ebuild':

~ means match any revision of the base version specified.  So in the
above example (~net-libs/libnet-1.0.2a), we would match versions
'1.0.2a', '1.0.2a-r1', '1.0.2a-r2', etc...

So the correct answer(?) would be to mask:

 >~dev-db/mysql-4.1.20

Then you get bugfixes but no new versions.

-- 
Jim Ramsay
"Me fail English?  That's unpossible!"

-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: world favorites: pros and cons

[Jim Ramsay]

Daniel da Veiga wrote:
> Take a look at the "~" operator...

That's what I get for not RTFM first!

From 'man ebuild':

~ means match any revision of the base version specified.  So in the
above example (~net-libs/libnet-1.0.2a), we would match versions
'1.0.2a', '1.0.2a-r1', '1.0.2a-r2', etc...

So the correct answer(?) would be to mask:

 >~dev-db/mysql-4.1.20

Then you get bugfixes but no new versions.

-- 
Jim Ramsay
"Me fail English?  That's unpossible!"

--
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: world favorites: pros and cons

[Enrico Weigelt]

* Jim Ramsay <i.am@jimramsay.com> wrote:

<snip>

> Then you could do:
> 
>  >=dev-db/mysql-4.2
> 
> This would allow any mysql-4.1.*, including 4.1.20-r1 and 4.1.99
> 
> Though you do raise a good point... it would be nice to specify a "real
> upstream version number" differently from the Gentoo "interim release
> number" such as -r2.  Can you think of a sane way to do that?

Simply normalize the version numbers to a fixed amount of digit 
and use another one for gentoo's revisions ?

I'm using 4 digits at my own buildsystem, but sometimes the upstream
requires them all for itself, so we need some more, ie. 6

<snip>

> Masking
>  >dev-db/mysql-4.1.20+
> should allow mysql-4.1.20-r1 but not mysql-4.1.21

"=dev-db/mysql-4.1.20.*" seems to be cleaner to me.

Maybe 

"=dev-db/mysql-4.2.20" could be considered equal.


cu
-- 
---------------------------------------------------------------------
 Enrico Weigelt    ==   metux IT service - http://www.metux.de/
---------------------------------------------------------------------
 Please visit the OpenSource QM Taskforce:
 	http://wiki.metux.de/public/OpenSource_QM_Taskforce
 Patches / Fixes for a lot dozens of packages in dozens of versions:
	http://patches.metux.de/
---------------------------------------------------------------------
-- 
gentoo-user@gentoo.org mailing list