You are not logged in.

#1 2006-01-15 06:38:24

Joffer
Member
Registered: 2006-01-12
Posts: 69

[Solved] How "dangerous" is testing branch?

How "dangerous" to use is the testing branch? I've used Gentoo before, and used a lot of the ~x86/~x64 ebuilds. Would it be comparable?

Offline

#2 2006-01-15 07:01:15

Dusty
Schwag Merchant
From: Medicine Hat, Alberta, Canada
Registered: 2004-01-18
Posts: 5,986
Website

Re: [Solved] How "dangerous" is testing branch?

I don't know anything about Gentoo so couldn't say if it compares. In general, testing stuff does have bugs (they could be major bugs) and is undocumented. The point is for people to help find bugs before they go current. Its usually used with major package upgrades that potentially have a lot of things to test.

I've used testing on my main system for specific packages I don't want to wait for, but I wouldn't suggest it as a general rule for a production system. Definately not for a server! If you like to play around with new stuff, figure out how it works and what breaks it, testing is for you.

Dusty

Offline

#3 2006-01-15 07:38:22

Gullible Jones
Member
Registered: 2004-12-29
Posts: 4,863

Re: [Solved] How "dangerous" is testing branch?

I use Testing, and find it to be quite stable. Occasionally something won't work exactly how it should, but it doesn't do any catastrophic borking... I wouldn't use it for a server or workstation, but it's fine for a desktop system.

Offline

#4 2006-01-15 16:28:06

Joffer
Member
Registered: 2006-01-12
Posts: 69

Re: [Solved] How "dangerous" is testing branch?

Dusty wrote:

I've used testing on my main system for specific packages I don't want to wait for, but I wouldn't suggest it as a general rule for a production system. Definately not for a server! If you like to play around with new stuff, figure out how it works and what breaks it, testing is for you.

Dusty

Is there a way to only enable a spesific package in testing and not the entire repo? In gentoo we have packages.unmask etc we can add spesific packages to to unmask (eg. use packages currently in testing)

Offline

#5 2006-01-15 18:20:53

Snowman
Developer/Forum Fellow
From: Montreal, Canada
Registered: 2004-08-20
Posts: 5,212

Re: [Solved] How "dangerous" is testing branch?

Joffer wrote:
Dusty wrote:

I've used testing on my main system for specific packages I don't want to wait for, but I wouldn't suggest it as a general rule for a production system. Definately not for a server! If you like to play around with new stuff, figure out how it works and what breaks it, testing is for you.

Dusty

Is there a way to only enable a spesific package in testing and not the entire repo? In gentoo we have packages.unmask etc we can add spesific packages to to unmask (eg. use packages currently in testing)

Yes. Enable the testing repo and place it after your other repo in pacman.conf. To install a package:
pacman -S testing/package
Remember that packages in testing might depends on other packages in testing.

Offline

#6 2006-01-15 18:32:30

Cerebral
Forum Fellow
From: Waterloo, ON, CA
Registered: 2005-04-08
Posts: 3,108
Website

Re: [Solved] How "dangerous" is testing branch?

Just a point, snowman, about using testing in this way.  I recently filed a bug report about wierd dependancy issues I was seeing:
http://bugs.archlinux.org/task/3779

The long and the short of it is that putting [testing] after [current] "breaks every logic that pacman uses to install these things from testing."  So, I guess you have to be careful doing it this way - you may get strange dependancy problems.

Not sure if there's a better way to do it though. tongue

Offline

#7 2006-01-15 18:44:08

Snowman
Developer/Forum Fellow
From: Montreal, Canada
Registered: 2004-08-20
Posts: 5,212

Re: [Solved] How "dangerous" is testing branch?

Cerebral wrote:

Just a point, snowman, about using testing in this way.  I recently filed a bug report about wierd dependancy issues I was seeing:
http://bugs.archlinux.org/task/3779

The long and the short of it is that putting [testing] after [current] "breaks every logic that pacman uses to install these things from testing."  So, I guess you have to be careful doing it this way - you may get strange dependancy problems.

Not sure if there's a better way to do it though. tongue

I am aware of that bug report. That's why I added that last sentence about being carefull about dependencies. I'm not sure about a better way either.  You could use jacman (other pacman GUI might have that feature) to get the dependency tree of the app you want to install from testing and if these dependencies are in testing, then you could install them too.

Offline

#8 2006-01-15 18:56:46

Cerebral
Forum Fellow
From: Waterloo, ON, CA
Registered: 2005-04-08
Posts: 3,108
Website

Re: [Solved] How "dangerous" is testing branch?

Snowman wrote:

I am aware of that bug report. That's why I added that last sentence about being carefull about dependencies. I'm not sure about a better way either.  You could use jacman (other pacman GUI might have that feature) to get the dependency tree of the app you want to install from testing and if these dependencies are in testing, then you could install them too.

Fair enough; afaik my problem wasn't related to dependencies located in [testing], but this thread isn't to discuss that.  smile 

I just wanted to warn Joffer to keep an eye out for odd stuff that may be caused by this setup in pacman.conf.

Offline

#9 2006-01-15 18:59:53

ozar
Member
From: USA
Registered: 2005-02-18
Posts: 1,686

Re: [Solved] How "dangerous" is testing branch?

Hmm... guess you might say I'm more of a "current and extra" kind of guy, but the apps that I have used from testing seemed to have worked just fine.


oz

Offline

#10 2006-01-16 22:04:06

Joffer
Member
Registered: 2006-01-12
Posts: 69

Re: [Solved] How "dangerous" is testing branch?

Snowman wrote:
Cerebral wrote:

Just a point, snowman, about using testing in this way.  I recently filed a bug report about wierd dependancy issues I was seeing:
http://bugs.archlinux.org/task/3779

The long and the short of it is that putting [testing] after [current] "breaks every logic that pacman uses to install these things from testing."  So, I guess you have to be careful doing it this way - you may get strange dependancy problems.

Not sure if there's a better way to do it though. tongue

I am aware of that bug report. That's why I added that last sentence about being carefull about dependencies. I'm not sure about a better way either.  You could use jacman (other pacman GUI might have that feature) to get the dependency tree of the app you want to install from testing and if these dependencies are in testing, then you could install them too.

So what you are saying is if I put [testing] after [current] (and [extra]) and it depends on package X and Y, it will fetch those packages from [current] (or [extra]) if they exists there, possibly giving me some trouble because there may be an updated version of X and/or Y in [testing] which my testing/package needs; where current/X perhaps does not supply the function it does in testing/X? Puhh, that was a long hard sentence yikes

Offline

#11 2006-01-16 22:11:15

Cerebral
Forum Fellow
From: Waterloo, ON, CA
Registered: 2005-04-08
Posts: 3,108
Website

Re: [Solved] How "dangerous" is testing branch?

Joffer wrote:

So what you are saying is if I put [testing] after [current] (and [extra]) and it depends on package X and Y, it will fetch those packages from [current] (or [extra]) if they exists there, possibly giving me some trouble because there may be an updated version of X and/or Y in [testing] which my testing/package needs; where current/X perhaps does not supply the function it does in testing/X? Puhh, that was a long hard sentence yikes

Yeah, I think that's the long and the short of it. wink

Offline

#12 2006-01-19 01:14:48

Joffer
Member
Registered: 2006-01-12
Posts: 69

Re: [Solved] How "dangerous" is testing branch?

Ok. I've added [Solved] to the topic name just to stop this thread as I've gotten the answer to my question.

Offline

#13 2006-01-19 08:58:25

Romashka
Forum Fellow
Registered: 2005-12-07
Posts: 1,054

Re: [Solved] How "dangerous" is testing branch?

Now I always use testing on my 3 vmware machines and everything is OK. smile
I don't have many packages and I had no problems with any package in testing.
Only one thing: once there was gcc-4.1 in testing, now it is 4.0.3, so on every upgrade Pacman tells me this, but i just ignore the message. smile


to live is to die

Offline

#14 2006-01-19 17:50:27

Snowman
Developer/Forum Fellow
From: Montreal, Canada
Registered: 2004-08-20
Posts: 5,212

Re: [Solved] How "dangerous" is testing branch?

Romashka wrote:

Now I always use testing on my 3 vmware machines and everything is OK. smile
I don't have many packages and I had no problems with any package in testing.
Only one thing: once there was gcc-4.1 in testing, now it is 4.0.3, so on every upgrade Pacman tells me this, but i just ignore the message. smile

You might want to downgrade your gcc to gcc 4.0.3.  If 4.1 was removed from testing, then it was  probably for a good reason.  Recently, I've read another post referencing gcc 4.1 and I couldn't even find it in the gnu web page. I'm assuming that gcc 4.1 had a bug/problem so it might be safer not to use it.

Offline

#15 2006-01-20 08:43:47

Romashka
Forum Fellow
Registered: 2005-12-07
Posts: 1,054

Re: [Solved] How "dangerous" is testing branch?

I'm not using gcc, I install only binary packages and don't compile anything on my VMware machines. (Right now I use MinGW for development).
I'll downgrade it when I'll be compiling something with it.
(I really don't like Pacman's dependency on GCC and other stuff. That's why I wished to split ABS and Pacman on Flyspray some time ago.)

Anyway it is recommended to view what packages Pacman wants to install from [testing] to see if it can break anything.


to live is to die

Offline

Board footer

Powered by FluxBB