On Fri, Nov 6, 2009 at 12:35 AM, doorknob60 wrote:

Here's how I installed my chroot, it works very well (I don't even use any lib32 or bin32 packages anymore, just this): http://wiki.archlinux.org/index.php/Arch64_Install_bundled_32bit_system

Or you can use the official mkarchroot tool: http://wiki.archlinux.org/index.php/DeveloperWiki:Building_in_a_Clean_Chroot

On Thu, Nov 5, 2009 at 9:29 PM, Eric Bélanger wrote:

...

On Thu, Nov 5, 2009 at 7:24 PM, Aaron Schaefer wrote:

...

So, my new machine is up and running (and I figured out my previous packaging issues!)...so I'm updating my jGnash package (http://www.archlinux.org/packages/community/i686/jgnash/) to the latest release and there is also currently a bug report on the package (http://bugs.archlinux.org/task/16665). The bug report correctly states that jGnash will not run with openjdk6 (jre works just fine), so what is the current policy for handling that fact?

I know that no other packages depend on jre directly, and the prefered method is now java-runtime, but doesn't that mean that openjdk6 users will just have this software silently fail?

In this case, make it depends on jre.  You could put a note in the PKGBUILD to explain this dependency.  And, when either openjdk or jgnash  release new versions, you could test to see if they work fine together so you could switch back the depends to java-runtime.

Also, if you're building

...

an i386 package on an x86_64 machine, is there an easy way to test the software to make sure that it's actually working on i386? Thanks in advance...

you could setup a i686 chroot on your x86_64 system.  I believe there's info in the wiki.

Eric

...

-- Aaron "ElasticDog" Schaefer

On Fri, Nov 6, 2009 at 1:34 AM, Heiko Baums wrote:

Am Fri, 6 Nov 2009 00:45:06 -0500 schrieb Eric Bélanger :

...

...

...

In this case, make it depends on jre.  You could put a note in the PKGBUILD to explain this dependency.

Btw., putting a note about a restricted dependency in the PKGBUILD is not the recommended way because the average user doesn't read the PKGBUILD if a binary package can't be installed due to a wrong dependency.

Heiko

That proposed note was not intended for users but for other TU or devs who might be tempted to "fix" the dependency by changing it to java-runtime.

On Fri, Nov 6, 2009 at 1:21 AM, Heiko Baums wrote:

Am Fri, 6 Nov 2009 00:29:49 -0500 schrieb Eric Bélanger :

...

...

I know that no other packages depend on jre directly, and the prefered method is now java-runtime, but doesn't that mean that openjdk6 users will just have this software silently fail?

In this case, make it depends on jre.  You could put a note in the PKGBUILD to explain this dependency.  And, when either openjdk or jgnash  release new versions, you could test to see if they work fine together so you could switch back the depends to java-runtime.

It would be better to set the dependency to java-runtime instead of jre so that openjdk6 users can install and test jgnash. If jgnash really fails to run with openjdk6 - I hadn't had any problems with openjdk6, yet - then a bug report should be filed to upstream of openjdk6 and/or jgnash.

I don't think it's the matter of downstream to restrict the dependency and to force the user using jre instead of openjdk6.

So, please, change the dependency to java-runtime.

Heiko

Yes, these problems should be reported upstream to openjdk6 and/or jgnash. If you change the depends to java-runtime, it might be a good idea to display a warning via a post-upgrade message that there might be problems with openjdk6. So that users won't open bug reports about it.