On =ibm-jdk-bin-1.4* the nsplugin is only available on x86 and on =ibm-jdk-1.5* it is only available on ppc and x86. On other arches please add package.use.mask entries for nsplugin.
The best solution probably is to package.use.mask nsplugin in the base profile and unmask it for x86 profiles, normal/hardened and ppc. I can implement this if you give the ok.
ok.. haven't followed portage development lately.. If I would have known that I could mask nsplugin on a per package base, I would have done so earlier on ppc64. From my point of view there are no objections. (in other words: this is the OK betelgeuse was looking for ^^)
you can't really use package.use.mask already. or rather, it will only affect systems with portage 2.1.1 installed, so you'll have to handle the other case too...
(In reply to comment #3) > you can't really use package.use.mask already. or rather, it will only affect > systems with portage 2.1.1 installed, so you'll have to handle the other case > too... > That has been handled for a long time with use arch testing in the ebuilds.
no objections from hardened. please dont forget the uclibc/x86 profiles where fitting. (that means binary only builds must depend directly on glibc or the libc they were linked with)
We should be doing this for any vm where a plugin isn't available on other archs =dev-java/sun-jdk-1.5* on amd64 =dev-java/sun-jre-bin-1.5* on amd64
I'm pretty sure we're cool to go ahead with this. In cases where portage < 2.1.1, the ebuilds already warn you when there isn't a plugin available for your particular arch, depsite having USE=nsplugin.
(In reply to comment #6) > We should be doing this for any vm where a plugin isn't available on other > archs > =dev-java/sun-jdk-1.5* on amd64 > =dev-java/sun-jre-bin-1.5* on amd64 Please do, we're fine with it :)
(In reply to comment #5) > no objections from hardened. > please dont forget the uclibc/x86 profiles where fitting. uclibc profiles package.mask glibc so I don't think they apply here
Created attachment 99143 [details] all package.use.mask files in the tree Ok. Here is my try for solving this problem. I catted all package.use.mask files in my checkout to this attachement. If no-one disagrees I will commit this next weekend.
I looked over your proposed package.use.mask files for ppc/ppc64 and they look like they will achieve the desired result, unmasking ibm-jdk-1.5 nsplugin for ppc32 profiles only
looks okie for x86 so when you commit it do it for the x86 stuff as well.
package.use.mask files committed
The hardened/x86/2.6 profile (which does not inherit from hardened/x86) was missed. Please add a package.use.mask to it as well.
(In reply to comment #14) > The hardened/x86/2.6 profile (which does not inherit from hardened/x86) was > missed. Please add a package.use.mask to it as well. Done. Thanks for the heads-up.