The package "sci-biology/emboss" was keyworded ~amd64 a few days ago. However, the "emboss" USE flag is still masked in the amd64 profiles. This means optional EMBOSS support cannot be enabled in the following packages: aaindex, CUTG, PRINTS, Prosite, rebase and TransFac. Since these are PDEPENDencies of EMBOSS but are built without EMBOSS support on amd64, EMBOSS installations on amd64 are more or less broken. (Try running "restrict" for instance. You should get an error message about missing data files.) Also, "emboss" should be in the default USE flags for amd64 (just like it is for x86, ppc, ppc64 and ppc-macos) now that EMBOSS is available on amd64, since most users expect EMBOSS support in the above-mentionned packages. The corect fix is to change the amd64 profiles by: - Removing "emboss" from "profiles/default-linux/amd64/use.mask" - Adding "emboss" to the default amd64 USE flags in: "profiles/default-linux/amd64/make.defaults" Thanks in advance,
we can't do that before emboss is stable, otherwise we'll have dependency issues on stable systems
> we can't do that before emboss is stable, otherwise we'll have dependency > issues on stable systems I am not sure this is the case. I could not find any stable package on amd64 depending on EMBOSS. Did I miss one? Also, even if that was the case, the correct thing to do would be, I think, to correct the dependency using constructs such as "!amd? ( emboss? ( sci-biology/emboss ) )" so that EMBOSS support is disabled on amd64 stable packages until EMBOSS is marked stable. Leaving something known to be broken in the testing branch is not acceptable when a solution exists. Please fix however appropriate or package mask EMBOSS on amd64 until you are ready to fix it.
The custom is to wait until the package is stable before de-use.masking. Also, !amd64?( emboss?( ) ) syntax is discouraged by the ebuild quiz and it's deprecated. Should be *always* avoided as possible.
!amd64? ( emboss? (... is UGLY (have a look at the ebuild quiz, it contains such a case, and secondly, it wont do what you think it does. It will just "mask" the flag for the specific package for every amd64 system, not just the stable ones. and even if there's not yet a package which has optional support for emboss: there might be one which gets stable before emboss itself, which then would break the tree again
According to comments from vapier on gentoo-dev, removing a keyword from use.mask may be done even if the package is in the testing branch. (And architecture specific dependencies are indeed prohibited. ;-) This seems like the best solution to the current problem. I think there are no dependency issues since no one would (or should) mark a package stable if it has an optional dependency on an unstable package. Please fix this issue as soon as possible, however you like. Otherwise, I will remove the amd64 keyword from the package; as the EMBOSS maintainer, I don't want EMBOSS keyworded on an architecture where I know it is broken.
are the 7 packages you listed in comment 0 the only ones that make use of emboss? If so, I'm fine with it, although i still don't like the idea of having something out of use.mask before the package required for the use flag is not stable, but in this situation it looks like there's no better solution :/
Yes, these six packages are the only ones that make use of the "emboss" USE flag.
removed emboss flag