Bugzilla Bug 53775

When I perform 'update -uD world', my system upgrades libgcrypt from version
1.1.94 to version 1.1.12. If I perform the same action again, it wants to
downgrade it back.

Reproducible: Always
Steps to Reproduce:
1.emerge -uD world
2.emerge -uD world
3.

Created an attachment (id=33159) [edit]
emerge info

Created an attachment (id=33160) [edit]
qpkq -I -v output

because of opencdk isnt it ?

vapier - yes.

opencdk-0.4.5 (x86 sparc alpha hppa) DEPENDS =dev-libs/libgcrypt-1.1.12 (x86 ppc sparc alpha ia64 amd64 hppa)
opencdk-0.5.1 (~x86 ppc ~sparc alpha amd64 ia64) DEPENDS =dev-libs/libgcrypt-1.1.12 (x86 ppc sparc alpha ia64 amd64 hppa)
opencdk-0.5.3 (~x86 ~ppc ~sparc ~mips ~alpha amd64) DEPENDS >=dev-libs/libgcrypt-1.1.91 (x86 ~ppc amd64)

So based on:
fgrep KEYWORD dev-libs/libgcrypt/libgcrypt*
dev-libs/libgcrypt/libgcrypt-1.1.12.ebuild:KEYWORDS="x86 ppc sparc alpha ia64 amd64 hppa"
dev-libs/libgcrypt/libgcrypt-1.1.91.ebuild:KEYWORDS="x86 ~ppc amd64"
dev-libs/libgcrypt/libgcrypt-1.1.92.ebuild:KEYWORDS="x86 ~amd64 ~sparc"
dev-libs/libgcrypt/libgcrypt-1.1.94.ebuild:KEYWORDS="x86 ~amd64 ~sparc ~mips ~alpha"
dev-libs/libgcrypt/libgcrypt-1.2.0.ebuild:KEYWORDS="~x86 ~amd64 ~sparc ~mips ~alpha ~hppa ~ia64 ~ppc"

Solution 1:
~x86 mask libgcrypt-1.1.2, libgcrypt-1.1.94
(that I unmasked)

Solution 2:

x86 unmask opencdk-0.5.3 (added 23 Jan 2004)

and optionally
x86 sparc unmask opencdk-0.5.1

Does solution 2 work for you liquidx (opencdk maintainer)?

i vote for stabilizing newer opencdk since libgcrypt-1.9.x is in stable and has
been for quite sometime

done - x86 opencdk-0.5.3.ebuild

Is there a deeper problem with portage/repoman not being able to detect these QA flaws?

Now the same problem occurs with two packages.

[ebuild     UD] dev-libs/libgcrypt-1.1.12 [1.1.94] 
[ebuild     UD] app-crypt/opencdk-0.4.5 [0.5.3] 

and then vice-versa.

update  gnutls from 0.8.10 to 0.8.12 and that should fix it... look in the
ebuild and I believe the 0.8.10 calls for a lower opencdk than the newer
ebuild.

Ok I updated to gnutls-1.0.4 and opencdk-0.5.3 and that appears to of fixed the
loop.

Good find Susie, I was stumped ;-).  Leonid work for you?

No, but I still don't see gnutls updated at x86. It's still gnutls-0.8.10. I've
tried to resync.

ok gnutls  0.8.12 and 1.0.4 x86 stable now. Any problems just repopen again.

http://packages.gentoo.org/search/?sstring=gnutls

Sorry for the previous. They both are not in x86 stable yet.

dont use packages.gentoo.org like that

it doesnt get updated immediately

re-open if when you *rsync* it's still not unmasked :P

http://www.gentoo.org/cgi-bin/viewcvs.cgi/net-libs/gnutls/gnutls-0.8.12.ebuild?r1=1.6&r2=1.7

I did miss 1.0.4 though. Will x86 that one when I commit a version. bump which I'm currently doing.

Elaborating on packages - it is updated at periodic intervals - viewcvs is current.

It's working. Thank you, guys.

closing so it doesn't show in my open bugs list. It has been fixed after all.