Bug 168710

Summary:	repoman does not seem to check profile's package.use.mask files
Product:	Portage Development	Reporter:	Duncan Coutts (RETIRED) <dcoutts>
Component:	Repoman	Assignee:	Portage team <dev-portage>
Status:	RESOLVED INVALID
Severity:	normal
Priority:	High
Version:	unspecified
Hardware:	All
OS:	Linux
Whiteboard:
Package list:		Runtime testing required:	---

Description Duncan Coutts (RETIRED) gentoo-dev

2007-02-28 11:54:51 UTC

I have a package that optionally depends on librsvg. In a couple profiles librsvg is p.masked. Therefore repoman rightly complains. To fix this I tried masking the appropriate use flag for my package on those profiles. I would have expected this to keep repoman happy but it does not. Fully masking the package on those profiles does work but seems excessive.

Details:
package dev-haskell/gtk2hs-0.9.11
ebuild in the Haskell overlay:
http://haskell.org/~gentoo/gentoo-haskell/dev-haskell/gtk2hs/gtk2hs-0.9.11.ebuild

profiles in question:
default-linux/sparc/sparc64/2006.1/2.4
default-linux/x86/no-nptl/2.4

DEPEND="gnome? ( >=gnome-base/librsvg-2.16 )"

I added the following to package.use.mask in those profile dirs:
dev-haskell/gtk2hs -gnome

repoman full complains with:
  DEPEND.bad                     2
   dev-haskell/gtk2hs/gtk2hs-0.9.11.ebuild: ~sparc(default-linux/sparc/sparc64/2006.1/2.4) ['>=gnome-base/librsvg-2.16']
   dev-haskell/gtk2hs/gtk2hs-0.9.11.ebuild: ~x86(default-linux/x86/no-nptl/2.4) ['>=gnome-base/librsvg-2.16']
  RDEPEND.bad                    2
   dev-haskell/gtk2hs/gtk2hs-0.9.11.ebuild: ~sparc(default-linux/sparc/sparc64/2006.1/2.4) ['>=gnome-base/librsvg-2.16']
   dev-haskell/gtk2hs/gtk2hs-0.9.11.ebuild: ~x86(default-linux/x86/no-nptl/2.4) ['>=gnome-base/librsvg-2.16']


Have I missed anything?

Comment 1 Zac Medico gentoo-dev

2007-02-28 17:41:56 UTC

(In reply to comment #0)
> I added the following to package.use.mask in those profile dirs:
> dev-haskell/gtk2hs -gnome

That's inverted.  It's gnome to mask and -gnome to unmask.  Yes, it's slightly counter intuitive, but it's consistent with the workings of use.mask and with the general behavior of "incrementals" that portage uses in various places.  A little known feature is that it's also possible to unmask entire packages in package.mask via -atom.