@4u wrote:
> After posting and closing the bug report:
> http://bugs.gentoo.org/show_bug.cgi?id=135870
> Jakub Moc noticed that the current >=virtual/x11-7.0 ebuild misses its
> task and creates trouble.

Indeed. To re-iterate here, I'll basically re-paste what I've said on
the bug, so that people don't have to jump to bugs.g.o.:

>=virtual/x11-7 is hiding breakage in ebuilds that are not ported for
modular X. The side effect is that dependencies like X? ( || ( foo bar
baz ) virtual/x11 ) fail once virtual/x11 gets emerged by one of those
broken ebuilds, because the dependency is already satisfied by
virtual/x11. If that virtual doesn't depend on either of foo bar baz,
then the dependency doesn't get emerged and a perfectly valid ebuild
without any missing dependencies fails.


> For example: This ebuild behaves partly like a ordinary meta build and
> installs imake. You need imake (more correctly xmfmk) to install tightvnc.

Yeah, as it is now, it's essentially a dumpspace for redundant
dependencies that are already stated in ebuilds fixed for modular X, but
that frequently don't get installed b/c of the problem described above.
We are mis-using a 'new style' virtual to produce yet another metabuild
that serves the only purpose - to hide borkage.

> For that reason I want to request the deletion of virtual/x11-7.0 and
> that at least some dependencies of virtual/x11 are moved to
> =>x11-base/xorg-x11-7.0 where these dependencies belong to IMHO.
> xorg-x11 is a meta ebuild.

Each ebuild should state its own dependencies. x11-base/xorg-x11 is a
metabuild for users' convenience, which should produce a pretty
full-featured X server install, but nothing more. It's not a dumpspace
for whatever redundant dependencies either.

So - IMHO we should stop shoving the real breakage under the carpet, if
ebuilds are not ported for modular X, they are broken and should be
fixed. If noone cares to fix them enough for some time, they'll probably
need to be package.masked and subsequently removed from the tree. Until
then, they'll bomb out, because they are broken, that's a perfectly
expected behaviour...

What we are instead doing now, is hiding the breakage by misusing
virtual/x11-7 to emerge most frequently missing deps, which is bloating
more and more, as more and more not broken ebuilds are hit by the
redundant virtual. Not good, really. Some examples of needless borkage
include:

http://bugs.gentoo.org/show_bug.cgi?id=123071
http://bugs.gentoo.org/show_bug.cgi?id=127617
http://bugs.gentoo.org/show_bug.cgi?id=128163
http://bugs.gentoo.org/show_bug.cgi?id=128353
http://bugs.gentoo.org/show_bug.cgi?id=128354

(plus numerous duplicates).

While the above bugs are marked fixed, they won't be really fixed until
>=virtual/x11-7 goes to /dev/null and stops causing more harm than good.

Sorry for a long post, but this problem really needs to be addressed.

-- 
Best regards,

 Jakub Moc
 mailto:jakub@gentoo.org
 GPG signature:
 http://subkeys.pgp.net:11371/pks/lookup?op=get&search=0xCEBA3D9E
 Primary key fingerprint: D2D7 933C 9BA1 C95B 2C95  B30F 8717 D5FD CEBA 3D9E

 ... still no signature   ;)