Allow multislot packages to be added to the world file

via --noreplace. Having these atoms in the world file
will trigger recommendations to run emaint in some
cases, like when running `emerge -e world`. The atoms
need to be in the world file to prevent multislot
packages from being removed by --depclean though.
(trunk r8436)

Bug #198129 - Prevent SLOT atoms like  sys-devel/binutils:0
from being inappropriately recorded in the world file when
USE=multislot is enabled.
(trunk r8437:8439)

svn path=/main/branches/2.1.2/; revision=8440

diff --git a/bin/emerge b/bin/emerge
index b9bebda6b578cd4c2e8118650deceb326afb4dbc..2c79af69e15cddfc7434d55931d08210d9ff2ab9 100755 (executable)
--- a/bin/emerge
+++ b/bin/emerge
@@ -735,17 +735,40 @@ def create_world_atom(pkg_key, metadata, args_set, root_config):
                # If the user gave a specific atom, store it as a
                # slot atom in the world file.
                slot_atom = "%s:%s" % (cp, metadata["SLOT"])
-               # First verify the slot is in the portage tree to avoid
-               # adding a bogus slot like that produced by multislot.
-               if portdb.match(slot_atom):
-                       # Now verify that the argument is precise enough to identify a
-                       # specific slot.
-                       matches = portdb.match(arg_atom)
+
+               # For USE=multislot, there are a couple of cases to
+               # handle here:
+               #
+               # 1) SLOT="0", but the real SLOT spontaneously changed to some
+               #    unknown value, so just record an unslotted atom.
+               #
+               # 2) SLOT comes from an installed package and there is no
+               #    matching SLOT in the portage tree.
+               #
+               # Make sure that the slot atom is available in either the
+               # portdb or the vardb, since otherwise the user certainly
+               # doesn't want the SLOT atom recorded in the world file
+               # (case 1 above).  If it's only available in the vardb,
+               # the user may be trying to prevent a USE=multislot
+               # package from being removed by --depclean (case 2 above).
+
+               mydb = portdb
+               if not portdb.match(slot_atom):
+                       # SLOT seems to come from an installed multislot package
+                       mydb = vardb
+               # If there is no installed package matching the SLOT atom,
+               # it probably changed SLOT spontaneously due to USE=multislot,
+               # so just record an unslotted atom.
+               if vardb.match(slot_atom):
+                       # Now verify that the argument is precise
+                       # enough to identify a specific slot.
+                       matches = mydb.match(arg_atom)
                        matched_slots = set()
                        for cpv in matches:
-                               matched_slots.add(portdb.aux_get(cpv, ["SLOT"])[0])
+                               matched_slots.add(mydb.aux_get(cpv, ["SLOT"])[0])
                        if len(matched_slots) == 1:
                                new_world_atom = slot_atom
+
        if new_world_atom == sets["world"].findAtomForPackage(pkg_key, metadata):
                # Both atoms would be identical, so there's nothing to add.
                return None