* Added another test- I believe it is a possible situation and one we need to

deal with. Currently, libalpm with AND without Nagy's patch fails it.
* Slight spacing update to pmenv.py.

pactest/tests/upgrade056.py

@@ -0,0 +1,24 @@
+self.description = "Upgrade a package that removes a provide but another package still does"
+
+lp1 = pmpkg("pkg1")
+lp1.depends = ["imaginary"]
+self.addpkg2db("local", lp1)
+
+lp2 = pmpkg("pkg2")
+lp2.provides = ["imaginary"]
+lp2.requiredby = [ "pkg1" ]
+self.addpkg2db("local", lp2)
+
+lp3 = pmpkg("pkg3")
+lp3.provides = ["imaginary"]
+self.addpkg2db("local", lp3)
+
+p = pmpkg("pkg2", "1.0-2")
+self.addpkg(p)
+
+self.args = "-U %s" % p.filename()
+
+self.addrule("PACMAN_RETCODE=0")
+self.addrule("PKG_EXIST=pkg1")
+self.addrule("PKG_VERSION=pkg2|1.0-2")
+self.addrule("PKG_EXIST=pkg3")