All mail-mta/ssmtp ebuilds explicitly list virtual/libc as a dependency. Because virtual/libc is part of the system package set, this is clearly unnecessary. In addition, because the explicit dependency is followed by the dependency resolver, this means that an emerge -e world (for example, after a GCC upgrade) pulls in glibc and gcc, when the previous emerge -e system has already re-merged them, leading to a considerably increased merging time due to the redundant double merge of these big packages. mail-mta/ssmtp should not depend on virtual/libc. In addition, if such a thing doesn't exist, a QA rule should automatically point out packages that are not in the system set but depend on packages in the system set.
(In reply to comment #0) > In addition, because the explicit dependency is followed by the dependency > resolver, this means that an emerge -e world (for example, after a GCC upgrade) > pulls in glibc and gcc, when the previous emerge -e system has already > re-merged them, leading to a considerably increased merging time due to the > redundant double merge of these big packages. > This would always happen with -e world; that's why tools like emwrap.sh [1] and and Guenther's upgrade script [2] were written. [1] http://forums.gentoo.org/viewtopic-t-282474.html [2] http://forums.gentoo.org/viewtopic-t-494331.html Not sure about the dep on virtual/libc in itself; just that removing it would not solve the above issue.
I just committet ssmtp-2.62 which fixes this bug.