With portage-2.2.7, I get this time: emerge -uDNpv world 109.79s user 1.15s system 99% cpu 1:51.03 total With portage-2.2.8, I get this time: emerge -uDNpv world 192.06s user 1.22s system 99% cpu 3:13.46 total I have bisected the problem to this commit: commit dee9512d2ffec94b4a9eb4ece5c8345e86d04d32 Author: Sebastian Luther <SebastianLuther@gmx.de> Date: Mon Dec 2 13:52:36 2013 +0100 Fix unnecessary rebuild (bug 487198) This one was caused by a mix of >= and < dependencies. Rename the test as requested by Sebastian to testSlotConflictMixedDependencies
Did the merge list change? Are there more slot operator rebuilds than before? If not, then attach the output with --debug for both versions.
I do not believe there was any change in the merge list -- I was running the same command over and over to bisect the issue, and there was no change to the portage tree or the set of installed packages during that time.
Since providing the original numbers, I re-installed media-libs/mesa with the llvm use flag disabled, which shaved quite a lot off the times. This was causing quite a lot of backtracking due to a conflict between media-libs/mesa and sys-devel/clang. However, dee9512d2ffec94b4a9eb4ece5c8345e86d04d32 is still 27 seconds slower than the revision before it. emerge --debug -uDNpv world > debug-c398a7e0a111081547ae1c4e5f723e0216310a57.log 2>&1 60.23s user 1.12s system 99% cpu 1:01.56 total emerge --debug -uDNpv world > debug-dee9512d2ffec94b4a9eb4ece5c8345e86d04d32.log 2>&1 87.77s user 1.15s system 99% cpu 1:29.07 total
Created attachment 367900 [details] Debug log for "good" revision
Created attachment 367902 [details] Debug log for "bad" revision
There's nothing obvious in the logs. The problem seems to be that it just does more work now.
