Here's an example of genlop averaging the build time for dev-db/postgresql-server and virtual/postgresql-server (giving a badly-off estimate) when it should only use the build time for dev-db/postgresql-server : $ genlop -c Currently merging 1 out of 1 * dev-db/postgresql-server-8.4.2 current merge time: 1 minute and 1 second. ETA: any time now. $ genlop -t postgresql-server * virtual/postgresql-server Mon May 26 10:00:03 2008 >>> virtual/postgresql-server-8.2 merge time: 2 seconds. Wed Apr 1 17:55:33 2009 >>> virtual/postgresql-server-8.2 merge time: 7 seconds. Thu Dec 17 21:10:11 2009 >>> dev-db/postgresql-server-8.4.2 merge time: 3 minutes and 41 seconds. Thu Dec 17 21:10:20 2009 >>> virtual/postgresql-server-8.4 merge time: 9 seconds.
This issue is due to backward compatibility with emerge.log entries produced by <portage-2.1.6 (bug #254972), but it should be safe to remove the backward compatibility now that portage-2.1.6.x has been stable for a long time.
How will that work out for emerge.log entries that were created long ago (goes back to 2007 on this particular installation, >=sys-apps/portage-2.1.6 emerged on october 2008) ?
Well, maybe some backward compatibility would still be useful. That doesn't prevent this bug from getting fixed.
The problem still exists 8 and a half year later with portage 2.3.24 and app-portage/genlop-0.30.9-r1. I had this issue with rust yesterday: After first and only emerge of dev-lang/rust: merge time: 56 minutes and 15 seconds. After first and only emerge of virtual/rust: merge time: 9 seconds. During the second emerge of dev-lang/rust: genlop -c Currently merging 1 out of 1 * dev-lang/rust-1.25.0 current merge time: 6 minutes and 7 seconds. ETA: 22 minutes and 5 seconds. And so, the estimate of the new emerge time was far too low.
Shameless plug: emlop doesn't have that bug and is much faster; it's waiting for inclusion in the portage tree at bug #649904 but you can already get it via layman or cargo. As for fixing genlop, I'm not sure how/if the format changed with portage 2.1.6 ? Didn't notice any difference while implementing emlop, although I do have log files that predate 2.1.6 (that's over 10 years at this stage...).