Dear fellow developers,
here's my promised detailed analysis of the impact of the checkins during that last two weeks and the current state of MonetDB 4.3.16:
To which changes caused which harmless or harmful differences, and to update the stable output according to the harmless differences, I did re-playing the changes that happened during the last two week on a per day per developer basis, i.e., in 6 steps: 2004-04-14_nielsnes, 2004-04-15_boncz, 2004-04-15_nielsnes, 2004-04-16_nielsnes, 2004-04-18_mlkersten, 2004-04-23_boncz.
YES, this means 6 times - compiling MonetDB from scratch, - running Mtest, - checking the output/differences, - running Mapprove where necessary - re-running Mtest, in case Mapprove was necessary (well, it only "cost" me a few hours on early Sunday morning...).
As you saw from my checkins, most changes turned out to be "harmless", and I updated the stable output as necessary (hoping, that the respective developers did indeed intend these changes...).
"Only" the broken join turned out to be indeed "sad truth". A respective bug report has been filed.
Well, that's all from me for now.
I'm off, again...
Stefan