estimate assertion trips with PREEMPT_EVERYWHERE + ICB mode

[bblum]

assertion is the ASSERT_FRACTIONAL one on proportion

not sure if fault of icb or not

only certain test cases (paradise lost) makes me think that it's likely an overflow bug, state spaces are too large and values overflow and proportion goes wildly out of the [0,1] range

log files are in tryeverywhere-iota/kappa, test cases:

ls-id-log-20160413230553.log.sQCClL.gz
ls-id-log-20160413230609.log.vsbc40.gz
ls-id-log-20160413231101.log.O4ugAn.gz
ls-id-log-20160414063316.log.CbIB5S.gz
ls-id-log-20160413231238.log.wEfU2S.gz
ls-id-log-20160413230526.log.dWhggD.gz
ls-id-log-20160418051413.log.onyVjn.gz
ls-id-log-20160413230325.log.3aYxeJ.gz
ls-id-log-20160414094745.log.Y3qBwG.gz
ls-id-log-20160414192119.log.bxobzo.gz
ls-id-log-20160415150729.log.iRiabW.gz
ls-id-log-20160414203256.log.Y0ojgo.gz
ls-id-log-20160414195748.log.i4HVsr.gz
ls-id-log-20160414174725.log.ROFJgb.gz
ls-id-log-20160415093136.log.i6SpZH.gz
ls-id-log-20160415051036.log.uNhHft.gz
ls-id-log-20160414202616.log.oilaME.gz
ls-id-log-20160415071822.log.vhaSVO.gz
ls-id-log-20160413234706.log.u8zuc9.gz
ls-id-log-20160428012235.log.cpCdEB.gz
ls-id-log-20160414090844.log.UDCBMf.gz
ls-id-log-20160413230809.log.NEqL3w.gz
ls-id-log-20160414090430.log.Pxuzun.gz
ls-id-log-20160415150746.log.IQJaTH.gz
ls-id-log-20160418033334.log.rJdK9w.gz
ls-id-log-20160414090457.log.QvITEl.gz
ls-id-log-20160414190851.log.gm5HhS.gz
ls-id-log-20160413230420.log.UEINlm.gz
ls-id-log-20160413230934.log.3P3qpZ.gz
ls-id-log-20160413230237.log.wtWWHl.gz
ls-id-log-20160413230236.log.PnfAzm.gz
ls-id-log-20160413230233.log.MirLeF.gz
ls-id-log-20160413230236.log.u73Nc9.gz
ls-id-log-20160429073418.log.IIk1cI.gz
ls-id-log-20160414124825.log.NJ8W8Y.gz
ls-id-log-20160416112227.log.tSNVd5.gz
ls-id-log-20160414103850.log.hPgN8R.gz
ls-id-log-20160414142610.log.M5kK5n.gz
ls-id-log-20160415132149.log.JmBhdb.gz
ls-id-log-20160414021631.log.0N6AwT.gz
ls-id-log-20160413230446.log.BL4A3a.gz
ls-id-log-20160415210833.log.FLz1GE.gz
ls-id-log-20160413230235.log.Ctysts.gz
ls-id-log-20160413230234.log.FfVmjF.gz
ls-id-log-20160414000303.log.k53A9L.gz
ls-id-log-20160413230234.log.Tcxs6E.gz
ls-id-log-20160414215001.log.Z2TPkK.gz
ls-id-log-20160413230236.log.ZllgCf.gz
ls-id-log-20160413231436.log.CQoPSv.gz
ls-id-log-20160416202733.log.AEsuob.gz
ls-id-log-20160413230336.log.qaa7qg.gz
ls-id-log-20160413230236.log.fSPUYv.gz
ls-id-log-20160415054410.log.vsgAXQ.gz
ls-id-log-20160415154201.log.138nLX.gz
ls-id-log-20160413230219.log.6TUQcX.gz
ls-id-log-20160413230235.log.eiRfTL.gz
ls-id-log-20160414173752.log.nl697a.gz
ls-id-log-20160418065256.log.kmbL4D.gz
ls-id-log-20160413230236.log.sEdYq7.gz
ls-id-log-20160413230233.log.clXemn.gz
ls-id-log-20160414091316.log.MwW9Y1.gz
ls-id-log-20160416200612.log.8ZC47w.gz
ls-id-log-20160416102310.log.NLt10J.gz
ls-id-log-20160414141326.log.WeKlfu.gz
ls-id-log-20160416070422.log.rKKYFZ.gz
ls-id-log-20160413230229.log.Yh9PeU.gz
ls-id-log-20160413230234.log.93hC9d.gz
ls-id-log-20160414090508.log.RqLOUM.gz
ls-id-log-20160413230154.log.LRUfjL.gz
ls-id-log-20160414214654.log.lcoNuB.gz
ls-id-log-20160413230340.log.oTXjWT.gz
ls-id-log-20160413230235.log.zU57th.gz
ls-id-log-20160414090342.log.axjrhu.gz
ls-id-log-20160414123944.log.5DbrsB.gz
ls-id-log-20160502212139.log.Z61lQZ.gz
ls-id-log-20160502213540.log.E0fQQs.gz
ls-id-log-20160502213534.log.FqcAEU.gz
ls-id-log-20160502213532.log.YOoAye.gz
ls-id-log-20160502213543.log.KB3nty.gz
ls-id-log-20160502213537.log.q8fGNu.gz
ls-id-log-20160502213538.log.1rA2F8.gz

[bblum]

after commenting out the "val \in [0,1]" assertion, i started getting values like "-0.0000%". so more likely than overflow is some crazy floating point imprecision problem when the proportion is incredibly tiny.