On Fri, 20 May 2022 at 15:31, Hans-Peter Nilsson wrote: > > > From: Jonathan Wakely > > Date: Fri, 20 May 2022 11:03:40 +0200 > > > > Ok to commit (without renaming)? > > > > I'm OK with the timeout factor, but we could also solve it differently > > so that it doesn't take nearly 5 minutes, as in the attached patch. > > The testDiscreteDist function can be parametrized with the number of > > iterations to perform. Would you rather do that? > > Yes thanks, down from 4m39s to 2.7 seconds, so very much > preferable! Nice, thanks for testing it. > (To the skeptics: the coverage intended with the test, is > IMHO reached with all non-simulator targets also running > this. Nothing target-dependent here.) Indeed. > Also in line with many other depth-level-cousin test-files > named value.cc. Still many more others seem to be > candidates for such pruning, judging by the time it takes > for a 'RUNTESTFLAGS=--target_board=cris-sim\ > conformance.exp=values.cc' to get to *that* values.cc. > > Though, some -DSIMULATOR_TEST-adjusted files use dg-options, > others dg-additional-options. It seems the difference is > that by using dg-options, you lose "-include bits/stdc++.h". > Likely not intended. If so, should Someone fix that by > preapproval but regtested? Ah good point. I've pushed the attached patch now. This adjusts the values.cc files for some other distributions, and does so using dg-additional-options not dg-options. Feel free to change existing (mis)uses of dg-options to be dg-additional-options. I've tested it on x86_64-linux, normally and also with -DSIMULATOR_TEST in the test flags, just to ensure I didn't introduce a silly syntax error for that case. I haven't tested it on a real sim though.