Laurynas Biveinis wrote: > Hi, > > After following the mailing list for a while I got impression that it > is not trivial to interpret "make -k check" results, i.e. if you get > FAILs while testing your changes, that does not necessarily mean that > your code actually broke something, as FAILs are rather common.
Well, not common. I think on i686, there are a couple fails that are known (4 or 5) out of thousands of testcases. > > So I guess that there is no baseline as such, and the only meaningful > way to use the results is to compare testsuite outputs with and > without your changes? > More or less. You can look at the gcc-testresults archives, and your baseline should match up pretty closely with one of hte ones there.
