Re: fix kennerism

[kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)]

    Unless your machine is very slow, it can manage two bootstraps/two
    regression testins overnight, so you can update your tree, check if
    your patches still apply, run a script with bootstrap/make check on
    the unmodified CVS, patches application and second bootstrap/make
    check (in a different obj tree, so you can see everything in the
    morning).

I don't follow.  The issue isn't changes that occured *before* the test
starts, since I run with those, but changes that occur between when the
test starts and the commit is done.

I don't see what the two runs proposed above accomplishes (or actually
even what the second run *is*), but I see a real disadvantage of having
more than one tree, which is the significant potential of getting confused
as to what state each tree is in.

I'm a big fan of KISS: the more complex the procedure, the higher the chances
of procedural errors.   And there's only so much you win with automation:
"To err is human, but to really mess up requires a computer."