Re: test results for i586-pc-sco3.2v5.0.4

[Craig Burley <burley at gnu dot org>]

>Is it my imagination, or are these tests taking longer and longer
>to run?   These took 7 1/2 hours on a dual pentium 100 system...

If you've recently (in the past 12 hours or so) updated your
tree via CVS, you might have gotten hold of a new test case
added to the g77 test suite (by Dave Love, at my request) that
really beats on a known deficiency in the gcc back end.  The
effect is that a relatively small program -- a "one-liner" that
is basically a fairly long, complicated line -- takes "forever"
to compile on some systems (many minutes on my Pentium II 233MHz
w/128MB RAM, for example).

The submission that had that test case Dave already added also
had a patch to fix the deficiency, but I don't think Jeff Law
has added it yet, and it's just the kind of patch that somebody
like Jeff or Jim should look over, and might want to ask me to
re-do using a different approach, so don't be surprised if it
doesn't get added right away.

In the meantime, you can look up the latest addition(s) to the
gcc/testsuite/g77.f-torture/compile/ directory to determine
which is the "``infinite'' compiler time" test (I forget the
name I asked Dave to give it offhand) and disable it by
either deleting it or renaming it to whatever.f-takes-too-long
or something.

        tq vm, (burley)