Results for gcc-2.95 19990602 (prerelease) testsuite on mips-sgi-irix5.2

[craig@jcb-sc.com]

>Setting a breakpoint on G77_ttynam_0 revealed that HAVE_TTYNAME had
>not been set, so it just executed:
>
>        errno = ENOSYS;
>        return -1;
>
>Which seems wrong for a function declared as:
>
>      /* Character */ void G77_ttynam_0 (char *ret_val, ftnlen ret_val_len, integer *lunit)
>
>Maybe you should also set ret_val = 0, or check whether ttynam
>succeeded (how can you test that in g77?)

Indeed, that's a bug, which I just fixed (on the mainline).  Actually,
setting the CHARACTER*(*) string pointed to be ret_val, length ret_val_len,
to all spaces is what the code now does instead of "return -1;".

>The second question is why HAVE_TTYNAME was not set, given that `man
>ttyname' says the function exists.  Considering that neither the
>testcase nor the library function have changed in a long time, the
>answer is probably yet another random failure.  Fortunately, this one
>was for good, because it revealed a bug, either in G77_ttynam_0 or in
>the testcase.

I'll assume a random failure as well.  But if you notice this happening
again, please do investigate it!

Thanks!

        tq vm, (burley)