Re: Small update to reversed_comparison_code

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Richard" == Richard Kenner <kenner@vlsi1.ultra.nyu.edu> writes:

    Richard> True, but how long do you think the latent bug would have
    Richard> taken to get fixed if the patch had immediately been
    Richard> removed from the branch?  I'd guess far longer, if ever
    Richard> (and probably never).

Perhaps this is one key point where we disagree.  I would argue that
it would have gotten fixed.  Not being able to fix the SPARC bug would
have SPARCed interest in the latent bug.  Anyhow, it took two weeks to
fix it this way -- that's an effective infinity for people trying to
work on the compiler.

But, this is getting tedious.  You're not going to change your mind;
neither am I.  Let's spare everyone else.

    Richard> There are a few cases where regressions *are*
    Richard> system-specific.  In those cases, if they are critical
    Richard> systems, either the patch is removed or the problem
    Richard> (often a latent bug elsewhere) is fixed.  But there have
    Richard> been a number of failures in less critical systems that
    Richard> have been deemed less important than the fix.

But that's a judgement call for ACT management to make.  When someone
checks something into GCC, they're affecting Red Hat, CodeSourcery,
ACT -- and much more important, the entire FSF.

    Richard> Not true in the case of triggering latent bugs.  Look at
    Richard> RTH's patch, the one we've been discussing.  It was a fix
    Richard> to an obscure bug in the way structures are returned in
    Richard> the 64-bit Sparc ABI.  It's reasonable for somebody
    Richard> working in that area to understand the relevant ABI and
    Richard> the config files in question.  But do we *really* want to
    Richard> say that understanding the intricacies of reload_cse is a
    Richard> prerequisite to fixing a bug in structure returning for
    Richard> Sparc?

No.  We just want them to know that there's a problem, and ask for
help, *before* inconveniencing everyone else.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com