Re: Small update to reversed_comparison_code

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

    Really I hadn't planned to join this debate but I'm a bit annoyed -- more
    than a bit, actually -- by your assertion that the last couple weeks' pain
    was "the only way" to find latent bugs.  

That's not the way I read the traffic.

The Sparc bootstrap was not broken "for the last couple of weeks".  It had
a miscompare that came and went for a while ever since a change done by
RTH into the mainline and branch on February 8.

But it wasn't until quite recently (within the last week) that this
was noticed.

    Break all the bootstraps you want in your private trees by sharing
    patches on the mailing list amongst those people working on the
    problem,

To say that change "broke" the bootstrap misses the point entirely.

    "break something and leave it broken" mindset rules GCC.

I don't think anybody proposed that, and certainly not me.

The idea is *not* to leave it broken.  We have a serious latent bug in
reload_cse which is now actively being worked on.  Had RTH's change
been rejected by an automatic daemon, *then* we would have been in a
situation where we were leaving something broken.  This way, we are
working on fixing it.