Re: script to automate binary search for regression

[Phil Edwards <phil at jaj dot com>]

Digging this out of the archives...


On Wed, Dec 18, 2002 at 11:44:20AM -0800, Janis Johnson wrote:
> On Mon, Dec 16, 2002 at 09:43:41PM -0500, Phil Edwards wrote:
> > On Mon, Dec 16, 2002 at 09:23:50AM -0800, Janis Johnson wrote:
> > > The length of the script is mostly due to comments and lots of error
> > > checking.  I've tested it fairly thoroughly.  If you find this helpful,
> > > let me know and I'll ask about adding it to contrib.
> > 
> > Yes, please.  This is very useful.  We should probably also check in an
> > initial "config file", and some simple skeleton scripts for the cvs calls.
> > And yes, by "we" I mean I'll help once this one's checked in.  :-)
> > 
> > > #! /bin/sh
> > 
> > Since the "let" builtin is a POSIX thing, and it's been made woefully clear
> > that GCC can't depend on /bin/sh actually being POSIX, this script should
> > either a) document that fact, or b) just use a different shell.
> 
> I don't know if Phil has authority to say this can go into contrib, but

I don't know either.


> here's the latest version of my framework for hunting regressions, with
> a few more nifty ways to cut down a search.  It works for bash, but
> probably not other shells, and depends on an extension in GNU date.
> 
> This script doesn't do the real work; that's done by scripts the user
> provides to get the sources, do the build, and run the test, which will
> vary greatly depending on the kind of bug being hunted.  I can put
> examples of those somewhere.
> 
> May I add this to contrib?
> 
> Janis Johnson
> IBM Linux Technology Center, OzLabs North
> 
> --- /tmp/empty	Wed Dec 18 11:32:07 2002
> +++ contrib/reg_search	Wed Dec 18 11:30:32 2002
> @@ -0,0 +1,273 @@
> +#! /bin/bash
> +
[...]

How about making a contrib/reg_search directory, checking the script in
there, along with some sample config files?

Any objections from anyone?


Phil

-- 
I would therefore like to posit that computing's central challenge, viz. "How
not to make a mess of it," has /not/ been met.
                                                 - Edsger Dijkstra, 1930-2002