From mboxrd@z Thu Jan  1 00:00:00 1970
Date: Fri, 22 Jul 2011 19:53:30 +0200
From: Lucio De Re <lucio@proxima.alt.za>
To: Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net>
Message-ID: <20110722175330.GB2065@fangle.proxima.alt.za>
References: <2fb2c31fa3cb67d0f3fe70f628f24bf6@ladd.quanstro.net>
	<8e8648e7029cff027f2fbfc4920762ae@quintile.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <8e8648e7029cff027f2fbfc4920762ae@quintile.net>
User-Agent: Mutt/1.5.20 (2009-06-14)
Subject: Re: [9fans] Plan 9 Go (Was: GNU/Linux/Plan 9 disto)
Topicbox-Message-UUID: 04cbd7be-ead7-11e9-9d60-3106f5b1d025

On Fri, Jul 22, 2011 at 05:35:24PM +0100, Steve Simon wrote:
>
> My understanding is that bison can be convinced to output some static
> tables which the Go authors munge into some C source (with awk).
>
> These tables allow the Go parser to generate more useful and
> accurate error messages.
>
That matches my recollections from some superficial inspection a few days
ago.  Thanks for bringing these memories back and clarifying them for me.

> I believe using bison is not mandatory if you don't mind hacking
> the code a little, but you won't get the nicer error handling.
>
That suggests that porting Bison is a "must" as anything else will
not meet the criterion of being included in the Go distribution.
Bootstrapping Bison seems to be my main obstacle (it relies on Bison
features, not least the very same %error-verbose I have encountered in
Go), but it can be overcome by using cross-processing from either NetBSD
or UBUNTU, so that's just some effort on my part.

> This is educated guesswork from looking at the code a few months ago,
> please don't treat it as gospel.
>
Pretty good analysis, if my memory is anything to go by.

++L