Re: [PATCH] use "unresolved reference" message

[Joel E. Denny]

Hi Alex.

On Thu, 10 Sep 2009, Alex Rozenman wrote:

> I would like to install the following patch.

It looks ok, but I didn't see a ChangeLog or git log entry.  Also, I 
noticed your last patch didn't format the git log entry correctly.  The 
first line should fit nicely into the Subject line of an email message (if 
you avoid line wrapping in a 72-column ChangeLog entry including the 
leading tab, you should be completely fine), and you should skip a line 
afterward.  This makes the output of commands like 'git format-patch' and 
'git branch -v' much more legible.

> The patch adds additional
> sub-messages in case of "invalid references" errors. Some issues in the
> implementation:
> 
> 1. I used "printf("%.*s", len, ptr); Is this stuff portable ? I suppose it's
> quite old, anyway it looks tricky. Isn't strndup better ?

Off the top of my head, I don't know the answer.  However, I don't 
actually see that in your code.  Where should I look?

> 2. As you can see, I used the whole rule location (rule->location) for the
> extra messages, that caused an ugly shift for multi-line rules (see in
> test-suite). Maybe we really need to keep sub-messages aligned ?

I'd be fine with that.

> 3. In the NEWS file (symbolic/named references) . I googled:
> "symbolic references" -> ~41,000

How many of those are in the context of parser generators and grammars, 
where "symbol" has another meaning?

> "named references" -> ~14,000 (including copies of our thread).

Here in the US, I get completely different results from google.

"named references" -> 117,000 (with quotes) or 27,300,000 (without quotes)

"symbolic references" -> 48,300 (with quotes) or 826,000 (without quotes)

But these are very generic terms, so maybe we should be looking for 
precedent in the context of parser generators and grammars.