Re: regressions in error message formatting

[Ranjit Mathew <rmathew at gmail dot com>]

Per Bothner wrote:
> I noticed various error messages had reverted to using
> old-style accent grave in place of single quotes.  I.e.
> PR163.java:7: error: final field 'foo' may not have been initialized
> became:
> PR163.java:7: error: final field `foo' may not have been initialized
> 
> This was well-meaningly fixed by Ranjit Mathew on 2004-07-04,
> I assume to be consistent with other error message.  Unfortunately,
> the fix is in the wrong direction:  The old convention was to
> quote variables `thus' but this is ugly in modern fonts and is
> not considered correct use of the accent grave charecter.

Yes, I agree completely - I didn't like the GCC error
quoting convention but I wanted the front-end files to be
consistent in what they show. I now see that my "fix" was
not the right thing at all. :-( ~sigh~


> At some point we'll fix all the error messages, but let's at
> least not make things worse.

Agreed. Sorry for the disruption.


> Furthermore, it appears some error messages were changed
> to start with a capital letter (see the final_assign_error
> function).  I don't believe that is consistent with our
> error message conventions.

Ok, I didn't know about this.


> Since this is a regression, I think we should try to fix
> this for 4.0.0.  If it is easier to fix all of them, rather
> than just the ones that have changed, that is fine IMO.

Agreed again. However, I would need some time (a week to two)
for taking this up as I am a bit too caught up with work
deadlines right now and come back home a bit too late. :-(

If that's unacceptable, may I request a volunteer for this
please?

Shall I continue to consider these changes "obvious"
requiring no explicit approval?

Thanks,
Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT gmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/