Re: Re-revised patch for numbered format specifiers in diagnostics

[Zack Weinberg <zack at codesourcery dot com>]

Zack Weinberg <zack@codesourcery.com> writes:

> [Sorry, I seem to have lost your message.]
>
> I took your revised patch, the one that used strings as an intermediate form,
> and revised it again.  It now does not need to copy a scratch pretty_printer
> structure over the real one, nor does it need to parse the format string three
> times.  I am still not totally happy with it, but I don't think we can do
> better without rewriting pretty-print.c and all its "subclass"ers, which I am
> not motivated to do.  What do you think?
>
> This gets no regressions on i686-linux (c, c++, objc, objc++).  Also, I
> hand-tested a couple of diagnostics that use %n$ in the Turkish translation;
> they work.

Jakub not having commented, I am checking this in unilaterally.  It
may not be perfect, but it does fix this long-standing bug to my
satisfaction.

zw