Re: Pango-1.11.99 "Just In Case" released [unstable]

["Matthias Clasen" <matthias clasen gmail com>]

On 2/27/06, Behdad Esfahbod <behdad cs toronto edu> wrote:
> On Mon, 27 Feb 2006, Federico Mena Quintero wrote:
>
> > On Sun, 2006-02-26 at 21:47 -0500, Behdad Esfahbod wrote:
> > > * Insert '?' chars instead of invalid UTF-8 sequences in
> > >   pango_layout_set_text [#33195]
> >
> > A long, long time ago we mande the decision to assume that
> > pango_layout_set_text() gets a valid UTF-8 string.  Doing otherwise
> > would kill performance.  This requirement is even documented in that
> > function's docs.
>
> You sure?   Until just before 0.11.6 it was validating the string
> all the time.  And none of us noticed anything in performance
> analysis, right?
>
> > Can't we add something to the docs instead, along the
> > lines of "if you crash inside this function, you passed an invalid UTF-8
> > sequence, doofus!".
>
> Do you really think people read docs when they see crashes?  What
> Owen suggested and I agree is, if we don't see that validation in
> any profiles, let it be.
>

Also note that we spent some time last year optimizing g_utf8_validate()