Re: [patch] serious problem in pango_layout_set_text()

[Owen Taylor <otaylor redhat com>]

Sven Neumann <sven gimp org> writes:

> Hi,
> 
> Havoc Pennington <hp redhat com> writes:
> 
> > Sven Neumann <sven gimp org> writes:
> > > 
> > > attached is a patch that fixes a problem with pango_layout_set_text():
> > > If an invalid UTF8 string is set, this function used to spit out a 
> > > warning and returned in an inconsistent state (layout->text g_free'd)
> > > which causes the function to crash when being called the next time.
> > > 
> > > The patch adds a gboolean return value that indicates success of the
> > > operation to give the application programmer a chance to react to this
> > > problem (for example by displaying "[ Invalid UTF8 string ]" instead).
> > 
> > All Pango and GTK functions require valid UTF-8; we aren't going to
> > require error-checking on all of them. Invalid UTF-8 is considered a
> > g_warning() or g_return_if_fail() situation.
> > 
> > Instead, you should g_utf8_validate() all text on its way into your
> > app (e.g. when loading a file), and then assume all text in the app is
> > valid UTF-8.
> 
> I've expected you or Owen would say this. Why are you validating the
> string in pango_layout_set_text() then? If you decide not to add the
> return value indicating success

The primary reason for validating here is that if things explode somewhere
deep in the guts of PangoLayout, the programmer has very little chance of
figuring out that the mistake was invalid UTF-8.

> please at least return the PangoLayout object in a sane state when
> bailing out with an error.

See my other mail.

                                        Owen