Re: GTK adjustement changes create incompatible behaviour between versions?

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

On Mon, Sep 22, 2008 at 12:41 PM, Patryk Zawadzki <patrys pld-linux org> wrote:
> 2008/9/22 Josselin Mouette <joss debian org>:
>> The difference is that you are knowingly breaking hundreds of
>> applications. You are able to fix those distributed by GNOME in a snap
>> and that's all good, but you let downstream distributors with hundreds
>> of broken applications in their hands. The only solution we have is to
>> revert this change, because we simply can't afford it.
>
> Not true, we can avoid all the breakage inside GtkSpinButton.
>
>> Even if we fix all our packages (and that would take a few weeks), the
>> problem of non-free applications remains. GTK+ is not just a random
>> library that you can afford to break a each and every release, it is one
>> of the more widely used toolkits. If the wording of the documentation
>> was not clear enough, that's bad, but it's too late, you can't break it
>> now.
>>
>> Now for more constructive discussion:
>>      * would it work if GtkSpinButtons (and maybe GtkScale) forced
>>        page_size to be 0 in the constructors and in ..._set_adjustment?
>
> That's what I proposed. If it's != 0, force it to 0 and issue a
> warning. Spin controls have no concept of data set so there is no such
> thing as a page and therefore there's no use for limiting the offset
> to a certain subrange of (min, max). It's a two-three line change for
> GTK and was already worked around in libglade (patch posted earlier in
> this thread).
>

That seems like a reasonable proposal, assuming that most of the
breakage is in spin buttons.