Re: GTK+ 2.10.7 released

[Yevgen Muntyan <muntyan tamu edu>]

Owen Taylor wrote:

> On Sat, 2007-01-06 at 18:31 -0600, Yevgen Muntyan wrote:
>  
>
> > Behdad Esfahbod wrote:
> >
> >    
> >
> > > On Sat, 2007-01-06 at 00:24 -0800, Sergei Steshenko wrote:
> > >
> > >
> > >      
> > >
> > > > [really impressive list of bugs deleted].
> > > >
> > > > As an end gtk+ user I just want to know which of the two
> > > > gtk+2.8.20 <-> 2.10.7 is less buggy.
> > > >   
> > > >
> > > >        
> > > Indeed 2.10.7.  The impressive list of bugs you saw has only been fixed
> > > in 2.10.7.  Another set fixed in 2.10.6.  Same for 2.10.5, ...  None of
> > > them fixed in 2.8.x.
> > >
> > >
> > >      
> > There is a natural question, how many of those bugs didn't even
> > exist in 2.8.x?
> >    
>
> If you want to do that research, then well, you have the list of bug
> numbers. But note that what you ask above "how many of these bugs didn't
> even exist in 2.8.x" isn't really interesting. The interesting question
> is *how many of these bugs represent regressions from 2.8.x*.
Yes, it is an interesting question. But I replied to "The impressive list
of bugs you saw has only been fixed in 2.10.7... None of them fixed in
2.8.x.". Number of bugs fixed in gtk-2.10 does not tell anything about
how much buggier gtk-2.8 is, you need to actually look at the bugs.
Then, "buggier" means (at least I read it this way) "more bugs". Not
clear what it means, to have "more bugs", of course; but it is clear
that non-fixed regressions make gtk-2.10 more buggier. And so on.

> > > This idea of thinking about newer series as buggier versions is
> > > fundamentally bogus.  The way you should be thinking about 2.10 is 2.8
> > > plus some new functionality plus lots of bug fixes.  It typically is the
> > > case that if you don't use the new functionality, it's just added value,
> > > not many bugs introduced, etc.
> > >
> > >      
> > Right, you *should* think that way, or GTK *should* be that way
> > but it's not quite so. Take infamous GtkFileChooser thing - more
> > bugs, less usable.
> >    
>
> The old file selector (GtkFileSelector) was not buggier in GTK+-2.4
> after GtkFileChooser was introduced. In fact, I'm pretty sure I fixed a
> number of GtkFileSelector bugs in the GTK+-2.4 development cycle.
>  
Um, GtkFileChooser is an example of what's broken in Gtk-2.10
comparing to Gtk-2.8. More bugs, less usable than in Gtk-2.8.

> > I am not claiming that Gtk-2.10 is worse than Gtk-2.8, my opinion
> > about it is very subjective, I didn't count bugs or something; some
> > things are certainly better in 2.10 (e.g. printing is sort of there).
> > But Sergei's question is completely reasonable. You know about
> > that debian thing, that folks do not (or did not) want to use Gtk-2.10
> > because it's buggy, don't you? 
> >    
>
> I have no clue what you are talking about. (Though maybe people who are
> actually active GTK+ developers currently do.) But the fact that some
> people hit some bug with some GTK+-2.10 release that affected what they
> were doing has basically zero bearing onto comparing the current state
> of the (unmaintained) GTK+-2.8.x branch and the (maintained) GTK+-2.10.x
> branch.
>  
It's not about comparing state of this and that, it's the question
"If we upgrade, do we get broken desktop?" Crash in Open dialog
is bad enough for some people. Here's what I'm talking about:
http://oskuro.net/blog/freesoftware/gnome-2.16-etch-2006-10-06-21-45
http://lists.debian.org/debian-gtk-gnome/2006/11/msg00021.html

> > It's not just like someone says "You
> > guys suck, 2.10 is broken and is no better than 2.8". There are things
> > that make some people believe something is wrong with Gtk-2.10, and
> > you can't simply say those things are bogus because version number
> > is bigger.
> >
> > I understand that Sergei's question didn't make any sense on
> > any gtk list, because it's taken as "You suck" (and maybe it's really
> > what he meant too). But then it should be simply ignored, answers
> > like "No, 2.10 is better because it must be better" do no good and
> > only may add to one's belief that GTK developers ignore problems
> > or do not admit problems do exist (no, I am not thinking this way).
> >    
>
> Reread Sergei's original question ... I don't know if there was a hidden
> agenda to it, but it looks simple enough, and I think Behdad's answer
> was polite and pretty much spot on.
I read only what Behdad quoted, I am not subscribed to whatever
list Sergei's letter was posted to. So I read the following:

"Is Gtk-2.10 more buggy than Gtk-2.8?"
"No, because there are bugs fixed in Gtk-2.10"

As it is, it's simply wrong.

> If I was answering I'd probably have
> been even more terse and said:
>
> - GTK+-2.10 is a stable branch
> - GTK+-2.10 is the only currently maintained stable branch
> - The GTK+ developers have no reason to think it's buggier or more
>   memory leak prone than 2.8.x
>  
This would actually be better, because it's clear and you can't
argue about it :)

Best regards,
Yevgen