2013/6/19 Sven Van Caekenberghe <sven@stfx.eu>

On 19 Jun 2013, at 22:55, Nicolas Cellier <nicolas.cellier.aka.nice@gmail.com> wrote:

> Yes, this part is perfectly correct.
> It's just that I used ZnInvalidUTF8 in TextConverter which introduces a dependency of TextConverter on Zinc.

OK, but technically, it is a dependency on Zinc-Character-Encoding-Core, which is independent of the Zinc HTTP code itself. Just like Zinc-Resource-Meta-Core (containing ZnUrl and ZnMimeType) is independent of the HTTP code proper. These sub packages where created specifically to break unwanted dependencies, as requested by Pavel et al.

> This only make sense if in the long term we replace those TextConverters by cleaner Zinc encoders/decoders.

Yes, that is the idea, eventually. But never in 2.0.

> 2013/6/19 Sven Van Caekenberghe <sven@stfx.eu>
>
> On 19 Jun 2013, at 22:33, Nicolas Cellier <nicolas.cellier.aka.nice@gmail.com> wrote:
>
> > The change is very light, I applied it almost unchanged on Squeak (can't remember which version was the first).
> > The only thing we neeed is to catch an UTF8Error.
> >
> > Since in Pharo there are two encoder/decoder hierarchies, and since that of Zinc is cleaner, I made a bet on future and chose to make UTF8Error a subclass of existing already Zn Encoding/Decoding exception.
> >
> > But we still use the old TextConverter hierarchy in MC, even in Pharo 3.0.
> > So the dependency on Zinc is totally arbitrary and un-necessary right now; it does not exist in Smalltalk version.
> > It's just a bet on the future.
> > For a backport to 2.0, I would just use an UTF8Error like in Squeak and avoid any dependency on Zinc.
>
> Well ZnInvalidUTF8 is now a full part of Zinc itself (Zinc-Character-Encoding-Core to be exact). It is already in 3.0 (with your code) and I guess we'll upgrade Zinc in 2.0 as well. But is is a dependency, that is correct.
>
> Anyway, we have to double check.
>
> > Nicolas
> >
> >
> > 2013/6/19 GOUBIER Thierry <thierry.goubier@cea.fr>
> > St�phane,
> >
> > I'll probably have a look at Nicolas fix anyway, but as it required also a change to Zinc... then I started to worry. But it may be better than trying to fix ZipArchive incorrect encoding issues from the outside (by forcing an utf8 conversion out of the contents of the ZipArchive members).
> >
> > If I manage to make sense of it, I'll put a enhancement request in FogBuz with a slice.
> >
> > Yes, I think 30 is not that unstable, like 2.0 was before ... except that I'm fairly dependent on the RPackage infrastructure, and I prefer to wait until the RPackage refactoring is done ;-)
> >
> > Thierry
> >
> > ________________________________________
> > De : Pharo-dev [pharo-dev-bounces@lists.pharo.org] de la part de St�phane Ducasse [stephane.ducasse@inria.fr]
> > Date d'envoi : mercredi 19 juin 2013 21:21
> > � : Pharo Development List
> > Objet : Re: [Pharo-dev] you may need to update your configurations
> >
> > >
> > >
> > >> Christophe.
> > >>
> > >> [1] https://pharo.fogbugz.com/default.asp?10801
> > >
> > > Ouch. Does not bode too well for backporting that to Pharo 2.0.
> > >
> > > I'm feeling a bit down. It's either enduring bugs in 2.0 for a year or so or fighting every few days with 3.0 and knowing that anyway it's not production ready. I had to cope with 1.4 not being able to handle utf8 in the same way.
> >
> > Thierry what we can do is the following:
> > have a look at the fix of nicolas and we can try to add to th 2.0 batch but we should pay attention
> > not to introduce other bugs (because I'm afraid it will).
> >
> > BTW 30 is not that instable.
> >
> >
> > > The window for my next serious developpement with Pharo is around 2014, so I guess I could just sit and wait.
> > >
> > > By the way, why the change in MonticelloFileTree-Core-ChristopheDemarey.97 breaks Pharo 2.0 ?
> >
> >
> >
>
>
>