On Sat, 2009-02-07 at 11:44 +0000, Michael Drake wrote:
> I think the main things blocking NetSurf 2.0 release are:
> 
>  1.  NetSurf doesn't work at all on most modern linux distros. The
>      throbber throbs but pages don't load. I believe it was discovered
>      to be due to some libcurl / openssl thing.
> 
>      Joty has reported it on Ubuntu 8.10 and so has someone else reported
>      it for Ubuntu 8.10 in #netsurf recently.
> 
>      We've had reports of the same issue from other distro users like
>      Fedora for quite a long time.

This is most likely because we ask for the SSL context. If the
underlying libcurl is not built for openssl support, then it'll just
return an error, thus preventing the fetch occurring.

The correct solution to this problem (as noted on TODO-jmb) is to remove
all the openssl-specific nonsense in our own code and use libcurl's new
(>=7.19.1 iirc) certificate chain inspection APIs when certificate
verification fails. For a fully comprehensive solution, libcurl will
need to grow support for these APIs when it is built against <>openssl.

[GTK frontend foo]

I'm mildly dubious about Accept-Language UI actually being a blocker.
Still, it's just a text field in some configuration dialogue, so
shouldn't be hard to solve.

>  4.  Textarea / text input / form submission issues.

These should be treated separately.

I would appreciate it if someone could actually investigate the supposed
form submission issues (and provide some standalone test cases for
them). The last one I looked at had nothing to do with the data we
submitted at all.

Somewhere on test.ns-b.org is a script that will echo POST data back to
you, so you can test stuff sensibly. That site's protected using basic
auth. Shout if you want (reminding of) the access details.

>  5.  Get Hubbub up to latest spec. and tested.

Andi, have you some time to give to this? I'm not sure I'm going to get
to it any time soon.

>  6.  Maybe the new CSS library.

Once the selection engine is finished we'll have a better idea of where
we stand. There's still a bunch of outstanding functionality, but that
can most likely be dealt with at the hackfest.

The selection engine needs a complete set of tests before I'm willing to
think about making any kind of libcss release. Note that any release of
libcss will be a 0.x.y, as the API's not guaranteed to be stable
initially.

> I think that's everything for the Core and the GTK front end. I don't
> think there is any problem releasing the RISC OS front end stuff in its
> current state. Not sure about BeOS & AmigaOS front ends.

There is one RO-frontend related crash on the tracker. That must be
fixed before release.

Talking of the tracker, it desperately needs triaging -- there's a whole
bunch of reports for issues which have been fixed or have since
disappeared. If someone can find some time to deal with this, that'd be
great.


J.


Reply via email to