On 15.04.2015 23:20, Ben Reser wrote:
> On 4/15/15 8:11 AM, Julian Foad wrote:
>> On 9 April 2015, Stefan Fuhrmann wrote:
>>> As of now, there are 2 issues scheduled for 1.9.0:
>>> http://subversion.tigris.org/issues/show_bug.cgi?id=4560
>>> http://subversion.tigris.org/issues/show_bug.cgi?id=4565
>> #4565 is now closed.
> The blockers in the STATUS file are now merged. This leaves Issue #4560 and
> Bert's 2 changes that can't go into 1.9.1 or later. Bert's changes look like
> there's a significant possibility they will be approved soon. I was somewhat
> hoping to roll the RC1 today but I think I'll wait a bit longer for those two
> changes.
>
> However, I think we should just punt on Issue #4560. It looks to me like
> either a documentation bug or an implementation bug. The contents of a hash
> are not something that I'm inclined to believe should hold up a release. So
> if
> someone knowledge on this issue can chime in and give some opinions it would
> be
> appreciated.
I'll talk about this with stsp and the rest of the gang here in Austin
today, thanks for the reminder.
> The other issue that exists but isn't really documented in STATUS or the issue
> tracker right now is the issue with SWIG 3.0.3+ and the Python bindings. We
> discussed this on IRC today. We have a few options available to us. Make
> configure refuse to work with SWIG 3.0.3+, putting some somewhat ugly changes
> into the bindings to deal with the problem, or doing nothing and telling
> people
> to use the generated sources included with the tarball if they have a newer
> SWIG. My inclination is not to hold up RC1 for this. It is something we can
> put in subsequent RCs if we have them or subsequent 1.9.x releases.
There are actually two issues with Swig 3.0.3+:
* The first is the whitespace-eating bug that converts comments to
invalid code; we know we can work around that to make the code
compile with both Swig 2 and Swig 3
* The second problem has to do with default argument handling in Swig
and I've not yet found a way to make the Python bindings generated
with Swig 3 actually work, even when they compile.
My recommendation is to not support Swig 3 at all for now.
-- Brane
