On Tue, 2010-09-14 at 18:50 -0600, Shane Hathaway wrote:
> On 09/14/2010 05:18 PM, Ian Bicking wrote:
> > WebOb 1.0!
> > This doesn't represent any particular development event, simply an
> > acknowledgement that WebOb is well past what would be called "1.0" --
> > stable, complete, with a well-defined scope, and as a dependency for
> > many other projects its API is and should be very stable.
> > Thanks to everyone who helped, and in particular to Sergey Schetinin for
> > his ongoing help with many a bug fix.
> That's great news!
> I wanted to celebrate by upgrading a project to WebOb 1.0, but
> unfortunately, the new version broke some tests. There is a bug in
> either WebOb 1.0 or BFG 1.2.1. The BFG traverse() function now causes
> this exception:
> DeprecationWarning: The class attr [default_]charset is deprecated
> This is raised by request.py. WebOb 0.9.8 used the warnings module to
> issue this warning, while WebOb 1.0 raises the warning as an exception.
> I think WebOb needs to go back to using the warnings module, since I
> don't think raising a warning as an exception is correct.
> Here is the relevant change:
> We could alternatively say that BFG needs to remove the default_charset
> attribute from its request object, but if that's the case, then WebOb
> should issue a deprecation error, not a warning.
I just fixed this in BFG 1.3a14. It might be possible to do a 1.2.X dot
release that worked with WebOb 1.0, although with WebOb 1.0 there's no
way to support code that relied on request.GET, request.POST, and
request.params returning bytestrings when request.charset was None (some
very old code, written pre-BFG-0.7.0 relies on this behavior). It also
implies that the "repoze.bfg.request.make_request_ascii" subscriber
function would need to be removed because as far as I can tell there's
no way to do that. We alternately pin a new 1.2.X release to WebOb <=
0.9.9 if anyone had a problem with that.
Whatever we decide to do, these are all nits, so thanks for WebOb, Ian!
It helps a lot to not need to think regularly very hard about the job it
Repoze-dev mailing list