On 2 August 2012 13:08, Raphael Hertzog <hert...@debian.org> wrote: > On Thu, 02 Aug 2012, Raphael Hertzog wrote: > > On Thu, 02 Aug 2012, Raphael Hertzog wrote: > > > If the problem is indeed not in Django, then we can temporarily disable > > > the test suite and upload the package. > > > > I just tried to build the current python-django 1.4-1 in Debian Sid and > it > > also failed. So this tends to confirm that something else broke Django's > > test suite (since the test suite worked when I uploaded 1.4-1 to sid). > > > > Now we need to find the culprit (and fix it or work-around it). > > Apparently the build works fine in wheezy so I have built it in wheezy and > uploaded it in sid. > > sid and wheezy differ in their python2.7 version: 2.7.3-2 vs 2.7.3~rc2-2.1 > > So the regression might be between those two versions. > > Doko, python-django test suite fails in sid but not in wheezy. The failure > looks like a HTMLParser regression. Do you know of any possible regression > in python 2.7.3 about this? > > I see in the upstream changelog a “- Issue #14538: HTMLParser can now > parse correctly start tags that contain a bare '/'.” maybe this could be > related? > > I also found https://code.djangoproject.com/ticket/18239 which might imply > that Django is relying on some internals of HTMLParser so it would > actually be a bug in Django in that case... > > I was just looking into this and these are the changes in HTMLParser.py
diff ../HTMLParser.py-old ../HTMLParser.py 25c25 < tagfind = re.compile('[a-zA-Z][-.a-zA-Z0-9:_]*') --- > tagfind = re.compile('([a-zA-Z][-.a-zA-Z0-9:_]*)(?:\s|/(?!>))*') 31c31 < r'[\s/]*((?<=[\'"\s/])[^\s/>][^\s/=>]*)(\s*=+\s*' --- > r'((?<=[\'"\s/])[^\s/>][^\s/=>]*)(\s*=+\s*' 292c292 < self.lasttag = tag = rawdata[i+1:k].lower() --- > self.lasttag = tag = match.group(1).lower() -- David Watson dwat...@debian.org