WUI cannot handle non-ASCII node nicknames

[zooko]

I used a non-ascii character in the nickname of my node in tahoe.cfg:

/usr/lib/python2.6/dist-packages/nevow/rend.py, line 777 in string
775
776 def string(context, data):
777    return context.tag.clear()[str(data)]
778
Locals
data	u'Zooko\'s ec2 micro instance named "\u03b12"'
context	WovenContext(tag=Tag(u'td', attributes={u'class': u'nickname mine'}), remembrances={'nevow.inevow.IData': u'Zooko\'s ec2 micro instance named "\u03b12"'})
Globals
string	<function string at 0xa0d95a4>
data	<function data at 0xa0d95dc>

<type 'exceptions.UnicodeEncodeError'>: 'ascii' codec can't encode character u'\u03b1' in position 34: ordinal not in range(128)

[warner]

Sounds like welcome.xhtml line 19 (the n:render="string" n:data="my_nickname" clause) is the trouble. Maybe there's a different kind of n:render= tag that can mean "expect unicode", maybe even n:render="unicode". Or perhaps web/root.py (line 193, in data_my_nickname()) will need to be changed to encode the value that it returns.

[davidsarah]

There probably needs to be a .decode('utf-8') in there.

It should probably just be context.tag.clear()[data].

[warner]

looks trivial, if davidsarah doesn't land this by the weekend, I will.

[davidsarah]

Working on a test now.

[davidsarah]

This was more complicated than it looked; we have to change the n:render type because we cannot change Nevow. The storage status page had the same bug. Also, there was an additional unicode vs. str confusion in NativeStorageServer.get_nickname.

Anyway, it is done: ​https://github.com/tahoe-lafs/tahoe-lafs/pull/25

[amiller]

Review:

1. D-S fixed the bug by Replacing "n:render='string'" with "n:render='data'" in affected xhtml templates, which is a workaround to prevent Nevow from attempting an ascii decoding. D-S had to write a few new mockup structures in test_web.py to simulate the original bug. I ran the tests and looked to see if any other templates were missed.

2. D-S removed a .decode('utf-8') from get_nickname, which is probably correct (since .decode('utf-8') is generally considered useless, doesn't exist in Python3, etc.), but what's the consequence? I put this line back in and all the tests still pass.

[davidsarah]

Re: point 2, I made the system tests exercise the non-ASCII case with the real server code, and now these tests fail without the .decode('utf-8') change. Thanks, that was a big hole in the test coverage.

[davidsarah]

Fixed in 0d4a826, system tests improved in 5c4ca67.