[tahoe-lafs-trac-stream] [Tahoe-LAFS] #3874: Incorrect encode in allmydata.testing.web._FakeTahoeUriHandler.render_GET

[Tahoe-LAFS]

#3874: Incorrect encode in allmydata.testing.web._FakeTahoeUriHandler.render_GET
---------------------+---------------------------
 Reporter:  exarkun  |          Owner:
     Type:  defect   |         Status:  new
 Priority:  normal   |      Milestone:  undecided
Component:  unknown  |        Version:  n/a
 Keywords:           |  Launchpad Bug:
---------------------+---------------------------
 `render_GET` does:

 {{{
         for arg, value in uri.query:
             if arg == u"uri":
                 capability = value
         # it's legal to use the form "/uri/<capability>"
         if capability is None and request.postpath and
 request.postpath[0]:
             capability = request.postpath[0]

         # if we don't yet have a capability, that's an error
         if capability is None:
             request.setResponseCode(http.BAD_REQUEST)
             return b"GET /uri requires uri="

         # the user gave us a capability; if our Grid doesn't have any
         # data for it, that's an error.
         capability = capability.encode('ascii')
 }}}

 So it reads `capability` from either a DecodedURL's `query` attribute or
 from `request.postpath[0]`.  `DecodedURL.query` is certainly a str.

 `request.postpath` comes from `twisted.web.server.Request.process`:

 {{{
     self.postpath = list(map(unquote, self.path[1:].split(b'/')))
 }}}


 `self.path` must be `bytes` since it is being split with `bytes` but it
 comes from `twisted.web.http.Request.requestReceived`:

 {{{
         x = self.uri.split(b'?', 1)

         if len(x) == 1:
             self.path = self.uri
         else:
             self.path, argstring = x
             self.args = parse_qs(argstring, 1)
 }}}

 `requestReceived` also documents `path` as `bytes` and then derives
 `self.uri` from it so ... probably bytes.

 Thus, in Tahoe, `capability` is either str or bytes depending on which
 codepath finds it.  And if it is bytes then `.encode('ascii')` fails.

 which seems to confirm that it is bytes.

--
Ticket URL: <https://tahoe-lafs.org/trac/tahoe-lafs/ticket/3874>
Tahoe-LAFS <https://Tahoe-LAFS.org>
secure decentralized storage