gwui.py:260:render:TypeError: WebKitWebView.load_html_string() argument 1 must be string without null bytes, not str

[Tom "spot" Callaway]

Fedora has gotten a few bugs filed with the below traceback:

gwui.py:260:render:TypeError: WebKitWebView.load_html_string() argument 1 must be string without null bytes, not str

Traceback (most recent call last):
  File "/usr/lib/python2.6/site-packages/gwibber/gwui.py", line 376, in on_stream_change
    self.update()
  File "/usr/lib/python2.6/site-packages/gwibber/gwui.py", line 391, in update
    self.messages.update(self.navigation.selected_stream)
  File "/usr/lib/python2.6/site-packages/gwibber/gwui.py", line 723, in update
    self.messages = self.message_view.render([selected_stream], count)
  File "/usr/lib/python2.6/site-packages/gwibber/gwui.py", line 781, in render
    accounts=accounts)
  File "/usr/lib/python2.6/site-packages/gwibber/gwui.py", line 260, in render
    self.load_html_string(content, "file://%s/" % os.path.dirname(template_path))
TypeError: WebKitWebView.load_html_string() argument 1 must be string without null bytes, not str

Thanks to the help of Toshio Kuratomi, I was able to narrow this down to a simple test case:

"""CODE STARTS"""
import os, webkit
template_path = '/usr/share/gwibber/ui/themes/default/template.mako'
content = 'foo \x00'

web = webkit.WebView()
web.load_html_string(content, "file://%s/" % os.path.dirname(template_path))
"""CODE ENDS"""

Somehow, in certain situations, \x00 ends up in the content string, and load_html_string chokes on it. Since content contains all of the processed messages which are trying to be displayed (but is used after this load_html_string() call), the safest bet seemed to be to simply replace it with a " " whenever it is found in content. This resolves the test case, and would seem to resolve the gwibber crashes as well. I've attached a patch which should stop these crashes from occurring.

[Tom "spot" Callaway (tcallawa)]

Full debugging info available here:

https://bugzilla.redhat.com/show_bug.cgi?id=630944
https://bugzilla.redhat.com/show_bug.cgi?id=620170

[David Klasinc (bigwhale)]

This still exist in current trunk and I was hit by it today. Proposed patch will fix the error. I modified it a little bit and also added some code that will prevent \x00 characters being in the database in the first place.

[Victor Vargas (kamus)]

Ken, would you mind to review and check this patch please :)

[ahmadzxc (ahmad-almomani5)]

thanks all
but i don’t know how to apply these patches any help please !!!!

[Ken VanDine (ken-vandine)]

The handling of the null bytes was fixed in the service at some point in the 3.2 series.