cgitb fails when evaluating various reprs/getattrs. The bug
demonstrator is cgitbbug.py. Below is the run output
for two bugs. Fixes are possible by putting try: except
around various evaluations. Although the test is
artificial we found these could happen in real code.

This is with 2.4a3 Windows XP

Sample output

C:\Python24>.\python \tmp\cgitbbug.py repr1
2.4a3 (#56, Sep  2 2004, 20:50:21) [MSC v.1310 32 bit
(Intel)]
Traceback (most recent call last):
  File "\tmp\cgitbbug.py", line 36, in ?
    print cgitb.html(sys.exc_info(),context=3)
  File "C:\python24\lib\cgitb.py", line 155, in html
    dump.append('%s = %s' % (name,
pydoc.html.repr(value)))
  File "C:\python24\lib\pydoc.py", line 352, in repr
    return Repr.repr(self, object)
  File "C:\python24\lib\repr.py", line 24, in repr
    return self.repr1(x, self.maxlevel)
  File "C:\python24\lib\pydoc.py", line 359, in repr1
    return self.escape(cram(stripid(repr(x)),
self.maxother))
  File "\tmp\cgitbbug.py", line 12, in __repr__
    if self._mad=='repr1': raise ValueError('repr1
madness')
ValueError: repr1 madness

C:\Python24>.\python \tmp\cgitbbug.py getattr1
2.4a3 (#56, Sep  2 2004, 20:50:21) [MSC v.1310 32 bit
(Intel)]
Traceback (most recent call last):
  File "\tmp\cgitbbug.py", line 36, in ?
    print cgitb.html(sys.exc_info(),context=3)
  File "C:\python24\lib\cgitb.py", line 129, in html
    vars = scanvars(reader, frame, locals)
  File "C:\python24\lib\cgitb.py", line 83, in scanvars
    value = getattr(parent, token, __UNDEF__)
  File "\tmp\cgitbbug.py", line 16, in __getattr__
    if self._mad=='getattr1': raise
ValueError('getattr1 madness')
ValueError: getattr1 madness

C:\Python24>

Logged In: YES 
user_id=6946

Bug fix added

Logged In: YES 
user_id=31392

The formatting of this patch is wrong.  There is a long line
(>80 chars) and missing whitespace around commas, operators,
etc.

Confirmed on trunk.

Anyone interested in running with this as it's still a problem in all versions?

The output now is

c:\Users\Mark\PythonIssues>cgitbbug.py getattr1
3.4.1 (v3.4.1:c0e311e010fc, May 18 2014, 10:38:22) [MSC v.1600 32 bit (Intel)]
Traceback (most recent call last):
  File "C:\Users\Mark\PythonIssues\cgitbbug.py", line 34, in <module>
    d.h('aaa')
  File "C:\Users\Mark\PythonIssues\cgitbbug.py", line 23, in _f
    raiseError()
  File "C:\Users\Mark\PythonIssues\cgitbbug.py", line 3, in raiseError
    raise ValueError('original error')
ValueError: original error

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\Mark\PythonIssues\cgitbbug.py", line 36, in <module>
    print(cgitb.html(sys.exc_info(),context=3))
  File "c:\python34\lib\cgitb.py", line 136, in html
    vars = scanvars(reader, frame, locals)
  File "c:\python34\lib\cgitb.py", line 88, in scanvars
    value = getattr(parent, token, __UNDEF__)
  File "C:\Users\Mark\PythonIssues\cgitbbug.py", line 16, in __getattr__
    if self._mad=='getattr1': raise ValueError('getattr1 madness')
ValueError: getattr1 madness

Thus the madness has been fixed and this can be closed as "out of date".

I'll repeat the post I made to BreamoreBoy regarding this bug:

"re: http://bugs.python.org/issue1047397

this bug is now 10 years old. I'm not sure why it's to be considered
closed because the original intent of the bug report was that the html
output of the cgitb module could be confused and errors during
evaluation could make the html output fail. So far as I can tell by
modifying the original cgitbbug.py to run under python 3.4.1 it still
fails to correctly produce html when a bug occurs in cgitb.html.
Producing a more reasonable traceback is certainly a step forward, but
I think that has been true since at least 3.1. The whole point of
cgitb is to produce something reasonable for an html consumer."

I have re-created the original 002 patch with mods for python 3.4.1. It's a bit of a shame that this wasn't fixed 10 years ago. Hopefully any style pedants will find this patch more acceptable. Personally I don't think style errors should prevent a working fix from being adopted. I would have just fixed the fix style and carped afterwards.

cgi/cgitb are deprecated as per PEP 594, so there won't be further enhancements to them.