I need some help interpreting this error

[Stestagg]

I don't particularly like to encourage this shotgun help request because,
as previous commenter suggests, debugging this yourself is best.

Sometimes debugging is super hard, and especially so when uncommon
situations occur, but it's always far easier to debug things when you have
visibility into the system under test.

However, in this case, the email code is super complex, and this scenario
also looks very uncommon, but not unique: (
https://github.com/Sydius/mbox-to-txt/issues/2), so you successfully
nerd-sniped me :).

My *guess*, from reading the python standard library source code is that
you came across an email with some content in the subject line that is
considered a "surrogate", roughly, some badly encoded unicode or binary
data in it.

When this happens, the code in some situations (depending on the policy...)
may return a header.Header() instance, rather than a
headerregistry.UniqueUnstructuredHeader
(which would have had a headerregistry.BaseHeader (mro: str) dynamically
attached).

header.Header() does not inherit from str, and thus would throw the
traceback you observed.

Your suggestion of a try: catch: may make sense, alternately, you could
wrap the result in a call to str():

if sbstrip in str(msghdr["subject"]):

which should attempt to encode the binary into some form of string object
for comparison (I haven't checked exactly what would happen, beyond: it
tries).

It should be possible to create a test mbox with some funky bytes in the
subject, and try to reproduce it that way.

Steve


On Wed, Feb 17, 2021 at 5:07 PM Chris Green <cl at isbd.net> wrote:

> Stefan Ram <ram at zedat.fu-berlin.de> wrote:
> > Chris Green <cl at isbd.net> writes:
> > >But msghdr["subject"] is surely just a string isn't it?  Why is it
> > >complaining about something of type 'Header'?
> >
> >   What would you do to debug-print the type of an object?
> >
> I don't know, what would I do?  :-)
>
> Without knowing what provokes the problem I could be waiting for days
> or weeks even before I see the error again.  As I'd need to print the
> type for every message I'd get some big logs or I'd need to add a try:
> except to trap the specific error.
>
> --
> Chris Green
> ·
> --
> https://mail.python.org/mailman/listinfo/python-list
>