Survey existing FTP servers and clients to figure out how LIST works

[exarkun]

Currently we special case a few options in ftp_LIST.  This sucks and we should
do something better.

[adiroiban]

Beside the current code:

        # bug in konqueror
        if path == "-a":
            path = ''
        # bug in gFTP 2.0.15
        if path == "-aL":
            path = ''
        # bug in Nautilus 2.10.0
        if path == "-L":
            path = ''
        # bug in ange-ftp
        if path == "-la":
            path = ''

I have also discovered the following clients:

        # bug in FireFTP
        if path == "-al":
            path = ''

        # bug in BareFTP
        if path == "-La":
            path = ''

Maybe we can use a regex or something simple like:

if path.lower() in ['-a', '-l', '-la', '-al']:
    path = ''

[spiv]

Replying to adiroiban:

Maybe we can use a regex or something simple like: {{{ if path.lower() in ['-a', '-l', '-la', '-al']: path = }}}

I like this approach. Obviously it needs tests etc. I'll make a patch, if I don't fall asleep or otherwise get distracted first :)

[spiv]

(In [34787]) Branch to ftp-list-flags-1333

[therve]

1. Please name the test test_LISTWithBinLsFlags and add a docstring.

2. This needs a NEWS fragment.

Thanks, merge once those points are fixed!

[therve]

#5791 has been closed a duplicate of this bug.

[adiroiban]

I have attached a patch on top of branch 1333

I have renamed the test and add docstring. Also added docstring and renaming for neighbor tests.

It looks like the branch already has a news file.

---

Please switch to git/github. Working with SVN is so slow, Trac code navigation is slow and uploading patched to tickets is not fun.

[tomprince]

(In [37410]) Branching to ftp-list-flags-1333-2.

[tomprince]

(In [37413]) Merge ftp-list-flags-1333-2: Handle bogus options to ftp's LIST command.

Author: spiv, adiroiban Reviewers: therve, tom.prince Fixes: #1333

The twisted.protocols.ftp.FTP server now treats "LIST -La", "LIST -al", and all other combinations of ordering and case of the "-l" and "-a" flags the same: by ignoring them rather than treating them as a pathname.

[tom.prince]

(I accidentally put my review on #5411 by mistake:)

Minor change: Prefer assertIn to assertTrue(... in ...).

I've fixed this up and merged forward.

Will merge once build results are in.