When Wget retrieves a file through FTP, it first downloads a .listing
file and parses it for information about the files and other metadata.
Some servers may serve invalid .listing files. This patch checks for one
such known inconsistency wherein multiple lines in a listing file have
the same name. Such a filesystem is clearly not possible and hence we
eliminate duplicate entries here.
Signed-off-by: Darshit Shah <darnir@gmail.com>
Wget was susceptible to a symlink attack which could create arbitrary
files, directories or symbolic links and set their permissions when
retrieving a directory recursively through FTP. This commit changes the
default settings in Wget such that Wget no longer creates local symbolic
links, but rather traverses them and retrieves the pointed-to file in
such a retrieval.
The old behaviour can be attained by passing the --retr-symlinks=no
option to the Wget invokation command.
Add (lots) of documentation for various parts of the test suite in the
form of Python docstrings. Also, clean up some of the redundant code and
fix indentation issues.
The server_conf hook and the server_sett() methods were no longer
required. The server configuration is currently being done by
server_conf() method in the server.
Some command line switches are passed to Wget unconditionally. These
switches should exist in a single place instead of being redundantly
defined in each test file. We add the following two switches by default
here:
1. --debug: This causes wget to be most verbose and output a lot of
debugging information. Hence, if a test fails, the test log should
provide valuable information.
2. --no-config: Users may have their own wgetrc files on the system.
However, for the tests, we want Wget to run with vanilla settings.
Hence, disbale loading any config files.
This is a squashed commit of the following from parallel-wget:
ecd6977 Refactor mainly the test cases classes
d26c8eb Create package test for test case classes
507383d Move server classes to package server.protocol
195393b Create package conf where rules and hooks are put
42e482a Create package exc and move TestFailed to exc
82f44f3 Fix a typo in Test-Proto.py
31e5f33 From WgetTest.py move WgetFile to misc
422171d Create package misc, move ColourTerm.py to misc
Squashed Commit from parallel-wget of:
b31e6e4 Add support for HTTPS Servers
b828a6e Sleep for n seconds before calling Wget Executable
7effa90 Support programatically setting Handler class variables
7e1f4c1 Correct the call to stop_HTTP_Server
f616192 Improve error handling when wget executable isn't available
31868fe Split large function to improve readability and extensibility
Squashed Commit, of the following commits:
7743384 Update documentation to reflect changes in code
b703633 Add feature that allows to ensure that Wget correctly crawls the website in recursive mode
0758f47 Add new test for recursive spider mode
43bb61b Smartly guess content type header
d4d0e63 Support substring replace in File Contents too
f578500 Compatibility fix with multiple servers
8b1a9b6 Extend Functionality to support spawning multiple servers
e84192a Use the provided calls to shutdown server instead of rewriting it
99659f3 Improve Documentation
cb94e52 Slight code cleanup. Remove unused code
886ac1a Shift to new Threading Model from Multiprocessing model
e74c2ec Add new test for POST Requests
48644f1 Print diff when file contents don't match
b6f9efe Add tests for Cookie support
4c9e6b4 Document pending work
e13bc90 Add new test to ensure Content Disposition and Auth work together
60d1f4d Add new Test for Continue command
738b299 Add test, Test-Head
9b9d16b Edit non-unique TEST_NAME variable
ae958db Minor optimizations to the way Server Rules are executed
50b4f0c The rules need not be a defaultdict.
dccc154 Introducing Python based Test Environment
It is done via one-time call OPENSSL_config. It will fix this error
`OpenSSL: error:140920F8:SSL routines:SSL3_GET_SERVER_HELLO:unknown
cipher returned`. OpenSSL with GOST-support and rebuilding wget
required
