Hook on to CheckUserInsertPrivateEventRow and CheckUserInsertLogEventRow
to override the IP, XFF and User-Agent string when the user is the
abuse filter user for log events.
These two hooks are being added as log entries are being removed from
cu_changes and added into two new tables. Because the columns and their
names are different for these tables, reusing the same hook won't work
for callers that rely on setting values for a specific column name.
Edits and log entries performed by the abuse filter user need to be
marked as being by the software (and not using the IP, XFF and
User-Agent provided in the main request).
These hooks will not be run until the appropriate config is set to
write to the two new tables. Until that point using the one currently
defined hook will work for all actions.
Bug: T324907
Bug: T44345
Depends-On: I7c7754323ade9a8d96273c1742f30b1b5fbe5828
Follow-Up: Idd77545af94f9f9930d9ff38ab6423a72e680df9
Change-Id: Id78417e9d95220946f110afbe1430df5b3bb4f4f
The motivation is to have a single immutable object providing
information about the action. It can represent the current
action being filtered, but also a past action stored in the
abuse log. It will hopefully help us get rid of passing
User(Identity) and Title/LinkTarget objects around together.
Change-Id: I52fa3a7ea14c98d33607d4260acfed3d3ba60f65
For fixing bugs like T65632, T105325, or T264104, we will need
to update code in more than one place at once. To prevent
regressions, create an integration test which tests the whole
pipeline, from the request submission to variable evaluation.
Edits are simulated using action=edit API call because the hook
AbuseFilter uses is run from EditPage.
To increase confidence in test coverage, remove some annotations
from AbuseFilterConsequencesTest or make them less greedy.
Ideally, it would only test consequences.
This patch includes refactoring of AbuseFilterCreateAccountTestTrait
which now only inserts the user into the database if it really
should be created.
It also restores test coverage of some other classes.
Change-Id: I661f4e0e2bcac4770e499708fca4e4e153f31fed
Change the IP to 127.0.0.1 (to indicate an internal IP), and blank
the XFF and UA when the performer of an action being logged by
CheckUser is the abuse filter user. Actions performed by the abuse
filter user can only be initated by the software, and as such should
not use the request's IP, XFF and UA. Also test the newly added
code.
Bug: T44345
Depends-On: I28acaaebd2d0067b700da0930e7b7ba924fa5c1c
Change-Id: Idd77545af94f9f9930d9ff38ab6423a72e680df9
A new core facility written for this use case.
Bug: T310662
Depends-On: I26b1cdba0a06ad16ad8bb71b455e1b6180924d17
Change-Id: I2b902d034a8c3308c0ba9878b69e873ca8fbda52
Almost all callers already provide an Authority in the form
of a User object, so mostly just need to change the typehints
Depends-On: I58661943c7e1acb6ff09798ee1a30be0fde3f459
Change-Id: I2ad86859c8194c14d7331f58db62b7cff4698085
This was most definitely my intention when I introduced the concept of
"generic vars", so it's a bit surprising to discover, 3.5 years later,
that the timestamp isn't computed there.
Also make the timestamp always be a string for consistency, since that's
the type documented on mw.org. I've manually checked all filters on
Wikimedia wikis using the timestamp variable, and added explicit int
casts where needed (although I think they'd still work due to implicit
casts).
Change-Id: Ib6e15225dd95c2eead7e48c200d203d6918e0c18
We want to make sure that all parameters are valid regardless of whether
there's a match.
Also make the minimum number of parameters = 2, so it's easier to switch
between this function and ip_in_range.
Change-Id: I141558a7ef4533485e315b3d93ea9b64f0959db7
Added support for ip_in_ranges which allow multiple ranges to be
checked at the same time. If the IP is in any of the ranges, the
function returns true.
Bug: T305017
Change-Id: Ic75c87ecd4cacf47ce2ff1b04173405230ff81d0
Convert a few integration tests to unit tests now that it's possible,
split the AbuseFilterSaveTest file into three different classes.
Change-Id: Ia2c0d7ab878b20a89324336a532abdc44f1e6b74
Introduce shorter methods, one for each steps, so that it's easier to
understand what the code is doing and figure out if the order makes
sense. The ConsequencesExecutor test is now a proper unit test. Also
simplify AbuseFilterConsequencesTest, removing old/wrong logic and
fixing two expected values that were actually wrong (but worked because
of the aforementioned wrong logic).
The only functional changes should be:
- We pick the longest block *after* checking the ConsequenceDisabler
consequences, so e.g. if a filter has a long block + warn and another
filter has a shorter block, we still keep the second one if warn will
disable the block.
- Remove disallow in presence of dangerous actions after checking
ConsequenceDisabler's and deduplicating blocks. Otherwise we may
remove disallow for filters where block (etc.) doesn't end up being
disabled. We may also want to consider not removing disallow at all,
now that messages are customizable.
Bug: T303059
Change-Id: If00adbf2056758222eaaea70b16d3b4f89502c20
- Use a /64 range for IPv6 instead of /16.
- Fix a curious and serious bug for IPv6, where grouping by range
would only use the first (!) number of the IP address, due to the
'v6-' prefix returned by IP::toHex.
- Fail hard if the identifier is unknown -- it's not something that's
supposed to happen.
- Include the type name in each identifier, instead of prefixing all
type names to all identifiers. This makes it easier to understand the
parts of the key.
- Test the whole lot.
Bug: T211101
Change-Id: I54c4209f2f0d5a4c5e7b81bed240ca3e28a2ded7
assertStatusMessage is being added to MediaWikiTestCaseTrait, rename
a method of the same name in FilterValidatorTest to avoid conflicts.
Change-Id: I642a3b620ab4d8ad620f7a1253fed98d6796883d
NeededBy: Ic01715b9a55444d3df6b5d4097e78cb8ac082b3e
This is a plain value object that represents the action being filtered,
replacing associative arrays that were being used up to this point.
We should now check whether it's possible to make it not require an
accountname (which complicates things), and then use it in related
classes as well, e.g. Parameters.
Change-Id: I9550c14819b600c97c46b632cc1c2d447972d69c
This replaces the previous pattern of callers having to use
RevisionLookup if the result was 'implicit'. Also, in some cases where
we were just hiding things if the visibility was !== true, properly
handle the implicit case by using the new method. Make the new method
return string constants rather than bool|string.
The new method also fixes some potential info leaks which happened when
the row was hidden, the user could view suppressed AbuseLog entries, but
the associated revision was also deleted and the user couldn't see it
(this shouldn't be relevant for WMF wikis since AF deletion is
oversight-level).
Also add a bunch of tests for the various cases to ensure we don't
regress again.
Bug: T261532
Change-Id: I929f865acf5d207b739cb3af043f70cb59243ee0
ParserStatus is now more lightweight, and doesn't know about "result"
and "from cache". Instead, it has an isValid() method which is merely a
shorthand for checking whether getException() is null.
Introduce a child class, RuleCheckerStatus, which knows about result and
cache and can be (un)serialized.
This removes the ambiguity of the $result field, and helps the
transition to a new RuleChecker class.
Change-Id: I0dac7ab4febbfdabe72596631db630411d967ab5
The old parser now has the correct name "Evaluator", so the
ParserFactory name was outdated. Additionally, the plan is to create a
new RuleChecker class, acting as a facade for the different
parsing-related stages (lexer, parser, evaluator, etc.), which is what
most if not all callers should use. The RuleCheckerFactory still returns
a FilterEvaluator for now.
Also, "Parser" is a specific term defining *how* things happen
internally, whereas "RuleChecker" describes *what* callers should expect
from the new class.
Change-Id: I25b47a162d933c1e385175aae715ca38872b1442
Remove unnecessary setters, injecting everything in the constructor.
These were leftovers from before the introduction of ParserFactory.
Remove public access to the conds used, include the information inside
the returned ParserStatus instead, and consequently simplify callers.
Change-Id: I0a30e044877c6c858af3ff73f819d5ec7c4cc769
So that the method can be typehinted in core.
Also add phan-var to fix broken master build due to typehint additions
in core.
Change-Id: I4a072e00ffeeb437753fc3d3c1f15de9929df510
This commit adds a class AFPSyntaxChecker which can statically analyze
a filter code to detect the following errors:
- unbound variables (which comes in two modes: conservative and liberal,
default to conservative)
- unused variables (disabled by default for compatibilty)
- assignment on built-in identifiers
- function application's arity mismatch
- function application's invalid function name
- non-string literal in the first argument of set / set_var
The existing parser and evaluator are modified as follows:
- The new (caching) evaluator no longer needs to perform variable
hoisting at runtime.
- Note that for array assignment, this changes the semantics.
- The new parser is more lenient, reducing parsing errors.
The static analyzer will catch these errors instead, allowing us
to give a much better error message and reduces the complexity of
the parser.
* The parser now allows function name to be any identifier.
* The parser now allows arity mismatch to occur.
* The parser now allows the first argument of set to be any expression.
Concretely, obvious changes that users will see are:
1. a := [1]; false & (a[] := 2); a[0] === 1
would evaluate to true, while it used to evaluate to the undefined value
due to hoisting
2. f(1)
will now error with 'f is not a valid function' as opposed to
'Unexpected "T_BRACE"'
3. length
will now error with 'Illegal use of built-in identifier "length"'
as opposed to 'Expected a ('
Appendix: conservative and liberal mode
The conservative mode is completely compatible with the current evaluator.
That is,
false & (a := 1); a
will not deem `a` as unbound, though this is actually undesirable because
`a` would then be bound to the troublesome undefined value.
The liberal mode rejects the above pattern by deeming `a` as unbound.
However, it also rejects
true & (a := 1); a
even though (a := 1) is always executed. Since there are several filters
in Wikimedia projects that rely on this behavior, we default the mode
to conservative for now.
Note that even the liberal mode doesn't really respect lexical scope
appeared in some other programming languages (see also T234690).
For instance:
(if true then (a := 1) else (a := 2) end); a
would be accepted by the liberal checker, even though under lexical scope,
`a` would be unbound. However, it is unlikely that lexical scope
will be suitable for the filter language, as most filters in
Wikimedia projects that have user-defined variable do violate lexical scope.
Bug: T260903
Bug: T238709
Bug: T237610
Bug: T234690
Bug: T231536
Change-Id: Ic6d030503e554933f8d220c6f87b680505918ae2
Create a dedicated "Exception" sub-namespace and remove the "AFP"
prefix, a leftover from the pre-namespace era.
Change-Id: I7e5fded9316d8b7d1628bc1a6ba8b1879ac901e1
