Should fix#2426.
Essentially, we disregard expressions used inside annotations, if these
annotations occur in a file that has `from __future__ import annotations`, as
this prevents the annotations from being evaluated.
* Adds fix for __init_subclass__ bug.
* Adds test case.
* Move test on name.
I think it makes more sense here, alongside the other "special" method names.
Fixes#1969.
The points-to analysis does not know that the assignment `input = raw_input`
cannot fail under Python 2, and so there are two possible values that `input`
could point-to after exiting the exception handler: the built-in `input`, or the
built-in `raw_input`. In the latter case we do not want to report the alert, and
so adding a check that the given function does not point-to the built-in
`raw_input` suffices.
Should fix#1833, #2137, and #2187.
Internally, comprehensions are (at present) elaborated into local functions and
iterators as described in [PEP-289](https://www.python.org/dev/peps/pep-0289/).
That is, something like:
```
g = (x**2 for x in range(10))
```
becomes something akin to
```
def __gen(exp):
for x in exp:
yield x**2
g = __gen(iter(range(10)))
```
In the context of the top-level of a class, this means `__gen` looks as if it is
a method of the class, and in particular `exp` looks like it's the `self`
argument of this method, which leads the points-to analysis to think that `exp`
is an instance of the surrounding class itself.
The fix in this case is pretty simple: we look for occurrences of `exp` (in fact
called `.0` internally -- carefully chosen to _not_ be a valid Python
identifier) and explicitly exclude this parameter from being classified as a
`self` parameter.
Fixes#1832.
In the taint sink, we add an additional check that the given control-flow node
can indeed point to a value that is mutable. This takes care of the guard on the
type.
If and when we get around to adding configurations for all of the taint
analyses, we may want to implement this as a barrier instead, pruning any steps
that go through a type test where the type is not mutable.
Fixes https://github.com/Semmle/ql/issues/1572
Adjust mock so it's more aligned with what the flask code actually does. Tests
were passing before, even though we didn't handle the case in real code :\
If the legacy configuration is only enabled if there are no other
configurations, defining a configuration in an imported library can lead to
unwanted results. For example, code that uses `any(MyTaintKind t).taints(node)`
would *stop* working, if it did not define its own configuration. (this actually
happened to us)
We performed a dist-compare to ensure there is not a performance deg ration by
doing this. Results at https://git.semmle.com/gist/rasmuswl/a1eca07f3a92f5f65ee78d733e5d260e
Tests that were affected by this:
- RockPaperScissors + Simple: new edges because no configuration was defined for
SqlInjectionTaint or CommandInjectionTaint
- CleartextLogging + CleartextStorage: new edges because no configuration was
defined before, AND duplicate deges.
- TestNode: new edges because no configuration was defined before
- PathInjection: Duplicate edges
- TarSlip: Duplicate edges
- CommandInjection: Duplicate edges
- ReflectedXss: Duplicate edges
- SqlInjection: Duplicate edges
- CodeInjection: Duplicate edges
- StackTraceExposure: Duplicate edges
- UnsafeDeserialization: Duplicate edges
- UrlRedirect: Duplicate edges
We don't use a locked-down version of six, so some internal things probably
changed from the version used last time, and the versoin I have installed.
Long term fix would be to use a specific version of six for tests!
Due to internal PR#35123 we now actually run the tests under
`python/ql/test/2/...`
These seems like a regression, since the tests state that N is ok, but A and J
should not be allowed.
For now we can accept them, so we don't block all other Python PRs
Due to internal PR#35123 we now actually run the tests under
`python/ql/test/2/...`
Since we haven't done this in a while, test output has changed a bit. These
changes look perfectly fine.
This was brought up on the LGTM.com forums here:
https://discuss.lgtm.com/t/warn-when-always-failing-assert-is-reachable-rather-than-unreachable/2436
Essentially, in a complex chain of `elif` statements, like
```python
if x < 0:
...
elif x >= 0:
...
else:
...
```
the `else` clause is redundant, since the preceding conditions completely
exhaust the possible values for `x` (assuming `x` is an integer). Rather than
promoting the final `elif` clause to an `else` clause, it is common to instead
raise an explicit exception in the `else` clause. During execution, this
exception will never actually be raised, but its presence indicates that the
preceding conditions are intended to cover all possible cases.
I think it's a fair point. This is a clear instance where the alert, even if it
is technically correct, is not useful for the end user.
Also, I decided to make the exclusion fairly restrictive: it only applies if
the unreachable statement is an `assert False, ...` or `raise ...`, and only
if said statement is the first in the `else` block. Any other statements will
still be reported.
This is vulnerable to SQL injection because of the quotes around %s -- added
some code that highlights this in test.py
Since our examples did this in the safe query, I ended up rewriting them
completely, causing a lot of trouble for myself :D
