Summary:
Before, we were using a set domain of strings to model a boolean domain.
An explicit boolean domain makes it a bit clear what's going on.
There are two things to note here:
(1) This actually changed the semantics from the old set domain. The set domain wouldn't warn if the lock is held on only one side of a branch, which isn't what we want.
(2) We can't actually test this because the modeling for `Lock.lock()` etc doesn't work :(.
The reason is that the models (which do things like adding attributes for `Lock.lock`) are analyzed for Infer, but not for the checkers.
We'll have to add separate models for thread safety.
Reviewed By: peterogithub
Differential Revision: D4242487
fbshipit-source-id: 9fc599d
Summary:
In Java, we handle unknown code by propagating behavior from the parameters of the unknown function call to the return value (or constructed object, in the case of a constructor). But we do this in a somewhat silly way--generating a new summary with these semantics at each unknown call site. Instead, this diff introduces these two options as predefined behaviors and adds specialized code for them.
As a side effect of this approach, unknown functions are no longer counted as passthroughs. This is ok; the original behavior was less of a reasoned decision and more of an unintended consequence of the way we decided to handle unknown code.
This new approach ought to be more efficient than the old one, and as a virtuous side effect it will be easier to specify how to handle unknown code in other languages like C++.
Reviewed By: jeremydubreil
Differential Revision: D4205624
fbshipit-source-id: bf97445
Summary: Run all java tests with project-root at `infer/tests`. Do it to keep things consistent between clang and java tests
Reviewed By: sblackshear
Differential Revision: D4233236
fbshipit-source-id: c3f24fd
Summary:
Let's introduce some concepts. A "known unknown" function is one for which no Java code exists (e.g., `native`, `abstract`, and `interface methods`). An "unknown unknown" function is one for which Java code may or may not exist, but we don't have the code or we choose not to analyze it (e.g., non-modeled methods from the core Java or Android libraries).
Previously, Quandary handled both known unknowns and unknown unknowns by propagating taint from the parameters of the unknown function to its return value. It turns out that it is really expensive to do this for known unknown functions. D4142697 was the diff that starting handling known unknown functions in this way, and bisecting shows that it was the start of the recent performance problems for Quandary.
This diff essentially reverts D4142697 by handling known unknowns as skips instead. Pragmatically, doing the propagation trick for Java/Android library functions (e.g., `String` functions!) matters much more, so i'm not too worried about the missed behaviors from this. Ideally, we will go back to the old handling once performance has improved (have lots of ideas there). But I need this to unblock me in the meantime.
Reviewed By: jeremydubreil
Differential Revision: D4205507
fbshipit-source-id: 79cb9c8
Summary:
Developers will sometimes write GuardedBy("T.f") with the intended semantics: "guarded by the field f of the object with type T in the current state".
We want to support this to avoid false positives.
Reviewed By: peterogithub
Differential Revision: D4197476
fbshipit-source-id: acd00d9
Summary:
The way interfaces are dealt with led to a false positive,
where tryLock() works OK for a Lock but not for a ReentrantLock.
The solution is just to provide the model.
While I am at it I am adding some more standard tests for Lock and ReentrantLock, which were not present.
Reviewed By: sblackshear
Differential Revision: D4204551
fbshipit-source-id: 9b6de28
Summary:
Record an abstraction of the bug traces in the tests. The abstraction of a
trace is the sequence of descriptions. In practice, descriptions are either
empty, or of the form "start/end/return from/call to procedure X". They seem
pretty stable.
Motivation: there is nothing testing the traces reported by Infer right now,
even though they are surfaced to developers. For instance, Quandary uses
--issues-txt instead of --issues-tests to make sure the traces do not regress.
This change would make this approach more widespread.
Reviewed By: sblackshear
Differential Revision: D4159597
fbshipit-source-id: 9c83952
Summary:
When loading results from a json file, sort them. This prints results in some
sane order for both --issues-test and --issues-txt, removing the need for
post-processing of the result.
Reviewed By: cristianoc
Differential Revision: D4167029
fbshipit-source-id: 37e9f1c
Summary:
- rename java.make -> javac.make, config.make -> java.make, and move to infer/tests/ so it's easier to use from infer/tests/build_systems/
- use these from ant's test Makefile, much code reuse!
- factor out common functionality between java and clang
A wrinkle: sorting is now done the same way for --issues-tests and
--issues-txt, which produces bogus (but still as deterministic) sorting for
--issues-txt. This is more of a cosmetic issue, but I hope to fix it in a later
diff that gets rid of calls to `sort` in favour of sorting directly from
`InferPrint`.
Reviewed By: jberdine
Differential Revision: D4166841
fbshipit-source-id: ed6f232
Summary: The thread safety checker is run independently of other analyses, using the command "infer -a threadsafety -- <build-command>".
Reviewed By: sblackshear
Differential Revision: D4148553
fbshipit-source-id: bc7b3f9
Summary:
Our patch to Javalib has been accepted, so we can parse programs with invokedynamic!
invokedynamic still crashes Sawja, but I have worked around this by replacing all invokedynamic's with invokestatic's before passing them to Sawja.
This means we can handle everything about invokedynamic except calling the correct function (I call a dummy function with the correct signature for now).
We can try to actually call the right method in the future.
Reviewed By: jvillard
Differential Revision: D4160384
fbshipit-source-id: a8ef4e1
Summary: When searching for cast errors, types that were not Java objects, e.g. arrays of primitive types were not taken into account, leading to incorrect class cast excpetion reports.
Reviewed By: sblackshear
Differential Revision: D4166184
fbshipit-source-id: 7157c95
Summary: If a procedure is both a source and a sink for the same value, and it's a sink first, you will get a false positive when applying the summary for the procedure.
Reviewed By: cristianoc
Differential Revision: D4145246
fbshipit-source-id: 97f0022
Summary: `make test` was always exiting with exit code 0, even in the case of test failures. This is definitely not what we want.
Reviewed By: sblackshear
Differential Revision: D4154912
fbshipit-source-id: 87b4b2b
Summary: Mark native methods as defined so that the analysis generates a summary for those methods. When analyzing Java projects compiled with Buck, the summaries for the dependencies methods of are retrieved from the classpath. In this case, having access to the summary is useful to access the attributes of a callee when the callee is part of a, previously analyzed, Buck target.
Reviewed By: sblackshear
Differential Revision: D4141362
fbshipit-source-id: 75888c8
Summary:
Analyses should handle methods whose code is unknown and methods whose summary is a no-op differently.
Previously, this was done correctly for some kinds of methods (e.g., native methods, which were recognized as unknown), but not for others (interface and abstract methods).
This diff makes sure we correctly treat all three kinds as unknown.
Reviewed By: jeremydubreil
Differential Revision: D4142697
fbshipit-source-id: c88cff3
Summary:
Instead of the custom filtering done by `InferPrint --issues-tests`, use the
filtering done by `infer` and run without filtering for our e2e tests. We still
test the filtering for our build systems integration tests, and this diff
restores that behaviour for the ant test (hence the bugs removed from
ant/issues.exp).
Also add internal exceptions to most tests to get more signal out of them (eg,
knowing when we add assertion failures and the like).
Retire the old `--issues-tests` to limit the number of ways we do filtering.
Reviewed By: jeremydubreil
Differential Revision: D4131308
fbshipit-source-id: 35805cc
Summary:
Our default strategy for handling unknown code is to propagate taint from the actuals to the return value.
But for commonly-used methods like `StringBuilder.append` (used every time you do `+` with a string in Java), this doesn't work.
The taint should be propagated to both the receiver and the return value in these cases.
I'm considering a solution where we always propagate taint to the receiver of unknown functions in the future, but I am concerned about the performance.
So let's stick with a few special string cases for now.
Reviewed By: cristianoc
Differential Revision: D4124355
fbshipit-source-id: 5b2a232
Summary: A must-have for reporting taint errors and any other interprocedural error where the trace is sufficiently complex.
Reviewed By: jvillard
Differential Revision: D4124072
fbshipit-source-id: 26b3b2b
Summary: A must-have for reporting taint errors and any other interprocedural error where the trace is sufficiently complex.
Reviewed By: jvillard
Differential Revision: D4106352
fbshipit-source-id: b2677e6
Summary: We want to skip readwrite locks for now, maybe report on their misuses later.
Reviewed By: sblackshear
Differential Revision: D4110998
fbshipit-source-id: 986f77e
Summary:
Previously, we recorded direct sinks as sinks and transitive sinks as passthroughs. This makes it difficult to create an expanded interprocedural trace when recording an error because we can't distinguish between sinks (which we want to expand) and passthroughs (which we don't). This diff changes recording of sinks so that a sink is now the *last* function in a trace to call a sink. To find out what the original sink was, the summary for the transitive sink in the trace will now need to be (recursively) expanded until we bottom out in the original sink.
Will do the same for sources in a follow-up diff.
Reviewed By: cristianoc
Differential Revision: D4103759
fbshipit-source-id: 6f435f5
Summary:
Needed to support upcoming diff(s) that change the nature of sources/sinks in a trace. Today they are the *original* source/sink, but in the future they will be the *transitive* source/sink (last procedure to return a source/call a sink).
This new convention will make the `returnAllSources`/`callAllSinks` form of these tests not so useful, since `returnAllSources`/`callAllSinks` will now show up as a single source/sink in the trace (at least without expanding the trace). By making these tests intraprocedural, we can make sure that we're still testing everything that we want to.
Reviewed By: cristianoc
Differential Revision: D4103754
fbshipit-source-id: 1733ecf
Summary:
See code comment about `throw exn` being translated as `return exn`.
This problem was revealed by D4081279, which started grabbing access paths from exceptions.
Reviewed By: jvillard
Differential Revision: D4096391
fbshipit-source-id: 9d91513
Summary: Doing `sychronized(A.class)` where `A` is an inner class was not previously recognized by the `GuardedBy` checker.
Reviewed By: peterogithub
Differential Revision: D4095094
fbshipit-source-id: c832f9e
Summary:
We issue a thread safety warning on a class not
marked ThreadSafe, when it has a super that is. This makes some sense. But,
it will be nice to remind that a super is so maeked, else the mesg could
seem out of context or surprising
Reviewed By: sblackshear
Differential Revision: D4075145
fbshipit-source-id: ebc2b83
Summary:
This diff revises the makefiles for java tests so that they are based on
the files actually produced and depended on, instead of the existing
imperative style. This is, I think, clearer and easier to modify, and
enables a little more parallelism.
Reviewed By: jvillard
Differential Revision: D4072560
fbshipit-source-id: c16d4bd
Summary:
Right now, taint gets lost if it flows into a constructor or procedure whose implementation is missing.
Since the core Java (e.g., String) and Android classes (e.g, Intent) are among these, this is bad.
We could handle this by writing a bunch of models instead, but that would be a lot of work (plus we may still miss cases).
Reviewed By: jvillard
Differential Revision: D4051591
fbshipit-source-id: 65851c8
Summary:
Before, if I wrote code like
```
x = src()
sink(x)
sink(x)
```
we would report three times instead of two.
The first flow would be double-reported.
Reviewed By: jeremydubreil
Differential Revision: D4024678
fbshipit-source-id: fcd5b30
Summary: when a method has writes to a field outside of synchrnoization, issue an appropriate error message identifying the fields
Reviewed By: sblackshear
Differential Revision: D4015612
fbshipit-source-id: 4f697fc
Sam Blackshear
Andrzej Kotulski
Dulma Churchill
Peter O'Hearn
Jules Villard
Jeremy Dubreil
Josh Berdine