Summary:
# Current design
Infer analysis is currently two staged:
1) proc-level callbacks calculate summary, including writing down the
issues if applicable.
2) file-level callbacks (formerly cluster callbacks, see the prev diff) are executed next; they are supposed to emit
additional issues that are impossible to emit based on mere
proc-context.
Currently RacerD and Starvation use file-level callback; in near future
we plan to onboard Nullsafe checker as well.
# Problem
Contract of callback (1) is clear: given a proc and existing
summary, the checker updates it and returns a modified summary. This
summary later on gets serialized (in-memory + external) and can be consumed by
other chechers. Issues written in summary will get reported when
analysis is over.
In constrast, contract of (2) is wild west: the function returns unit.
In practice, what the checkers do is create IssueLog and serialize it to
checker-specific directory.
Then another part of program (InferPrint.ml) knows about this side
effect, reads the error log for checkers and ultimately get it reported
together with errors written at stage (1).
This is problematic because it is hard to reason about the system and it
makes onboarding new checkers to (2) error-prone.
# This diff
This diff brings (2) on par with (1): now file-level callback has a
clear contract: it should be side effect free, and the only
responsibility is to fill out and return IssueLog.
Additionally, we make the notion of "checker-specific issue directory"
an official thing, so the checker only needs to specify the name,
everything else will be made automatically by orchestation layer,
including cleanup.
# Starvation
Implementing the new contract is starvation is possible and desirable, but involved: see comment
in the code, so we leave it up to the future work to fix that.
Reviewed By: ngorogiannis
Differential Revision: D20115024
fbshipit-source-id: fb2f9b7e6
Summary:
1. Some invariants are tricky enough to be documented. This is especially
important for cases related with error reporting. Lets document it.
2. Cluster callback -> File callback rename.
Reviewed By: ngorogiannis
Differential Revision: D20093932
fbshipit-source-id: e716f1f5b
Summary:
Previous implementation supported only stringy params (strings and
stringified bools). Current one exposes a proper variant `Annot.t`,
with support for all possible param values in Java except
numbers (more on that below).
This change is required for implementing `Nullsafe(LOCAL)` as the
annotation used to specify nullsafe behaviour has a more complex
structure than what we've dealt with before.
**Why support for number values was not added**: supporting numbers
requires using `int64`. Unfortunately, adding another variant `Vnum
int64` to `Annot.t` causes a runtime failure on assert in
`MaximumSharing.ml:133`. It seems that it might be enough to flip
`fail_on_nonstring` from `true` to `false`, but since this would
require additional testing and is not required for my case, I'll leave
checking this to whoever needs to use numeric annot params in future.
Reviewed By: ezgicicek
Differential Revision: D19855923
fbshipit-source-id: 878e33856
Summary:
The big one:
- stop using polymorphic `<>`, `<`, `>`, ..
- add `<>` to `PolyVariantEqual` escape hatch now that `<>` is as taboo as `=`
- Interestingly, there were a lot of uses of `Z.(x < y)`, which although
they seem to use `Z.lt` actually used polymorphic comparison. The actual
comparison infix operators of `Z` are cleverly hidden in `Z.Compare`
instead, which makes them impractical to use...
Reviewed By: jberdine
Differential Revision: D19861584
fbshipit-source-id: 5dce08ad9
Summary:
The domain has a notion of lock state (taken/not taken) and any access occurring will remember the current lock state.
Methods in Java can be `synchronized` meaning the lock is taken automatically at method start and release on return.
Currently, instead of starting analysis of a `synchronized` method with an initial lock state of "lock taken", the summary would be computed as if there is no lock, and then a caller would peek at whether the callee is `synchronized` and change the callee summary accordingly before taking it into account. Also, when an access happens (not a call) the analysis always consults the pdesc of the current method to check whether the method is `synchronized`.
Do things the right way: if method is `synchronized`, start with the lock taken. When the method exits, the lock-state postcondition is computed by releasing once the lock.
This is a behaviour-preserving change.
Reviewed By: mityal
Differential Revision: D19742559
fbshipit-source-id: 1d0fce3f6
Summary:
The goals are:
- Increase precision in C-languages by ditching access paths.
- Help with eventually sharing the abstract address module with RacerD.
- Reports are now language-mode specific (eg `->` in clang vs `.` in Java).
It's not exactly access expressions used here. Instead the pattern `(base, access list)` is used where `access` is `HilExp.Access.t`. This is done to ease the way `deriving` is used for creating two comparison functions, one that cares about the root variable and one that doesn't; and also because the main function that recurses over accesses (`normalise_access_list`) visits the accesses from innermost to outermost.
Also, kill some dead code.
Reviewed By: skcho
Differential Revision: D19741545
fbshipit-source-id: 013bf1a89
Summary:
The problem with is_override that it can be misleading: it does not
check that that class of the first method is a subtype of the second; in
fact it totally ignores the classes.
This method is publicly exposed so lets just call what it does.
Reviewed By: jvillard
Differential Revision: D19724295
fbshipit-source-id: dc0919193
Summary:
To emulate the `ThreadSafe` contract in C++/ObjC, reporting was gated behind a check that ensured a C++/ObjC class has a `std::mutex` member (plus other filters). This is reasonable, but it has some drawbacks
- other locks may be used, and therefore must be added to the member check;
- locking mechanisms that use the object itself as a monitor cannot be modelled (`synchronized` in ObjC)
RacerD already has `ThreadsDomain` which models our guess on whether a method is expected to run in a concurrent context, and which in C++/ObjC boils down to whether the method non-transitively acquires a lock. This should be a good enough indicator that the class should be checked regardless of whether the locks are member fields. This diff gates the C++/ObjC check on that abstract property.
Reviewed By: dulmarod
Differential Revision: D19558355
fbshipit-source-id: 229d7ff82
Summary:
- Thread the two types into one instead of having a record where the `path` field doesn't always make sense (`Class` case).
- Improved pretty printing of class objects (java only).
- Move starvation-specific stuff out of `AbstractAddress` (eg `make_java_synchronized`).
- Slight optimisation of `apply_subst` for when a parameter is used without additional accesses inside a method (then, the substitution need not modify the term substituted for the parameter in any way).
Reviewed By: ezgicicek
Differential Revision: D19639922
fbshipit-source-id: 1cebecf5d
Summary: This will be eventually used in RacerD. Further diffs up the stack will migrate away the starvation-specific aspects.
Reviewed By: ezgicicek
Differential Revision: D19623392
fbshipit-source-id: a72650718
Summary:
The "access path" memory model (equal access paths iff equal object addresses) is suited to when aliasing occurs only at the roots (i.e. variables). When there is intentional aliasing in the middle of an access path, this model will miss the aliasing. For instance if `[x.f] == [y.g]`, then also `[x.f.h] == [y.g.h]`, but the latter access paths are unequal.
In Java, non-static inner classes consistently alias `this.this$0` inside an inner class, which points to the "parent" outer-class object. So if two inner-class objects (belonging to different inner classes) access `this(type:InnerClassA).this$0.f` and `this(type:InnerClassB).this$0.f` the equality will be missed (many other combinations exist). This isn't strictly due to the memory model -- any alias analysis would have to do some class invariant inference to detect this.
For this purpose `AccessPath.inner_class_normalize` exists (it replaces `this.this$0` with `this` of the appropriate type), but this breaks the invariant that we know which formal parameter is at the root (there may not even exist a `this` parameter if the method is static). So this was buggy.
Here we simply recursively remove the synthetic field prefix of the accesses list, while computing forwards the object type. This is only applied when we check aliasing across threads. This will also allow actuals/parameters substitutions (stacked diff) which normalisation was breaking.
Reviewed By: jberdine
Differential Revision: D19601455
fbshipit-source-id: 7e42667b6
Summary: Keep the type name of the class as the key in the map constructed from class names to their methods in a file. This will be used later, and also why string?
Reviewed By: dulmarod
Differential Revision: D19557707
fbshipit-source-id: aa8569581
Summary: Currently, empty summaries are passed to the cluster callbacks. This is pointless and could potentially lead to recomputing already analysed summaries. This change passes only procnames to the callback, and `Ondemand` is used to load the summary or analyse as needed.
Reviewed By: jvillard
Differential Revision: D19544043
fbshipit-source-id: 28ab642c3
Summary: There is no need to provide type environments to cluster analysers, since the execution environment can be used to retrieve those on demand.
Reviewed By: jvillard
Differential Revision: D19543561
fbshipit-source-id: f9b064011
Summary: Deadlocks often result in two reports if not deduplicated (two traces), so there is some logic for doing that. Locks recently became an opaque type, with the `get_access_path` loophole supporting that deduplication ordering. Fix that here and remove `Lock.get_access_path`.
Reviewed By: jvillard
Differential Revision: D19465223
fbshipit-source-id: b597e3c65
Summary: Change `MayBlock` and `StrictModeCall` constructors from taking a string to a `Procname.t`, which was the sole source of that string anyway.
Reviewed By: ezgicicek
Differential Revision: D19465226
fbshipit-source-id: e3ed6ef88
Summary:
If a race exists in two or more overloads of the same method and we use only the class and method name in the report text, then the current bug hashing algorithm will identify the two reports as duplicates.
To avoid this, the report had the class, method and list of type parameters. This is unreadable, however, and redundant (the report is already located within the method in question). So at the risk of duplicates, use only class+method names.
Also, fix a bug in `Procname.pp_simplified ~withclass` where `withclass` was ignored for C++/ObjC methods.
Now:
> Read/Write race. Non-private method `FrescoVitoImageSpec.onCreateInitialState(...)` indirectly reads with synchronization from `factory.AnimatedFactoryProvider.sImpl`. Potentially races with unsynchronized write in method `FrescoVitoImageSpec.onEnteredWorkingRange(...)`.@ [Litho components are required to be thread safe because of multi-threaded layout](https://fburl.com/background-layout). Reporting because current class is annotated `MountSpec`, so we assume that this method can run in parallel with other non-private methods in the class (including itself).
Before
> Read/Write race. Non-private method `void FrescoVitoImageSpec.onCreateInitialState(ComponentContext,StateValue,StateValue,Uri,MultiUri,ImageOptions,FrescoContext,Object,ImageListener)` indirectly reads with synchronization from `factory.AnimatedFactoryProvider.sImpl`. Potentially races with unsynchronized write in method `FrescoVitoImageSpec.onEnteredWorkingRange(...)`.@ [Litho components are required to be thread safe because of multi-threaded layout](https://fburl.com/background-layout). Reporting because current class is annotated `MountSpec`, so we assume that this method can run in parallel with other non-private methods in the class (including itself).
Reviewed By: artempyanykh
Differential Revision: D19462277
fbshipit-source-id: aebc20d89
Summary:
Introduce a new notion of equality for comparing abstract addresses in distinct threads:
```
(** Abstract address for a lock. There are two notions of equality:
- Equality for comparing two addresses within the same thread/process/trace. Under this,
identical globals and identical class objects compare equal. Locks represented by access paths
rooted at method parameters must have equal access paths to compare equal. Paths rooted at
locals are ignored.
- Equality for comparing two addresses in two distinct threads/traces. Globals and class objects
are compared in the same way, but locks represented by access paths rooted at parameters need
only have equal access lists (ie [x.f.g == y.f.g]). This allows demonically aliasing
parameters in *distinct* threads. This relation is used in [may_deadlock]. *)
```
Reviewed By: skcho
Differential Revision: D19347307
fbshipit-source-id: 9f338731b
Summary: Access expressions can appear in casts, or sometimes other constructors, inside a `HilExp.t`. Extraction of the access expression can ignore those wrappers. Introduce a single function for doing that throughout the analyser.
Reviewed By: ezgicicek
Differential Revision: D19410673
fbshipit-source-id: a724cb466
Summary: In whole-program mode, analysing a method requires analysing first all constructors of the same class. This is not needed in normal mode, so gate that computation under `starvation_whole_program` for efficiency.
Reviewed By: artempyanykh
Differential Revision: D19393412
fbshipit-source-id: 2277e6b5e
Summary: Now that we have the kind of lock stored (global/class obj/path rooted at parameter), use it for comparison/equality, while ignoring the root variable of the access path, which is only used for printing.
Reviewed By: skcho
Differential Revision: D19346801
fbshipit-source-id: c65661dc6
Summary: Add command line option that directs racerd to treat all return values from unknown code (including abstract methods) as owned objects. This is essentially treating return values with full angelicism
Reviewed By: artempyanykh
Differential Revision: D19368375
fbshipit-source-id: 6a10153fa
Summary:
As part of enabling substitution of arguments into lock names in summaries, and potentially changing the equality relation over locks (so that we can eventually make equal `x.f.g` and `this.f.g` for example), the lock type needs to be elaborated so that the root of an access path is classified into:
- global variables are clearly separated from all other classes (invariant to substitution)
- class objects are also separate (also invariant, but identified by type)
- all other parameters remember their positional index (so that substitution becomes easier)
For now the comparison/equality is kept identical, so as to make easy CI comparisons.
Reviewed By: skcho
Differential Revision: D19232577
fbshipit-source-id: 0cd6a43db
Summary: Before going to a new lock representation that will allow, eg, substitutions with arguments for parameters on method calls, make the interface of the lock type abstract to ease the transition.
Reviewed By: mityal
Differential Revision: D19309359
fbshipit-source-id: 5277357ee
Summary:
Remove Clang and Java submodules of Typ.Fieldname. They are unnecessary and they reflect a fake dichotomy: there is only one fieldname type. To distinguish between fields of Java classes and other C constructs, there is a helper function provided, but the idea is simple: obtain the class type the field belongs to, and check if it's a Java class.
This diff still preserves behaviour, but removes as many functions as possible from the interface, to leave a small surface.
Reviewed By: mityal
Differential Revision: D18962423
fbshipit-source-id: ffe6933ee
Summary: Guava uses assertions to ensure a future can be gotten without blocking (this means that if the future is not done, the app will crash). This diff teaches the starvation analyser about a number of such assertions, by treating them as assumes (since we don't care about exceptions).
Reviewed By: jvillard
Differential Revision: D18893427
fbshipit-source-id: 4d26a202b
Summary: A future is guaranteed not to block if `isDone()` has returned true first. Add logic for supporting that by remembering the objects that we have called `isDone` on and by making `assume` do the right thing with that knowledge. All this is achieved with the attribute domain.
Reviewed By: ezgicicek
Differential Revision: D18833901
fbshipit-source-id: 7f4ea0cd1
Summary:
When retrieving a value from a container, we previously had an arbitrary hack which would
- In java, give no ownership to the returned object (trying to be sound)
- In C++ give conditional ownership to the current method's first argument (trying to be complete, but doing it badly, as the first argument may not be the `this` object in a static method, or we might be accessing it through another parameter altogether).
Harmonise both by using the existing ownership of the container as ownership value for the returned object (leaning towards completeness).
Reviewed By: jvillard
Differential Revision: D18882800
fbshipit-source-id: f98f8d315
Summary:
Also add logic for recognising excessive timeouts. Refactor the code
around timeouts a little.
Reviewed By: artempyanykh
Differential Revision: D18807836
fbshipit-source-id: df5a1b566
Summary:
`Object.wait()` must be called on a locked monitor and it releases the
lock immediately, as far as other threads are concerned
(it also magically re-takes the lock when the monitor is `notified`).
Starvation can only occur if the UI thread is waiting
a lock that is distinct to that being waited on.
The check present was over-approximate in that it was checking that there exists a lock held by the UI thread and the thread issuing the `wait`, but did not make sure that lock was *not* the one waited on.
Amusingly, the e2e test was correct, but the reporting code wasn't.
Reviewed By: dulmarod
Differential Revision: D18782919
fbshipit-source-id: b3b98239e
Summary: Similar to constructor established attributes, we do the same here for static initializers. That is, attributes of static properties are injected into the initial state of every method.
Reviewed By: ezgicicek
Differential Revision: D18763192
fbshipit-source-id: 3879a27c5
Summary:
This diff enables parsing and auto-formatting documentation
comments (aka docstrings).
I have looked at this entire diff and manually made some changes to
improve the formatting. In some cases it looked like it would take too
much time, or benefit from someone more familiar with the code doing
it, and I instead disabled auto-formatting docstrings in those files.
Also, there are some source files where the docstrings are invalid,
and some where the structure detected by the parser appears not to
match what was intended. Auto-formatting has been disabled for these
files.
Reviewed By: ezgicicek
Differential Revision: D18755888
fbshipit-source-id: 68d72465d
Summary:
RacerD is angelic in the sense that when a method has no summary, no accesses are added to the symbolic state when we call that method. However, when the method returns an object we then proceed to access, this leads to non-angelic behaviour: if the object is assumed to be un-owned, then the accesses may lead to a race.
This manifests itself on Litho components, which are generated code without sources and thus RacerD has no CFG to analyse, and therefore produces no summary. The `Builder` patterns used in these classes are ubiquitous, and full of spurious races due to the fact that the returned objects, even though freshly allocated, are un-owned as far as RacerD knows.
Here, instead of going full-angelic and assuming that every method without a summary returns an owned object (which is a bit extreme), we try to model the `Builder` pattern wrt ownership. That is, static `create` methods returning `Builder` types are assumed to return full ownership; and, non-static methods of a `Builder` class which return the same type as their receiver are assumed to return conditional ownership.
Reviewed By: ezgicicek
Differential Revision: D18748423
fbshipit-source-id: bd53d4b67
Summary:
One standard way to schedule work is by starting a thread. We treat this by
- Treating invocations of `start` on a receiver with the `Runnable _` attribute as scheduling that runnable for parallel execution in the background (as opposed to on the UI thread).
- If `start` is used on an object of a subclass of `Thread` everything already works thanks to the `get_exp_attributes` function which will implicitly ascribe to an expression the attribute `Runnable _` if the expression points to an object with a known `run` method. This will even take care of some degree of dynamic dispatch, yay!
- If `start` is used on a `Thread` object which has been created with a constructor call provided with a `Runnable` argument, we have to appropriately model that constructor call, which is what is done in `do_call`.
Reviewed By: jvillard
Differential Revision: D18726676
fbshipit-source-id: 0bd83c28e
Summary:
A current blind spot is when object construction stores specific executors / runnables to object fields, which are then never mutated and accessed from normal methods. IOW the attributes established in the constructor are necessary to report properly inside a normal method (assuming these attributes are not invalidated by method code).
To achieve this, first we retain a subset of the final state attributes in the summary (only those that affect instance variables, in constructor methods). Then, when we analyse a non-constructor method:
- we analyse all constructors
- remove all attributes from the attribute map whose key is not an expression of the form `this.x. ...`
- re-localise all remaining keys so that they appear as rooted in the `this` local variable of the current procedure
- join (intersect) all such attribute maps
- use the result in place of initial state as far as the attribute map is concerned for the analysis of the current procedure, which can now start.
This means we can catch idioms that use side-effectful initialisation for configuring certain object fields like executors or runnables.
Reviewed By: jvillard
Differential Revision: D18707890
fbshipit-source-id: 42ac6108f
Summary: Another way to schedule work in android is by posting it to a `Handler`. A handler can be constructed out of the main looper, which forces it to schedule work on the UI thread. To model all this, we add syntactic models for getting the main looper and for creating handlers, and dataflow attributes for tracking that an expression is a looper/the main looper, or a handler constructed out of a looper.
Reviewed By: skcho
Differential Revision: D18706768
fbshipit-source-id: 7c46e6913
Summary:
Instead of trying to figure out what runnable is directly passed to an executor,
use attributes to track the flow of a runnable. This has several advantages:
- Can track runnables across function return values.
- Can somewhat overcome the information loss under dynamic dispatch.
- Unifies handling with other attributes.
Reviewed By: skcho
Differential Revision: D18672676
fbshipit-source-id: a06a0e6ff
Summary:
- Unify treatment of modelled and annotated executors by making things go through attributes.
- Add a return attribute to summaries, so that we can track flows of thread guards/executors/future stuff through returned values.
- Dispatch modeled functions to model summaries.
This will help in following diffs where runnables will also go through attributes.
Reviewed By: skcho
Differential Revision: D18660185
fbshipit-source-id: e26b1083e
Nikos Gorogiannis
Mitya Lyubarskiy
Artem Pianykh
Jules Villard
Sungkeun Cho
Josh Berdine