Summary:
Array types where the length is not statically known were represented
using fresh variables. This diff:
- Makes array type length optional, reducing the amount of work needed
for renaming, substitution, and normalization.
- Revises uses of array length so that the length component of a
Tarray type represents only the statically determined constant
length of an array type, and the length component of a Sizeof
expression represents the dynamically determined length of an array
value.
- Restricts the type of static lengths from a general expression
(Sil.exp) to an integer (Sil.Int.t), enforcing that static types are
constant. This in particular ensures that types contain no
variables, and so are invariant under operations such as renaming
and substitution.
- Removes the type substitution and renaming functions typ_sub,
typ_normalize, and typ_captured_ren. Now that array type lengths
are constant integers, all of these functions are the identity.
Reviewed By: cristianoc
Differential Revision: D3387343
fbshipit-source-id: b5db768
Summary:
This diff extends Sizeof expressions with an optional expression for the
length of the final extensible array, if any. For example, sizeof a
simple array `sizeof(t[n])` is represented by (modulo subtyping info)
`Sizeof t (Some n)`, and sizeof a struct whose final member is an array
`sizeof(struct s {... t[n] f})` is represented by `Sizeof (struct s
{... t[n] f}) (Some n)`.
This is an intermediate step toward eliminating expressions from types,
the redundancy between the length in the types and in the sizeof
expressions will be eliminated later.
Reviewed By: cristianoc
Differential Revision: D3358763
fbshipit-source-id: 2239bca
Summary:
Reimplement command line options in preparation for uniformly passing
options from the top-level infer driver that invokes a build command
through the build system to the descendant infer processes.
All command line options of all executables are collected into Config,
and declared using a new CommandLineOption module that supports
maintining backward compatibility with the current command line
interface. Very few values representing command line options are
mutable now, as they are set once during parsing but are constant
thereafter. All ordering dependencies are contained within the
implementation of Config, and the implementation of Config is careful to
avoid unintended interactions and ordering dependencies between options.
Reviewed By: jvillard
Differential Revision: D3273345
fbshipit-source-id: 8e8c6fa
Summary:This wasn't used anywhere. Frontends that wish to do something like goto can
just set the targets of the goto as successors of the current node, no need for
a special instruction to do that.
Reviewed By: sblackshear
Differential Revision: D3179826
fb-gh-sync-id: 572a6f2
fbshipit-source-id: 572a6f2
Summary:public
Refactor Utils.SymOp into a separate module, bringing the failure_kind
type and associated operations.
Reviewed By: cristianoc
Differential Revision: D3161640
fb-gh-sync-id: be3d7c9
fbshipit-source-id: be3d7c9
Summary:public
Eliminate the use of the -open Utils command line option passed to the compiler in favor of `open! Utils` in each source file. While slightly convenient, this option causes more headaches than it is worth with other tools e.g. merlin.
Reviewed By: jvillard
Differential Revision: D3168193
fb-gh-sync-id: 4285ef6
fbshipit-source-id: 4285ef6
Summary:public
Refactor the ml_loc type and associated operations from Utils to Logging. Seems a better fit, and reduces dependencies.
Reviewed By: cristianoc
Differential Revision: D3161440
fb-gh-sync-id: 2e09c25
fbshipit-source-id: 2e09c25
Summary:public
Rename functions and arguments to be more uniform, and change several to simplify types by using Builtin.t.
Reviewed By: cristianoc
Differential Revision: D3107836
fb-gh-sync-id: 8445f79
fbshipit-source-id: 8445f79
Summary:public
This will allow SymExec to depend on Inferconfig with introducing
circular dependencies, as Inferconfig calls Builtin.is_registered.
Reviewed By: jeremydubreil
Differential Revision: D3100614
fb-gh-sync-id: 786cf62
fbshipit-source-id: 786cf62
Summary:public
Assert false have been observed in Procname when analyzing some C projects.
This diff changes the Procname API to make it safe for Java: the java functions in the module don't assert false now. This takes care of the errors observed in C projects.
The new API forces changes throughout the codebase. In particular, the constant propagation module was making assumptions that it would only be executed on Java code, triggering assert false on C. Now it is safe.
For the remaining functions in the Procname module, those for other languages, a special assert false in Utils is used to print stack traces. This is for future debugging.
Reviewed By: sblackshear
Differential Revision: D3054077
fb-gh-sync-id: a77f1d7
shipit-source-id: a77f1d7
Summary:public
In order to implement the lazy dynamic dispatch algorithm, we need to generate a procedure description based on the types encountered during the symbolic execution. This diff adds support for analyzing such a prodecure description directly, without having to first serialize it to disk, which is slow and not necessary.
Reviewed By: cristianoc
Differential Revision: D3028226
fb-gh-sync-id: 1b2360e
shipit-source-id: 1b2360e
Summary: public Taint errors are complex, and each type requires its own specialized recommendation.
Reviewed By: jeremydubreil
Differential Revision: D3025921
fb-gh-sync-id: 8d7b45b
shipit-source-id: 8d7b45b
Summary:public
Lazy dynamic dispatch handling works as follows:
Assuming a call of the form:
foo(a);
where the static type of `a` is `A`. If during the symbolic execution, the dynamic type of the variable `a` is `B` where `B <: A`, then we create on-demand a copy `foo(B)` of `foo(A)` where all the uses of the typed parameter `a` are replaced with a parameter of type `B`. Especially, if `foo` contains virtual call, say `get` where `a` is the receiver, then the call gets redirected to the overridden method in `B`, which simulates the runtime behavior of Java.
This lazy dynamic dispatch mode is only turn on for the tracing mode for now in order to avoid conflicts with sblackshear's approach for sound dynamic dispatch.
Reviewed By: sblackshear
Differential Revision: D2888922
fb-gh-sync-id: 3250c9e
shipit-source-id: 3250c9e
Summary:public
Remove back-end infrastructure that exists only when on-demand mode is disabled.
This, together with removing a few command-line options, sheds a lot of weight in the back-end.
No changes expected for on-demand mode.
Reviewed By: sblackshear
Differential Revision: D2960242
fb-gh-sync-id: 220d821
shipit-source-id: 220d821
Summary:public
In tracing mode, we translate the runtime checks done by the JVM, so the checks for null happen independently from the what happens before the dereference.
Reviewed By: cristianoc
Differential Revision: D2981515
fb-gh-sync-id: 695de07
shipit-source-id: 695de07
Summary:public
I have seen enough comments in this space by people during code review to switch on the analyses the compiler can already do. This diff is an automated renaming of unused identifiers to _, with a few additional changes made when reading the diff of the results for things that stood out as particularly strange. This base-lines all of the existing warnings. I'm not sure this is a good idea, since it might be better for those familiar with each part of the code to look at these warnings and use them as pointers to suspicious code.
Reviewed By: jeremydubreil
Differential Revision: D2938376
fb-gh-sync-id: 6e67817
shipit-source-id: 6e67817
Summary:
public
Remove the need to create a record with two elements to create a procedure description
Reviewed By: cristianoc
Differential Revision: D2872744
fb-gh-sync-id: d26bbdc
Summary:
public
Fist resolve the method name based on the dynamic types found during the symbolic execution, then run the analysis on-demand with the resolved method name.
Reviewed By: cristianoc
Differential Revision: D2815744
fb-gh-sync-id: c55accd
Summary:
public
In the case of Java, the name of the method is enough to lookup the summary of the callees and run the symbolic execution. This revision separates the case of method call in Java and in C, C++ and Objective C. Most of the code for executing method calls was Clang specific and this is an intermediate step to be able to run the capture and analysis on demand.
Reviewed By: sblackshear
Differential Revision: D2809171
fb-gh-sync-id: da62dce
Summary:
public
When reading from static fields, the translation was overwriting the list of class members with the list of static ones. The backend was only looking up fields from the list of non static fields.
Reviewed By: sblackshear
Differential Revision: D2801759
fb-gh-sync-id: fe8ed80
Summary:
public
Using Typename.t in the list of superclasses to match the type for the key of the type environment. This avoids to make back and forth convertions from typename to type (csu, mangled name).
Depends on D2786574
Reviewed By: jberdine
Differential Revision: D2792116
fb-gh-sync-id: 6100f1a
Summary:
public
This diff cleans up the detection of assertion failures in C, C++ and Objective C which was previously hacked on top of the tracing mode for Java. The code is also generalized to detect any custom errors which can be defined using the `__infer_fail` builtin, and the case of assertion failure is now just the specific case of translating `assert` using `__infer_fail` directly in the clang frontend.
Reviewed By: jberdine
Differential Revision: D2786574
fb-gh-sync-id: dd1e1cf
Summary:
public
Move the naming of types to it own module, so that it can be used by modules `Sil` depends from like `Procname`
Reviewed By: jberdine
Differential Revision: D2773148
fb-gh-sync-id: a89f595
Summary:
public
Move the representation of data-structure into it own module, so that it can be used by modules `Sil` depends from like `Procname`.
Reviewed By: jberdine
Differential Revision: D2772791
fb-gh-sync-id: cda4e3a
Summary: public Crashes during the analysis are classified as timeouts in the .specs file. In addition, when there is a timeout, it does not say *why* the timeout occurred (hard time, symops, or recursion). This diff adds this information to the .specs file and adds a "fail hard" mode where crashes and timeouts will actually stop the analysis in developer mode (but will still be hidden in the normal production mode).
Reviewed By: jeremydubreil
Differential Revision: D2725382
fb-gh-sync-id: b0b4e5e
Summary: public Did this by adding an option to rearrange that turns of error reporting.
Reviewed By: dulmarod
Differential Revision: D2768396
fb-gh-sync-id: 4898d2d
Summary:
public To deal with ObjC nullability and give meaningful error
messages, we introduced the ObjC_NULL attribute in the symbolic execution to
mean that the object carrying the attribute is null because it was the result
of a method call from a null object. However, one cannot add attributes to null,
so we had to delay nullifying the object in order to have the attribute until we
can assign it to a program variable. However, if the temp variable was used in a condition,
we were not taking into account that its meaning is null. This diff addresses that and fixes
many FPs that we have encounter.
Reviewed By: ddino
Differential Revision: D2765167
fb-gh-sync-id: c0878dd
Summary: public First in a series of diffs fixing the __nullable checker for obj-c, which currently reports false positives on property accesses via getters and setters. To fix this problem, we need a special implementation of letderef and set for property accesses that don't report NPE errors. Factoring out the execution of these instructions from the main symbolic executor will make this easier.
Reviewed By: jeremydubreil
Differential Revision: D2763923
fb-gh-sync-id: 883a184
Summary:
public This continues the work on adding builtins for getters and setters
and removing the generated code from the frontend. This is extending it to setters and
removing the preanalysis that is no longer needed.
Reviewed By: akotulski
Differential Revision: D2734499
fb-gh-sync-id: 7cf5749
Summary:
public The ivar corresponding to the property is only available in the ast when the
implementation of the peroperty is available. Otherwise we add an ivar with the correct type
and the default name to the tenv and use it in the getter (and later in the setter).
This was not causing crashes because the generated code was swallowing the Missing_fld exception.
Now it flags it.
Reviewed By: akotulski
Differential Revision: D2734217
fb-gh-sync-id: 21c62af
Summary: public
We were generating getters and setters in the frontend, and then removing them if they were not needed
in the preanalysis. This diff adds a builtin getter that gets called if we are going to skip the function. That
means, if there was a getter written by the developers we still use that one rather than the builtin.
Code for setter and cleanup of preanalysis will come in a next diff.
Reviewed By: jvillard
Differential Revision: D2702890
fb-gh-sync-id: d65f25e
Summary: public
Refactor how propositions are created from prune nodes, especially for
inequalities.
Reviewed By: cristianoc
Differential Revision: D2700119
fb-gh-sync-id: 86a70a3
Summary: public Refactoring Printfargs checker a bit to
make it callable from symbolic executor, then calling it.
Reviewed By: jeremydubreil
Differential Revision: D2361286
fb-gh-sync-id: 4b73855
Summary: public
modules are better for namespacing.
How I made this diff:
1. moved list_* functions from utils.ml{,i} to iList.ml{,i}
2. shell commands:
grep '^val ' infer/src/backend/iList.mli | cut -f 2 -d ' ' | tr '\n' ' '
# gives a list of former list_ functions that IList implements, fed into the loops below:
LISTNAMES=" compare equal append combine exists filter flatten flatten_options find fold_left fold_left2 for_all for_all2 hd iter iter2 length fold_right map mem nth partition rev rev_append rev_map sort split stable_sort tl drop_first drop_last rev_with_acc remove_duplicates remove_irrelevant_duplicates merge_sorted_nodup intersect mem_assoc assoc map2 to_string"
# replace " list_*" function calls with IList.* ones
for i in $LISTNAMES; do find . -name '*.ml' -exec sed -i -e "s/ list_$i\b/ IList.$i/g" \{\} \; ; done
# replace (list_* functions with (IList.* ones
for i in $LISTNAMES; do find . -name '*.ml' -exec sed -i -e "s/(list_$i\b/(IList.$i/g" \{\} \; ; done
# ditto with [
for i in $LISTNAMES; do find . -name '*.ml' -exec sed -i -e "s/\[list_$i\b/[IList.$i/g" \{\} \; ; done
3. Then fix up the rest by hand. In particular, stuff that called Utils.list_*
explicitely, and stuff that used the "Fail" exception that has moved to
IList. (may revisit this in the future)
Reviewed By: jeremydubreil, cristianoc
Differential Revision: D2550241
fb-gh-sync-id: cd64b10
Summary: public
Dictionary literals are normally implemented using
`+dictionaryWithObjects:forKeys:count:` but were modeled as
`+dictionaryWithObjectsAndKeys:`
In particular, `@{@"aaa": nil}` would trigger a sentinel error instead of an NPE.
This models dictionary literals as a special infer builtin that the backend
interprets so as to give NPEs when passed nil objects or keys.
Reviewed By: dulmarod
Differential Revision: D2550039
fb-gh-sync-id: 1a10656
Summary: @public Infer previously did not work correctly when a function returns the result of a skip function:
```
retUndef() {
x = undefined();
return x;
}
derefUndef() {
y = retUndef();
y.doSomething(); // Symexec_memory_error here, prevents spec inference
}
```
The problem is that angelic mode did not know to add the return value of `retUndef()` to the footprint.
This diff fixes the problem by adding return values marked with the `Aundef` attribute to the footprint.
This is done lazily (e.g., a value only gets added to the footprint when you try to deref it).
Reviewed By: @jvillard
Differential Revision: D2444929
Summary:
`get_resource_or_undef` attribute is weird and was causing problems for me in another diff.
This diff refactors the attribute categories to make resource and undef separate.
Summary:
Move proc_attributes to a separate module.
Field err_log, in common between proc desc and summary, can now be moved to ProcAttributes without creating cycles of dependencies.
Josh Berdine
Jeremy Dubreil
Sam Blackshear
Dulma Churchill
Jules Villard
Cristiano Calcagno
Dino Distefano
Peter O'Hearn