Summary:
Increases precision a bit. I didn't observe speed problems on what I tested. (But, who knows?)
Closes https://github.com/facebook/infer/pull/799
Reviewed By: jvillard
Differential Revision: D6284206
Pulled By: rgrig
fbshipit-source-id: 6f1e8631f
Summary:
Instead of emitting an ad-hoc builtin on variable declaration emit a new
metadata instruction. This allows us to remove the code matching on that
ad-hoc builtin that had to be inserted in several checkers.
Inferbo & pulse used that information meaningfully and had to undergo
some minor changes to cope with the new metada instruction.
Reviewed By: ezgicicek
Differential Revision: D14833100
fbshipit-source-id: 9b3009d22
Summary:
Add support for GuardedBy: we deviate from the spec as follows:
- No warnings issued for any access within a private method, unless that method is called from a public method and the lock isn't held when the access occurs.
- Warnings are suppressed with the general RacerD mechanism, ie `ThreadSafe(enableChecks=false)`
- GuardedBy warnings override thread-safety violation warnings on the same access, because GuardedBy has a clearer and simpler contract.
Also, some simplifications, cleanups and perf improvements (eg avoid unreportable procs at the top level as opposed to on each of their accesses).
Reviewed By: jeremydubreil
Differential Revision: D14506161
fbshipit-source-id: b7d794051
Summary:
Context: "quandary" traces optimise for space by only storing a call site (plus analysis element) in a summary, as opposed to a list of call sites plus the element (i.e., a trace). When forming a report, the trace is expanded to a full one by reading the summary of the called function, and then matching up the current element with one from the summary, iterating until the trace cannot be expanded any more. In the best case, this can give a quadratic saving, as a real trace gets longer the higher one goes in the call stack, and therefore the total cost of saving that trace in each summary is quadratic in the length of the trace. Quandary traces give a linear cost.
HOWEVER, these have been a source of many subtle bugs.
1. The trace expansion strategy is very arbitrary and cannot distinguish between expanded traces that are invalid (i.e., end with a call and not an originating point, such as a field access in RacerD). Plus the strategy does not explore all expansions, just the left-most one, meaning the left most may be invalid in the above sense, but another (not left-most) isn't even though it's not discovered by the expansion. This is fixable with major surgery.
2. All real traces that lead to the same endpoint are conflated -- this is to save space because there may be exponentially many such traces. That's OK, but these traces may have different locking contexts -- one may take the lock along the way, and another may not. The expansion cannot make sure that if we are reporting a trace we have recorded as taking the lock, will actually do so. This has resulted in very confusing race reports that are superficially false positives (even though they point to the existence of a real race).
3. Expansion completely breaks down in the java/buck integration when the trace goes through f -> g -> h and f,g,h are all in distinct buck targets F,G,H and F does not depend directly on H. In that case, the summary of h is simply not available when reporting/expanding in f, so the expanded trace comes out as truncated and invalid. These are filtered out, but the filtering is buggy and kills real races too.
This diff completely replaces quandary traces in RacerD with plain explicit traces.
- This will incur the quadratic space/time cost previously saved. See test plan: there is indeed a 30% increase in summary size, but there is no slowdown. In fact, on openssl there is a 10-20% perf increase.
- For each endpoint, up to a single trace is used, as before, so no exponential explosion. However, because there is no such thing as expansion, we cannot get it wrong and change the locking context of a trace.
- This diff is emulating the previous reporting format as much as possible to allow good signal from the CI. Further diffs up this stack will remove quandary-trace specific things, and simplify further the code.
- 2 is not fully addressed -- it will require pushing the `AccessSnapshot` structure inside `TraceElem`. Further diffs.
Reviewed By: jberdine
Differential Revision: D14405600
fbshipit-source-id: d239117aa
Summary:
This will be used in the future to determine what to do with destructors
in pulse.
Reviewed By: mbouaziz
Differential Revision: D14324759
fbshipit-source-id: bc3c34471
Summary:
The Eradicate backend is reporting nullable type errors, that are not always necessarily leading to null pointer exceptions.
For example, the analysis is designed to be consistent with the Java type system and report on the following code:
String foo(boolean test) {
Object object = test ? new Object() : null;
if (test) {
return object.toString(); // the analysis reports here
}
}
even though the code will not crash.
In order to make this aspect clear, this diff renames the warnings `Null Method Call` and `Null Field Access` into `Nullable Dereference`
Reviewed By: ngorogiannis
Differential Revision: D14001979
fbshipit-source-id: ff1285283
Summary:
The purpose these serve is unclear to me. From the comment I *think*
they were used to hint to the biabduction backend that smart pointers
are just pointers. That said, The tests still mostly pass even without
that (just a few `weak_ptr` tests changed from `NULL_DEREFERENCE` to
`Bad_footprint`).
Moreover, this extra dereference was added unreliably. For instance,
this piece of code:
```
auto x = std::make_unique<X>(some_X);
```
would either get the extra dereference or not depending on which headers
were picked for the C++ stdlib.
The extra dereference was tripping up the liveness checker (see later in
the stack), and probably most checkers too.
Reviewed By: mbouaziz
Differential Revision: D13991130
fbshipit-source-id: 462923595
Summary:
`make clean` would fail calling `xcodebuild` to clean. It also calls `ant
clean`. Instead, do the cleaning ourselves by deleting the appropriate stuff so
that it's faster and more reliable.
Also clean the OCaml build before cleaning the tests as that's what more
important most of the time.
Also don't delete the man pages as part of `make clean` as they are checked in.
Also fix ant's Makefile so that the build points at the right .class files (the
path to the .class files was wrong so `make -C infer/tests/build_systems/ant`
would start `ant` for *each* file instead of once).
Reviewed By: martintrojer
Differential Revision: D13956218
fbshipit-source-id: bce27fe11
Summary:
Before, the liveness pre-analysis would place extra instructions in the
CFG for either:
1. marking an `Ident.t` as dead, or
2. marking a `Pvar.t` as `= 0`
But we have no way of marking pvars dead without setting them to 0. This
is bad because setting pvars to 0 is not possible everywhere they are
dead. Indeed, we only do it when we haven't seen their address being
taken anyway. This prevents the following situation, recorded in our tests:
```
int address_taken() {
int** x;
int* y;
int i = 7;
y = &i;
x = &y;
// if we don't reason about taken addresses while adding nullify instructions,
// we'll add
// `nullify(y)` here and report a false NPE on the next line
return **x;
}
```
So we want to mark pvars as dead without nullifying them. This diff
extends the `Remove_temps` SIL instruction to accept pvars as well, and
so renames it to `ExitScope`.
Reviewed By: da319
Differential Revision: D13102953
fbshipit-source-id: aa7f03a52
Summary:
Useful to understand the changes in the pre-analysis, or to inspect the
CFG that checkers actually get.
This means that the pre-analysis always runs when we output the dotty,
but I don't really see a reason why not. In fact, we could probably
*always* store the CFGs as pre-analysed.
Reviewed By: mbouaziz
Differential Revision: D13102952
fbshipit-source-id: 89f3102ec
Summary:
When initialising a variable via semi-exotic means, the frontend loses
the information that the variable was initialised. For instance, it
translates:
```
struct Foo { int i; };
...
Foo s = {42};
```
as:
```
s.i := 42
```
This can be confusing for backends that need to know that `s` actually
got initialised, eg pulse.
The solution implemented here is to insert of dummy call to
`__variable_initiazition`:
```
__variable_initialization(&s);
s.i := 42;
```
Then checkers can recognise that this builtin function does what its
name says.
Reviewed By: mbouaziz
Differential Revision: D12887122
fbshipit-source-id: 6e7214438
Summary:
Fix the logic for computing duplicate symbols. It was broken at some point and some duplicate symbols creeped into our tests. Fix these, and add a test to avoid duplicate symbols detection to regress again.
Also, this removes one use of `Cfg.load`, on the way to removing file-wide CFGs from the database.
Reviewed By: ngorogiannis
Differential Revision: D10173349
fbshipit-source-id: a0d2365b3
Summary:
New clang in the plugin \o/
Changes that were needed:
- (minor) Some extra AST nodes
- defining a lambda and calling it in the same line (`[&x]() { x = 1; }()`) used to get translated as a call of the literal but now an intermediate variable gets created, which confuses uninit in one test. I added another test to showcase the limitation this is hitting: storing the lambda in a variable then calling it will not get caught by the checker.
The controller you requested could not be found.: facebook-clang-plugins
Reviewed By: jeremydubreil
Differential Revision: D10128626
fbshipit-source-id: 8ffd19f3c
Martin Trojer
Radu Grigore
Jules Villard
David Lively
Dino Distefano
Nikos Gorogiannis
Ezgi Çiçek
Mehdi Bouaziz
Jeremy Dubreil
Calascibetta Romain
Nikos Gorogiannis
Sungkeun Cho
Martino Luca