Summary:
It turns out HIL gets in the way of a precise heap analysis. For
instance, instead of:
```
n$0 = *&x.f
_ = delete(&x)
*&y = n$0
```
HIL tries hard to forget about intermediate variables and shows instead
```
_ = delete(&x)
*&y = *&x.f
```
Oops, that's a use-after-delete, whereas the original code was safe.
While it's easy to write SIL programs that are completely unsound for
HIL, they are not generated very often from the frontends. In fact, the
problem became apparent only when making the clang frontend translate
C++ temporaries destructors, which produces the situation above
routinely.
This diff makes the minimal amount of change to make Pulse build and
produce equivalent results (minus HIL bugs) starting from SIL instead of
HIL. The reporting sucks for now because we need to translate SIL
temporaries back into program access paths. This is done in the next
diff.
Reviewed By: mbouaziz
Differential Revision: D15824961
fbshipit-source-id: 8e4e2a3ed
Summary:
I realized that there was a discrepancy in the # of instructions between whether we run a single analysis or multiple analyses at the same time. It turns out that in biabduction, bufferoverrun and other HIL analyses we did Preanalysis step (which adds scope instructions and invokes liveness etc.) but not in others. This discrepancy results in inconsistent analysis results (e.g. in the new inefficient-keyset-iterator) that rely on instructions. We should be consistent. Hence, we now invoke Preanalysis in the frontend and remove all other uses in the rest of the checkers.
Consequently, I had to update the inefficient-keyset-checker to take the CFG resulting from Preanalysis with extra scoping instructions.
Reviewed By: mbouaziz, ngorogiannis, jvillard
Differential Revision: D15803492
fbshipit-source-id: 4e21eb610
Summary:
I was wondering what were the empty sessions and why inferbo was running twice.
Answer: the empty sessions were 'compute pre' and the second run of inferbo was the narrowing phase.
Reviewed By: ngorogiannis
Differential Revision: D15378138
fbshipit-source-id: 507a3df42
Summary:
- Makes sure that `start_session` and `finish_session` are well parenthesized
- Avoids a try finally when debug is disabled
Reviewed By: ngorogiannis
Differential Revision: D15371841
fbshipit-source-id: 340203edb
Summary:
Useful to know which disjunct is being executed. Reprinting them
wholesale is too spammy so compromise by outputting just enough to be
able to reconstruct the info "which disjunct was executed and which new
disjuncts were produced?".
Reviewed By: ngorogiannis
Differential Revision: D14753495
fbshipit-source-id: f5aa68160
Summary:
This provides a way for AI checkers to read the formals of a procedure,
or other things related to its `Procdesc.t`.
Reviewed By: mbouaziz
Differential Revision: D14258483
fbshipit-source-id: a28e28d3c
Summary:
Bundle all non-semantic-bearing instructions into a `Metadata _`
instruction in SIL.
- On a documentation level this makes clearer the distinction between
instructions that encode the semantics of the program and those that are
just hints for the various backend analysis.
- This makes it easier to add more of these auxiliary instructions in
the future. For example, the next diff introduces a new `Skip` auxiliary
instruction to replace the hacky `ExitScope([], Location.dummy)`.
- It also makes it easier to surface all current and future such
auxiliary instructions to HIL as the datatype for these syntactic hints
can be shared between SIL and HIL. This diff brings `Nullify` and
`Abstract` to HIL for free.
Reviewed By: ngorogiannis
Differential Revision: D14827674
fbshipit-source-id: f68fe2110
Summary:
Previously we would say that `lhs <= rhs` (or `lhs |- rhs`) when a
mapping existed between the abstract addresses of `lhs` and `rhs` such
that `mapping(lhs)` was a supergraph of `rhs`. In particular,
we had that `x |-> x' * x' |-> x'' |- x |-> x'`. This is not entirely
great, in particular once we get pairs of state representing footprint +
current state. I'm not sure I have an extremely compelling argument why
though, except that it's not the usual way we do implication in SL, but
there wasn't a compelling argument for the previous state of affairs
either.
This changes `|-` to be true only when `mapping(lhs) = rhs` (modulo only
considering the addresses reachable from the stack variables).
Reviewed By: jberdine
Differential Revision: D14568272
fbshipit-source-id: 1bb83950e
Summary: This helps convergence when `<=` is based on physical equality for example, and widening is implemented as `widen ~prev ~next = join prev next`.
Reviewed By: skcho
Differential Revision: D14568270
fbshipit-source-id: ded5ed296
Summary:
Open fewer sessions by wrapping AI operations together in the same HTML
node session. This allows us to also print more stuff, such as whether
the current loop computation has converged.
Reviewed By: skcho
Differential Revision: D14568274
fbshipit-source-id: d47110cf4
Summary:
This fixes (if in a hackish way) an inherently quadratic behaviour in
the disjunctive domain when analysing loops: If you start with some
disjuncts `D1 \/ ... \/ Dn` and go once around the loop, you will end up
with disjuncts `(D1 \/ ... \/ Dn) \/ (D1' \/ ... \/ Dn')` assuming that
for all `i`, `{ Di } body of loop { Di' }` (in practice there is the
added difficulty that the post of the body of the loop can be a
disjunction too instead of a single abstract state). Assuming this isn't
a fixpoint, we would then go around the loop again from `D1`, ..., `Dn`,
`D1'`, ..., `Dn'`. However we already know what the posts of `D1` to `Dn`
are!
This attempts to curb duplicate work by marking the disjuncts in `prev`
as "visited" and instructing symbolic execution to skip visited states.
Then, once convergence is detected (from within `widen` for now) we mark
again all states as unvisited so that whatever is after the loop gets
symbolically executed.
This is a hack because ideally the AI scheduler would know about
disjunctive domain and schedule individual disjuncts for analysis.
However that would be a much bigger change. Let's see if the hack is
enough for now.
Reviewed By: mbouaziz
Differential Revision: D14258491
fbshipit-source-id: 21454398c
Summary:
When joining two lists of disjuncts we try to ensure there isn't a state
that under-approximates another already in the list. This helps reduce
the number of disjuncts that are generated by conditionals and loops.
Before we would always just add more disjuncts unless they were
physically equal but now we do a subgraph computation to assess
under-approximation.
We only do this half-heartedly for now however, only taking into
consideration the "new" disjuncts vs the "old" ones. It probably makes
sense to do a full quadratic search to minimise the number of disjuncts
from time to time but this isn't done here.
Reviewed By: mbouaziz
Differential Revision: D14258482
fbshipit-source-id: c2dad4889
Summary:
The disjunctive domain shouldn't really be a set in the first place as
comparing abstract states for equality is expensive to do naively
(walking the whole maps representing the abstract heap). Moreover in
practice these sets have a small max size (currently 50 for pulse, the
only client), so switching them to plain lists makes sense.
Reviewed By: mbouaziz
Differential Revision: D14258489
fbshipit-source-id: c512169eb
Summary:
`AnalyzerNodesBasicCost` is just mapping instructions to abstract costs, it doesn't need to use AI.
Also it was keeping a map (node -> cost) for each node, this is completely removed.
Depends on D14028171
Reviewed By: ddino
Differential Revision: D14028249
fbshipit-source-id: 63f39261a
Summary:
- There is no need to use AI to compute a dot product: let's just fold over all nodes, but still do it in order (using the WTO) to report at the right place
- The previous version was computing a dot product on nodes for each node, which was quadratic, the new version is linear
- Report only once, the first time the threshold is reached (if in a loop, report at the loop head)
Reviewed By: ddino
Differential Revision: D14028171
fbshipit-source-id: b4a840c6e
Summary:
This will allow disjunctive analyzers to return sets of states as a
result instead of always returning one state. More precisely, this will
be needed for pulse when it becomes inter-procedural, if we take
summaries of functions to be disjunctive too (like, e.g., biabduction
does with several specs per function).
Reviewed By: mbouaziz
Differential Revision: D13537601
fbshipit-source-id: f54caf802
Summary:
Introduce machinery to do disjunctive HIL domains and use it for pulse,
but only in a mode that preserves the existing behaviour.
The disjunctive domain is a functor that turns any (HIL for now)
transfer function module into one operating on sets of elements of the
original domain. The behaviour of joins (and widenings, which are equal
to joins) can be chosen when instantiating the functor among 3
behaviours:
- `` `JoinAfter n`: when the set of disjuncts gets bigger than `n` the
underlying domain's join is called to collapse them into one state
- `` `UnderApproximateAfter n`: when the sest of disjuncts gets bigger
than `n` then just stop adding new states to it, drop any further states
on the floor. This corresponds to an under-approximation/bounded
approach.
- `` `NeverJoin`
The widening is always of the form ``
`UnderApproximateAfterNumIterations max_iter` for now since the only
user is pulse and I'm not sure what else would be useful.
Picking `` `JoinAfter 0` gives the same results as the non-disjunctive
domain since the underlying `join` will always be called. Make pulse use
this mode for now, and tune it in a next diff.
Reviewed By: mbouaziz
Differential Revision: D13431375
fbshipit-source-id: b93aa50e7
Summary:
A lot of functors that take a `Make{SIL,HIL}` can take a `{SIL,HIL}`
directly instead. This makes my head hurt a bit less.
Reviewed By: mbouaziz
Differential Revision: D13416967
fbshipit-source-id: eb0b33bc4
Summary:
`AccessExpression.t` and `HilExp.t` are about to become mutually
recursive, this will help distinguish the actual changes from the moving
of code around.
This deletes the file left around in the previous commit to preserve
callers of `AccessExpression`.
Reviewed By: mbouaziz
Differential Revision: D13377645
fbshipit-source-id: 71338d1f3
Summary:
It's useful for checkers to know when variables go out of scope to
perform garbage collection in their domains, especially for complex
domains with non-trivial joins. This makes the analyses more precise at
little cost.
This could have been added as a custom function call to a builtin, but I
decided against it because this instruction doesn't have the semantics
of any function call. It's better for each checker to explicitly not
deal with the custom instruction instead.
Reviewed By: jberdine
Differential Revision: D13102951
fbshipit-source-id: 33be22fab
Summary:
I hear that this scheduler is better. I want the best scheduler
possible. Also pulse's join is a bit complex so it might matter one day.
whydididothis
Reviewed By: mbouaziz
Differential Revision: D12958131
fbshipit-source-id: 3bd77ccba
Summary:
This may help running the id map bookkeeping on its own in the future
and makes the code slightly more readable in my opinion.
Reviewed By: mbouaziz
Differential Revision: D12858066
fbshipit-source-id: fea4aea63
Summary:
HIL wanted to do its own HTML printing, causing code duplication and hacks to
avoid double opening/closing files. Instead, pass a hook to print SIL
instructions or not.
This also makes the debug HTML be printed even in case of raised exceptions,
which is invaluable to debug crashes or even just reports in the case of
checkers that can raise `Stop_analysis` (pulse only for now).
This also print intermediate abstract states between instructions instead of
only at the start and end of nodes, for moar debugging.
Reviewed By: mbouaziz
Differential Revision: D12857425
fbshipit-source-id: 4ee6c88d6
Summary: Make the whole type private, introduce constructors for each variant, and deal with the consequences.
Reviewed By: da319
Differential Revision: D12825810
fbshipit-source-id: a01922812
Summary:
Seems useful to know when we're printing one instruction only, but not when we
print lots of them for readability.
Reviewed By: mbouaziz
Differential Revision: D12823481
fbshipit-source-id: 2beb339f2
Summary:
It terminates narrowing when new and old states are not comparable.
Since current narrowing does not use meet operations guaranteeing
termination of narrowing, it tries to terminate narrowing more
conservatively.
Reviewed By: mbouaziz
Differential Revision: D12815419
fbshipit-source-id: e8b45199e
Summary:
Instead of propagating a partial state give up the analysis of the
function entirely on error. The state after an error is mostly
non-sensical so until we know better just giving up makes sure the
analysis remains sensible and produce fewer spurious warnings.
Reviewed By: mbouaziz
Differential Revision: D10483979
fbshipit-source-id: 171ec8469
Summary:
New analysis in foetal form to detect invalid use of C++ objects after their
lifetime has ended. For now it has:
- A domain consisting of a graph of abstract locations representing the heap, a map from program variables to abstract locations representing the stack, and a set of locations known to be invalid (their lifetime has ended)
- The heap graph is unfolded lazily when we resolve accesses to the heap down to an abstract location. When we traverse a memory location we check that it's not known to be invalid.
- A simple transfer function reads and updates the stack and heap in a rudimentary way for now
- C++ `delete` is modeled as adding the location that its argument resolves to to the set of invalid locations
- Also, the domain has a really crappy join and widening for now (see comments in the code)
With this we already pass most of the "use after delete" tests from the
Ownership checker. The ones we don't pass are only because we are missing
models.
Reviewed By: mbouaziz
Differential Revision: D10383249
fbshipit-source-id: f414664cb
Summary:
In some error paths we may end up querying the state for the instruction
being executed, but that is only populated by biabduction. Now it's
populated by AI checkers too.
Reviewed By: jberdine
Differential Revision: D10381068
fbshipit-source-id: dca1325d7
Summary:
When the backend crashes we print which instruction/file/... we were analysing,
but because of recursion we can end up repeating that information all
the way to the toplevel call.
This makes sure we only print the innermost one, we don't care about the
calling context because the analysis is compositional.
Reviewed By: mbouaziz
Differential Revision: D10381141
fbshipit-source-id: 1c92bb861
Summary: Before this diff, the analysis would only lookup the attributes with the classname appearing in the instruction. However, it would fail to find those attributes for inherited and not overridden methods. With this diff, the attributes are now searched recursively in the super classes.
Reviewed By: mbouaziz
Differential Revision: D10007469
fbshipit-source-id: 77d721cba
Summary: This fixes some cases of false positives where the analysis will compare with the wrong overridden methods. This could later be improved with the possibility to do sub-typing comparison on the parameters.
Reviewed By: ngorogiannis
Differential Revision: D9985249
fbshipit-source-id: 7998d8619
Summary:
Now we see which file/procedure/instruction is responsible for a crash in the
backend. Biabduction and eradicate not supported yet for the instruction-level
debug.
Reviewed By: mbouaziz, da319
Differential Revision: D9915666
fbshipit-source-id: 279472305
Summary:
The internal concept of "kind" should in fact be named "severity" to match the convention used by many other tools, whereas the internal concept of "severity", i.e "HIGH", "MEDIUM" and "LOW" was never used and in any case redundant with the concept of "info", "warning", "error".
This diff maps both the "kind" and "severity" fields to value of the form "advice", "info", "warning", and "error" to be able to progressively migrate the code using the "kind" field.
Reviewed By: mbouaziz, jvillard
Differential Revision: D9187978
fbshipit-source-id: 447d89f51
Summary:
It adds relational domains to Inferbo: octagon of Apron and polyhedra of Elina.
- Each Mem domain value includes one relational value containing relations among symbols. The relational values are modified by the `Prune` and `Store` commands.
- Each abstract value includes three symbols, which represent integer value, array offset, and array size of an abstract value.
The relational domain is deactivated by default. Use the `--bo-relational-domain {oct, poly}` option for the activation, though Inferbo with the relational domains does not work at this point because some modifications of Apron and Elina we made has not been applied to their opam repositories yet.
Reviewed By: jvillard
Differential Revision: D8874102
fbshipit-source-id: 08e5883cb
Summary:
It adds relational domains to Inferbo: octagon of Apron and polyhedra of Elina.
- Each `Mem` domain value includes one relational value containing relations among *symbols*. The relational values are modified by the `Prune` and `Store` commands.
- Each abstract value includes three *symbols*, which represent integer value, array offset, and array size of an abstract value.
The relational domain is deactivated by default, so this diff should not make any differences in CI.
Use `--bo-relational-domain {oct, poly}` for the activation, though Inferbo with the relational domains does not work at this point because some modifications of Apron and Elina we made has not been applied to their opam repositories yet.
Reviewed By: mbouaziz, jvillard
Differential Revision: D8478542
fbshipit-source-id: 510ff53
Summary:
`make doc` will use `jbuilder` (which in turn uses `odoc`) to generate the
documentation for infer's modules. This is useful to browse the APIs of infer
and gives a more discoverable place to host more general documentation about
infer's internals.
Besides the actual plumbing necessary to generate the docs, this diff also
- Moves the various infer/src/*/README.md to index.mld files that make it to the generated docs
- Fixes some doc comments that would anger `ocamldoc`
Closes#435
Reviewed By: mbouaziz
Differential Revision: D8314572
fbshipit-source-id: 4a5c70e
Summary: Create mechanism for suppressing starvation reports. To do that, refactor and expose a Checkers function.
Reviewed By: mbouaziz
Differential Revision: D8259583
fbshipit-source-id: f5b5a63
Jules Villard
Ezgi Çiçek
Josh Berdine
Mehdi Bouaziz
Jeremy Dubreil
Nikos Gorogiannis
Daiva Naudziuniene
Sungkeun Cho