Summary: public
Refactor compute_results_to_parent to call collect_res_trans directly there.
So the job of PriorityNode.compute_results_to_parent is the following:
Given list of [res_trans_1; res_trans_2; ... ; res_trans_n],
1. Connect nodes linearly (ie. res_trans_1 with res_trans_2, res_trans_2 with res_trans_3) etc.
If res_trans_k has no nodes, try connecting res_trans_k-1 with res_trans_k+1 (this is what collect_res_trans always did)
2. Collect all ids, instrs and exps into one res_trans
3. Create node if needed and put there same ids and instrs as usual.
4. Connect new node with existing ones
4. Return new node structrue with exps = []
This will make function usable in more places so we don't have to create nodes over and over again
Reviewed By: ddino
Differential Revision: D2696351
fb-gh-sync-id: 9f4a9b4
Summary: public Make it uniform with other function call translations. This is just reordering, no functional change intended
Reviewed By: dulmarod
Differential Revision: D2696370
fb-gh-sync-id: 63656e3
Summary: public
Nodes in these functions can be created later (after translation of subexpressions is done).
Change triggers many differences in .dot files, but they are all about different numbering of
nodes.
There is no functional change in this diff.
Reviewed By: dulmarod
Differential Revision: D2699703
fb-gh-sync-id: a7b8d2a
Summary: public
In both functions, there is no need to pass anything in succ_nodes since nodes will be connected later anyway.
This will allow to create node later (will be done in separate diff)
Reviewed By: dulmarod
Differential Revision: D2699693
fb-gh-sync-id: 4396d31
Summary: public
Add qualifiers to global varible names. It affects both
normal global vars and class static fields
Reviewed By: dulmarod
Differential Revision: D2699927
fb-gh-sync-id: 1471faf
Summary: public
`this` can't be null in C++ methods, make backend aware of it.
Behavior for other languages remains the same
Reviewed By: dulmarod
Differential Revision: D2668945
fb-gh-sync-id: c85acbf
Summary: public Turn off this 'feature' without killing the code (yet)
We should be fine without it due to CAstProcessor that populates all location information.
Reviewed By: dulmarod
Differential Revision: D2699651
fb-gh-sync-id: cd7a184
Summary: public
Infer gets confused with this cast. It happens when objects try
to access superclass field/method, but we shouldn't change type in
this case (for the sake of backend)
Reviewed By: dulmarod
Differential Revision: D2663905
fb-gh-sync-id: bbf1cb2
Summary: public
I realized that this code was returning false for `android.content.Context <: android.content.Context`.
Reviewed By: cristianoc
Differential Revision: D2699112
fb-gh-sync-id: 4c01c42
Summary: public
This allows to run the checker and get feedback about potential expensive call stacks without having to annotate first all the methods that are overriding PerofrmanceCritical-annotated methods
Reviewed By: cristianoc
Differential Revision: D2693556
fb-gh-sync-id: cb60278
Summary: public
Currently payloads for different analyses are stored in a disjoint union.
After this diff, a record is used to enable running different analyses using the same spec files.
Reviewed By: jvillard
Differential Revision: D2695566
fb-gh-sync-id: 3d2886d
Summary: public Translate template instantiations exported by clang.
There is no need to deal with templated code - clang creates one class
for each implementation.
Reviewed By: dulmarod
Differential Revision: D2686269
fb-gh-sync-id: 9249a00
Summary: public
Not finding C++ method in decl map was considered impossible to happen.
It turns out it can happen because ast exporter wasn't exporting template
instantiations.
So instead, pick a default when method decl is not found. This will be right in
most cases and will allow us to analyze rest of the function without crashing
frontend
Reviewed By: dulmarod
Differential Revision: D2695653
fb-gh-sync-id: 4077c59
Summary: public
Using clang's method resolution. This means that, in method calls, clang gives you a pointer to the declaration of the method.
In some cases though, clang doesn't find the right method. For example, when it finds a method in a category, we
need to make it into a method in the corresponding class, because that's how we treat categories in Infer. Moreover,
when it finds a method in a protocol, that is not useful for us, since the implementation will be in some class. Finally,
sometimes the call is on an object of type id, in which case clang doesn't know what is the correct declaration. In
those cases, we fall back to what we were doing before of approximating the method resolution. We also refactor
some of the code.
Reviewed By: akotulski
Differential Revision: D2679766
fb-gh-sync-id: b79bb85
Summary: public
Support MaterializeTemporaryExpr which happens often in real life C++.
For example, it will happen in this code:
std::vector<int> v;
v.push_back(1);
// it's because std::vector<int>::push_back(const int &)
Strategy is to create variable that will store value of init expression (to provide storage)
and then return variable as a result of an expression
Reviewed By: ddino
Differential Revision: D2674340
fb-gh-sync-id: 077ed6a
Summary: public
Update facebook-clang-plugins. Changes:
- [major] new clang
- dump class template instantiations
- change naming for qual names of anonymous structs
Changes are due to:
1. Different setter properties are exported for defaults (more correct). Fix the test that was wrong
2. New unary operator type
3. Naming for qual names changed
Reviewed By: jvillard
Differential Revision: D2685734
fb-gh-sync-id: f2c568a
Summary: public
Use the analysis summary to store call stacks from PerformanceCritical-annotated methods to Expensive-annotated methods.
This use the on demand scheduling in order to make sure that the summary of the callee is always analyzed before the callers.
Reviewed By: cristianoc
Differential Revision: D2685347
fb-gh-sync-id: ab403d9
Summary: public
With this change, it will be possible to pass to InferAnalyze a list of folders to be
searched for spec files, through a file containing those paths, separated by the newline
character.
For example, if there's a `PathList.txt` file containing
/path/to/specs/folder1
/path/to/specs/folder2
/path/to/specs/folder3
Then invoking `infer --specs-dir-list-file PathList.txt [--other_args, ...] -- <build_cmd>`
will instruct the analyzer to lookup to those three folders whenever it searches for specs.
It's important to note that since the analyzer runs in parallel from different locations, it's necessary to pass **absolute** paths, or the analyzer will fail with an error.
Reviewed By: jvillard
Differential Revision: D2668700
fb-gh-sync-id: b407a57
Summary: public
Translate C++ overloaded operator calls. AST of their children looks slightly differently
which means that we have to be more permissive in params_trans.exps.
Difference:
CXXMemberCallExpr:
MemberExpr:
'ThisExpr' (it's part of method decl ref evaluation)
MethodDeclRef
Param1
Param2
...
CXXOperatorCallExpr
MethodDeclRef/FunctionDeclRef
'ThisExpr' (it's part of parameters list evaluation)
Param1
Param2
...
Reviewed By: dulmarod
Differential Revision: D2679503
fb-gh-sync-id: 1437f73
Summary: public
For some reason OpaqueValueExpr has extra field with subexpressions. Visit it when assigning location information
Reviewed By: dulmarod
Differential Revision: D2674708
fb-gh-sync-id: b91e142
Summary: public
We forgot that BlockExpr has BlockDecl inside, we need to visit it as well and assign proper location information to block body
Reviewed By: dulmarod
Differential Revision: D2565045
fb-gh-sync-id: c821a28
Summary: public
Add support for translation and calling c++ static methods.
They can be called in two ways:
ClassName::method()
classInstance.method()
Both of them have the same meaning, but AST produced is different
Reviewed By: dulmarod
Differential Revision: D2636489
fb-gh-sync-id: 9294a3f
Summary: public
I previously change this for transitive calls to Expensive-annotated methods but forgot to do the same when enforcing the subtyping rules
Reviewed By: cristianoc
Differential Revision: D2667142
fb-gh-sync-id: 0de5a34
Summary: public Plugin already exports fully qualified name, we don't need
to reconstruct it in ocaml again.
It wasn't used anywhere so this diff shouldn't change any behavior
Reviewed By: dulmarod
Differential Revision: D2668588
fb-gh-sync-id: 1ed1d61
Summary: public
New qual_name exporting logic in facebook-clang-plugins handles anonymous
classes in much better way. This allows us to simplify name generation for
classes that don't rely on a type of the object (which could be wrong for
methods from superclasses).
Reviewed By: dulmarod
Differential Revision: D2663810
fb-gh-sync-id: 08146b8
Summary: public
The static and global variables used in blocks don't appear in the ast as captured.
We need them however to try and find retain cycles involving those variables.
This diff adds a way of collecting the static variables used in blocks and treat them
like we treat other captured variables to find retain cycles.
Reviewed By: ddino
Differential Revision: D2663727
fb-gh-sync-id: d5b44ec
Summary: public Fix oversight where fields of base classes were
not exported.
Reviewed By: dulmarod
Differential Revision: D2652218
fb-gh-sync-id: 75b93ed
Summary: public
When method was called from within other method,
the type of `this` parameter didn't have pointer in type.
This was due to wrong logic that stripped pointer out of the
type.
We still need to strip the type when dereferencing reference variable.
Reviewed By: dulmarod
Differential Revision: D2652012
fb-gh-sync-id: 44552ac
Summary: public so that we don't get thousands of memory leaks in debug mode.
We will still get them if we pass ml_bucket internal.
Reviewed By: jvillard
Differential Revision: D2648200
fb-gh-sync-id: 8dce66f
Summary: public
Move dealing with init expressions from VarDecl_trans to standalone function.
This will be useful for constructor initializer.
I didn't improve anything, just moved stuff around so that no test breaks.
Main change is that we pass Sil code responsible for LHS (var dereference) as
result_trans.
Currently we assume that var_result_trans will have no nodes, no instructions,
no ids. This is true for current usages (it's always Sil.Lvar expression), but in the future it
will change.
Reviewed By: dulmarod
Differential Revision: D2647108
fb-gh-sync-id: cae60b6
Summary: public Categories are now also only translated on demand.
For this we needed links from the class to the categories which we have
now.
Reviewed By: akotulski
Differential Revision: D2647072
fb-gh-sync-id: 4afdf5f
Summary: public
Load and store the inferred attributes on the analysis summaries. The next step is to use the call stack from performance critical methods to expensive methods form the summary payload instead of modifying the attributes.
Reviewed By: sblackshear, cristianoc
Differential Revision: D2644530
fb-gh-sync-id: b99a8e3
Summary: public
returnStmt_trans adds all ids and instrs to its node. It shouldn't return them to parent then.
Reviewed By: dulmarod
Differential Revision: D2636829
fb-gh-sync-id: eb554b9
Summary: public This diff changes the way we treat enums in Infer.
1. The semantics of the translation is now correct, it was a bit incorrect before.
2. We don't add the enum types to the tenv anymore, which saves a lot of disk space
and avoids errors in the backend dealing with the enum type.
Reviewed By: akotulski
Differential Revision: D2641903
fb-gh-sync-id: 6295e5f
Summary: public
We should be able to distinguish if a method is annotated with Expensive and has been automatically annotated as calling an expensive method using internally the annotation CallsExpensive in the later case.
Reviewed By: cristianoc
Differential Revision: D2640478
fb-gh-sync-id: 32a7ee9
Summary: public
This adds the following subtyping rules:
- methods that are not annotated with Expensive cannot be overwritten by a method annotated with Expensive
- methods annotated with PerformanceCritical must be overwitten by method annotated with PerformanceCritical
Reviewed By: cristianoc
Differential Revision: D2636076
fb-gh-sync-id: eb616c9
Summary: public
Just works by running the analysis bottom-up and promoting any method as virtually annotated with `Expensive` whenever one of its callee is annotated with `Expensive`
Reviewed By: cristianoc
Differential Revision: D2635242
fb-gh-sync-id: 4401be6
Summary: public
The jar file of the models was opened twice before. This should reduce a little bit the IO of the capture.
Reviewed By: cristianoc
Differential Revision: D2626251
fb-gh-sync-id: 310bef4
Summary: public
When analysing projects we would not always display the numbers of procs and
files analysed. This fixes it, and also prints the stats before the reports for
readability. This way we also don't need to print "Analysis done" anymore.
Before:
$ infer -- gcc -c hello.c
Starting analysis (Infer version git-436690cf022a16313dda8447121a5934529e6e5c)
Analysis done
hello.c:5: error: NULL_DEREFERENCE
pointer s last assigned on line 4 could be null and is dereferenced at line 5, column 3
3. void test() {
4. int *s = NULL;
5. > *s = 42;
6. }
After:
$ infer -- gcc -c hello.c
Starting analysis (Infer version git-5b7ff3ac10d58e7ffd17c44574ab330ed07cb188)
Computing dependencies... 100%
Analyzing 1 cluster. 100%
Analyzed 1 procedure in 1 file
Found 1 issue
hello.c:5: error: NULL_DEREFERENCE
pointer s last assigned on line 4 could be null and is dereferenced at line 5, column 3
3. void test() {
4. int *s = NULL;
5. > *s = 42;
6. }
Also looked at the results on buck and gradle
Reviewed By: cristianoc
Differential Revision: D2636991
fb-gh-sync-id: 32b9a7c
Summary: public
This cleans up the output of infer. Before:
$ touch empty.c
$ infer -- gcc -c empty.c
Starting analysis (Infer version git-436690cf022a16313dda8447121a5934529e6e5c)
Analysis done
No issues found
$ infer -- javac Hello.java
Starting analysis (Infer version git-436690cf022a16313dda8447121a5934529e6e5c)
Analysis done
Hello.java:4: error: NULL_DEREFERENCE
object s last assigned on line 3 could be null and is dereferenced at line 4
2. int test() {
3. String s = null;
4. > return s.length();
5. }
6. }
Analyzed 2 procedures in 1 file
$
After:
$ infer -- gcc -c empty.c
Starting analysis (Infer version git-434faa7f70f6b9498615d3ead8c12bcfec6fc553)
Analyzing 0 clusters
Analysis done
No issues found
$ infer -- javac Hello.java
Starting analysis (Infer version git-434faa7f70f6b9498615d3ead8c12bcfec6fc553)
Computing dependencies... 100%
Analyzing 1 cluster. 100%
Analysis done
Found 1 issue
Hello.java:4: error: NULL_DEREFERENCE
object s last assigned on line 3 could be null and is dereferenced at line 4
2. int test() {
3. String s = null;
4. > return s.length();
5. }
6. }
Analyzed 2 procedures in 1 file
$
Also tested with buck, gradle.
Reviewed By: cristianoc
Differential Revision: D2636969
fb-gh-sync-id: 52f06f0
Summary: public
C++ allows to have one variable declared+initialized as a condition statement.
Handle these cases properly for `if` and `while` statements.
Reviewed By: dulmarod
Differential Revision: D2625774
fb-gh-sync-id: bac95b8
Summary: public
This is an initial version of the Expensive checker which only report violations on direct calls. The main objective is to setup all the files for this new checker.
The next steps are:
1) run the checker in interprocedural mode
2) Save in the summary of a method foo() the annotation attribute Expensive if a direct callee of foo is annotated with Expensive
3) Check that Expensive is enforced by subtyping, i.e. check that non-expensive method cannot be overwritten by a method annotated with Expensive
Reviewed By: cristianoc
Differential Revision: D2629947
fb-gh-sync-id: 0e06f85
Summary: public
Adding a progress bar to the analysis phase to check progress.
The progress bar for the analysing procedures part is just printing dots at the moment and will be
improved soon when we do all the multithreading inside Ocaml.
There is also a no_progress_bar option in the scripts to hide it, for instance in CI.
Reviewed By: cristianoc
Differential Revision: D2625763
fb-gh-sync-id: daf6f96
Summary: public
This is a non-functional refactoring to add a flag to trigger the translation of the procedure description of callees. This allows to check in which cases we still need to procedure description of callees to run the analysis.
Reviewed By: cristianoc
Differential Revision: D2620402
fb-gh-sync-id: 7ef5b5f
Summary: public
Translate CXXConstructExpr that are parts of variable initialization.
Reviewed By: dulmarod
Differential Revision: D2570750
fb-gh-sync-id: 708a457
Summary: public
Type of `this` argument to c++ method call has pointer type.
It's either raw pointer (for `->`) or reference (for `.`).
It was already correct in method declaration and method parameters, but it wasn't correct in method calls. Same thing will apply to constructor expressions.
As a result of this change, we won't expand type when calling methods.
Change to ast_expressions.ml fixes problem with autogenerating getters/setters that produced lvalue types after LValueToRValue cast.
Reviewed By: dulmarod
Differential Revision: D2605756
fb-gh-sync-id: 1027600
Summary: public
The context leaks were reported multiple times. If a leaks was found on method `f()` and `g()` calls `f()`, then the same leak was report both in `f()` and in `g()`.
Reviewed By: sblackshear
Differential Revision: D2598110
fb-gh-sync-id: ca90b57
Summary: public Translate classes and protocols on demand only.
This saves space in the tenv and time. The categories still cannot be translated on demand
because there is no pointer to them from the classes. Andrzej is fixing that.
We had to make cTypes_Decl and the modules for translating classes, protocols and categories mutually
recursive. Moreover, we had to reorder the method for adding classes to the tenv to receive a decl and
return a type.
Reviewed By: ddino
Differential Revision: D2595610
fb-gh-sync-id: ca76068
Summary: public
Extends the current activity leak checker to all sort of context leaks.
Reviewed By: sblackshear
Differential Revision: D2572548
fb-gh-sync-id: 9da18e4
Summary: public
The relative path was triggering a bug in ocamlbuild where it would try to
create ".." for some reason, which already exists. Hence `make -C infer clean
checkCopyright` would always fail for instance.
Reviewed By: jeremydubreil
Differential Revision: D2596307
fb-gh-sync-id: 0dd8d9f
Summary: public
The harness was created all the time but we do not report issues comming from the harness right now. The option:
infer --android-harness -- <build command>
Allows to create the harness
Reviewed By: sblackshear
Differential Revision: D2595211
fb-gh-sync-id: c7a6dc5
Summary: public
This allow to tell Infer to skip the translation of some files. This is especially useful to skip the translation of some generated files following the syntax:
> cat .inferconfig
{
"skip_translation": [
{
"language": "Java",
"source_contains": "_SHOULD_BE_SKIPPED_"
}
]
}
Reviewed By: cristianoc
Differential Revision: D2588095
fb-gh-sync-id: 3fda816
Summary: public
In C pre-increment/decrement returns rvalue, but in C++ it returns lvalue.
Make translation aware of the difference and treat these cases differently.
Reviewed By: ddino
Differential Revision: D2575136
fb-gh-sync-id: 952c095
Summary: public
Translating structs and C++ classes only on demand. This reduces the size of the tenv and
saves analysis time.
Reviewed By: ddino
Differential Revision: D2575008
fb-gh-sync-id: b29da2a
Summary: public
Update clang plugin version that has following changes:
1. Don't dump DeclContext as a part of BlockDecl
2. Add location information to C++ constructor initializers
This diff is making infer compatible with these changes by
1. Making infer compile
2. Reading location information from c++ constructor initializers so we don't miss update to line number
Reviewed By: dulmarod
Differential Revision: D2575066
fb-gh-sync-id: 6dc594a
Summary: public
Refactor translation of C++ method calls so that we get two parts:
1. Get Sil code for method address and this expression
2. Given method address, this expression and list of paramter statements, create Call instruction
This will allow us to share more code with C++ constructor calls that do (1) differently, but
(2) will be shared
Reviewed By: dulmarod
Differential Revision: D2570712
fb-gh-sync-id: 9c6c3e4
Summary: public
Remove spurious recursion between type and function declarations in the files
generated via `atdgen`. This speeds up the compilation time of `make clang` by
almost half (one-shot measurements: 59s -> 36s).
Silence warning 39 on this generated code to make compilation pass.
Reviewed By: akotulski
Differential Revision: D2570506
fb-gh-sync-id: 8edba22
Summary: public This diff fixes incorrect mangling of captured variables in blocks. Because they are formals,
they shouldn't be mangled, but this case was not taken into account. This caused an assert false in the
example infer/tests/codetoanalyze/objc/frontend/block/block.m which wasn't caught before because there
wasn't an endtoend test for it.
Reviewed By: akotulski
Differential Revision: D2560379
fb-gh-sync-id: db500b6
Summary: public
Adds incomplete translation of constructor bodies. Treat constructors as
methods with something 'extra'.
We still don't translate initializer lists, just pass the information to cTrans
where it's ignored
Reviewed By: dulmarod
Differential Revision: D2550214
fb-gh-sync-id: 102c13a
Summary: public anonymous types have file:line in its name.
Since file is relative path, type name can have '/' in its name.
This is very fragile since we might create file wiht typename in its name (for example for methods).
Replacing '/' with '_' should make frontend more resilient to failure.
Translation of anonymous structs is still pretty fragile (due to relative path in its name),
but at least it doesn't crash frontend
Reviewed By: dulmarod
Differential Revision: D2559936
fb-gh-sync-id: 647fd7f
Summary: public
Just some refactoring and renaming of Makefile variables.
Reviewed By: jeremydubreil
Differential Revision: D2555766
fb-gh-sync-id: 5b7d4ff
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
C++ assignment operation result is lvalue, while in C it was rvalue.
This leads to different AST produced by clang for then same code!
Use language information from clang (`-x` flag) to distinguish these cases.
More specifically, let's look at following code:
int r;
int f = (r = 3);
// type of (r = 3) expression:
// C/objC -> int rvalue
// C++/objC++ -> int lvalue
Existing code did extra dereference because it was rvalue in C and there was no cast afterwards
in C++ there will be extra LValueToRvalue cast when neccesary so we don't have to do extra dereference manually
Reference:
http://en.cppreference.com/w/c/language/value_category (search for 'assignment and compound assignment operators')
NOTE: AST output doesn't change when something is hidden behind `extern "C"`, so we should use global language information
Reviewed By: ddino
Differential Revision: D2549866
fb-gh-sync-id: b193b11
Summary: public Two cases were not handled properly so far:
1. Declaration of a reference variable missed reference bit in type
2. Parameters to a function expecting T& had type T.
The way to distinguish reference types from value types is to look
whether parameter is type 'T rvalue' or type 'T lvalue' (xvalue probably as well)
Unfortunately, we can't just say 'T lvalue' = 'T&' because it would break
a lot of things in our frontend.
However, we know that when parameter to a function call has type 'T lvalue', it has to be 'T&' type.
Same applies when init_expression type is lvalue.
So, the solution is to add wrapper function that looks at results of `instruction` function and
expected expression type. Then if it's lvalue, wrap the type in reference.
Do this wrapping magic only when we know that lvalue mean reference type.
The rest of the changes is to make frontend tests pass - since we use different fields
in the AST, some of them were incorrectly set before and no one noticed.
Reviewed By: cristianoc
Differential Revision: D2549991
fb-gh-sync-id: 067f5d5
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
This diff changes following things:
1. expression_info.type_ptr has type than decl_ref_info.type_ptr for reference types. Use type from decl_ref_info as a source of truth
2. reference types need to have one extra dereference that is not in AST. Add handling for this.
3. [small refactor] create function that creates temporary variable from res_trans expression and returns new res_trans.
Some caveats:
1. types are not quite right yet (see .dot files).
2. decl_ref_info might not be set for DeclRefExpr, make frontend crash in that case to catch when this happens
This is high risk change since it changes behavior of every translation on very widely used expr.
Reviewed By: @dulmarod
Differential Revision: D2540632
fb-gh-sync-id: aa28936
Summary: @public
This removes the old way of finding variable declarations to create sil variables and replaces it with
a a new way based on the map from pointers to declarations.
Basically, every variable dereference contains a pointer to the variable declaration, with that we can
build the corresponding sil variable.
Reviewed By: @akotulski
Differential Revision: D2536000
fb-gh-sync-id: dd29cf9
Andrzej Kotulski
Jules Villard
jrm
Sam Blackshear
Cristiano Calcagno
Dulma Rodriguez
martinoluca
Dino Distefano