[cfe-dev] Visiting statements of a CFG one time

[Andrew Sutton via cfe-dev]

I ran into exactly this problem last week. I wanted to visit top-level
statements (and expressions since expression-statements are just
expressions) and recursively work through their ASTs. I just wrote a small
pre-pass over the CFG elements that would construct a set of top-level
statements by noting which expressions occurred as subexpressions and then
extract the statements from there.

There's probably a better way to do this, but here's what my implementation
of that function looks like.

    void findStatements(CFGBlock *B, llvm::SmallVectorImpl<const Stmt *>
&SS) {
      llvm::SmallDenseMap<const Stmt*, bool> Map;

      // Mark subexpressions of each element in the block.
      for (auto I = B->begin(); I != B->end(); ++I) {
        CFGElement E = *I;
        if (auto SE = E.getAs<CFGStmt>()) {
          const Stmt *S = SE->getStmt();
          for (const Stmt *K : S->children())
            Map[K] = true;
        }
      }

      // Any expressions not in Map are statements.
      for (auto I = B->begin(); I != B->end(); ++I) {
        CFGElement E = *I;
        if (auto SE = E.getAs<CFGStmt>()) {
          const Stmt *S = SE->getStmt();
          if (Map.find(S) == Map.end())
            SS.push_back(S);
        }
      }
    }

Andrew Sutton


On Tue, Sep 8, 2020 at 9:44 AM Stefan Schulze Frielinghaus via cfe-dev <
cfe-dev at lists.llvm.org> wrote:

> Hi all,
>
> I'm trying to iterate over all top-level functions of a C file and perform
> control-flow analysis of each function body.  My current approach looks as
> follows:
>
> class MyASTVisitor : public DeclVisitor<MyASTVisitor> {
>   ASTContext &Context;
> public:
>   explicit MyASTVisitor(ASTContext &Context) : Context(Context) { }
>
>   void VisitFunctionDecl(FunctionDecl *FD) {
>     if (!FD->hasBody())
>       return;
>
>     FD->getNameInfo().getName().dump();
>
>     const CFG::BuildOptions opts;
>     auto CFG = CFG::buildCFG(FD, FD->getBody(), &Context, opts);
>     CFG->dump(Context.getLangOpts(), true);
>     // perform the actual analysis over the CFG
>   }
> };
>
> class MyConsumer : public ASTConsumer {
>   MyASTVisitor Visitor;
> public:
>   explicit MyConsumer(ASTContext &Context) : Visitor(Context) { }
>
>   bool HandleTopLevelDecl(DeclGroupRef DR) override {
>     for (auto *I : DR)
>       Visitor.Visit(I);
>     return true;
>   }
> };
>
> What confuses me right now is how the CFG is structured.  Consider the
> following C code:
>
> extern int *pi;
> extern unsigned char *pc;
> void foo(void) {
>     *(pi=(int *)pc, pi) = 42;
> }
>
> The dump of the corresponding CFG looks as follows:
>
>  [B2 (ENTRY)]
>    Succs (1): B1
>
>  [B1]
>    1: pi = (int *)pc
>    2: pi (ImplicitCastExpr, LValueToRValue, int *)
>    3: ... , [B1.2]
>    4: *([B1.3]) = 42
>    Preds (1): B2
>    Succs (1): B0
>
>  [B0 (EXIT)]
>    Preds (1): B1
>
> The single statement
>   *(pi=(int *)pc, pi) = 42;
> is broken up into 4 statements in the CFG.  That means, if I iterate over
> each
> statement of a block like
>
> for (const CFGBlock *block : *CFG) {
>   for (const auto &I : *block) {
>     if (Optional<CFGStmt> cs = I.getAs<CFGStmt>())
>       TF.Visit(cs->getStmt());
>   }
> }
>
> and the transfer functions TF recursively visit each subexpression, I will
> visit some of them twice.  My plan was actually to construct transfer
> functions
> of type ConstStmtVisitor in order to examine each statement recursively.
> In
> the past I ran into a similar problem which was related to wrongly set
> CFG::BuildOptions.  However, this does not seem to be the case this time.
>
> Is it possible to construct a CFG where statements are not split?  Or is
> there
> another approach to walk over all statements of a basic block exactly one
> time?
>
> Cheers,
> Stefan
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20200908/a8a2f9f7/attachment.html>