[PATCH] D62648: [Sema][Typo] Fix assertion failure for expressions with multiple typos

[David Goldman via Phabricator via cfe-commits]

dgoldman marked 2 inline comments as done.
dgoldman added inline comments.


================
Comment at: lib/Sema/SemaExprCXX.cpp:7713-7714
+          // Add the newly created typos to the TypoExprs list, even if they
+          // failed to apply. This allows them to be reaped although they won't
+          // emit any diagnostic.
+          SavedTypoExprs.set_union(TypoExprs);
----------------
rsmith wrote:
> What prevents diagnostics from being emitted for `TypoExpr`s we create for an outer transform that we end up discarding? Eg, in:
> 
> ```
> struct Y {};
> struct Z { int value; };
> struct A {
>   Y get_me_a_Y();
> };
> struct B {
>   Z get_me_a_Z();
> };
> A make_an_A();
> B make_a_B();
> int f() {
>   return make_an_E().get_me_a_Z().value;
> }
> ```
> 
> I'm concerned that we will:
>  * try correcting `make_me_an_E` (`TypoExpr` `T0`) to `make_me_an_A` (because that correction has the minimum edit distance) and discover that there is no `get_me_a_Z` member, creating a new `TypoExpr` `T1`, and recurse:
>    * try correcting `get_me_a_Z` (`TypoExpr` `T1`) to `get_me_a_Y` and discover that there is no `value` member
>    * no correction works: bail out of recursive step
>  * try correcting `make_me_an_E` (`TypoExpr` `T0`) to `make_me_a_B` and succeed
> 
> But now we still have `T1` in our list of `TypoExpr`s, and will presumably diagnose it (and take action below if it turned out to be ambiguous etc).
> 
Fixed by discarding the failed `TypoExpr`s


================
Comment at: lib/Sema/SemaExprCXX.cpp:7762-7787
     // Ensure none of the TypoExprs have multiple typo correction candidates
     // with the same edit length that pass all the checks and filters.
     // TODO: Properly handle various permutations of possible corrections when
     // there is more than one potentially ambiguous typo correction.
     // Also, disable typo correction while attempting the transform when
     // handling potentially ambiguous typo corrections as any new TypoExprs will
     // have been introduced by the application of one of the correction
----------------
rsmith wrote:
> What happens if an ambiguous `TypoExpr` is created as a result of one of the outer level transformations?
> 
> In that case, I think that we will try alternatives for that `TypoExpr` here, but that `TypoExpr` is not in the expression we're transforming (it was created within `RecursiveTransformLoop` and isn't part of `E`), so we're just redoing the same transformation we already did but with typo-correction disabled. This means that the transform will fail (because we hit a typo and can't correct it), so we'll accept the original set of corrections despite them being ambiguous.
So what do you recommend here? Checking for the non-ambiguity in `RecursiveTransformLoop` itself?


Repository:
  rC Clang

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D62648/new/

https://reviews.llvm.org/D62648