[libc++][NFC] Make AssertionInfoMatcher::CheckMessageMatches Stricter

[hawkinsw]

Rather than allow for a message to be considered a match for the actual assertion if it is anywhere in the assertion text, make sure that the expected and the actual assertion are identical.

Addresses #77701

[llvmbot]

@llvm/pr-subscribers-libcxx

Author: Will Hawkins (hawkinsw)

Changes

Rather than allow for a message to be considered a match for the actual assertion if it is anywhere in the assertion text, make sure that the expected and the actual assertion are identical.

Addresses #77701

---

Full diff: https://github.com/llvm/llvm-project/pull/77721.diff

1 Files Affected:

• (modified) libcxx/test/support/check_assertion.h (+1-2)

diff --git a/libcxx/test/support/check_assertion.h b/libcxx/test/support/check_assertion.h
index 98dd95b11556e6..01e296d9138d0c 100644
--- a/libcxx/test/support/check_assertion.h
+++ b/libcxx/test/support/check_assertion.h
@@ -89,8 +89,7 @@ struct AssertionInfoMatcher {
     std::size_t found_at = got_msg.find(msg_);
     if (found_at == std::string_view::npos)
       return false;
-    // Allow any match
-    return true;
+    return found_at == 0 && got_msg.size() == msg_.size();
   }
 private:
   bool is_empty_;

[hawkinsw]

Please feel free to ignore this change if it is unwanted. The tests that I ran locally all seem to pass with the stricter checking. I'll wait to see what it looks like on a full run of CI.

[mordante]

In general I like this patch. However there seems to be unrelated changes.

[ldionne]

I am fine with this change if it passes CI (and without the unrelated changes). I thought we took advantage of this "hand wavyness" in more places, but if we don't and the CI is happy, I think this is an improvement.

[hawkinsw]

Asking for another review from @mordante just out of caution!

[hawkinsw]

Pinging @var-const given his work on #77883. I will wait for his "go" before moving forward.

[mordante]

Thanks LGTM! Let's ship it.