Closed junjihashimoto closed 4 years ago
Thanks a lot for reporting this. I have merged a slightly modified version of your commit: 9183ec82ecc092515e1bac5cd1fdb1824097aeb9 .
The bug is due to the following behavior. When parsing a parameter declaration, we first parse all "declaration specifiers", and then the identifier. The problem highlighted by that test is when we have a declaration A B.C
where the intended reading is "B.C
of type A
", with B.C
being a Haskell interpreter. However, if B
is a C type, we parse A
and B
as declaration specifiers, and we cannot proceed further.
I think for now it's OK to just not accept such declarations. I'll amend the quickcheck properties so that they're not considered valid to begin with.
OK, 4a4bf5edd42687298366f0a8bdad86adb986e9af should fix this. Thanks again!
Thank you for quick help and updating the test.
Quickcheck fails, when we use the seed(996343399). But current test-code does not show the failed string(
GeOIop36R9BfZ2zktb61 T.FhQ.l
). This PR uses the seed and shows the failed string.