A kekulization failure from a SMILES benchmark #1900
Description
In results presented at the recent ICCS regarding a SMILES benchmark (https://www.nextmovesoftware.com/products/SMILESBenchmark_ICCS_May2018.pdf), there was a single SMILES string (of the 47K written out by the CDK) that RDKit 2018.03.1 failed to kekulize. Given that (at least) CDK itself, OEChem and Open Babel all succeeded in kekulizing it, it's likely that it is kekulizable.
Unfortunately, it's a buckyball derivative (perhaps the same issue reported by Brian C at #1740?). The CDK aromatic SMILES is:
c12c3c4c5c6c7C8c5c-9c%10c%11c%12c%13c%14c%15c%11c9c%16c%17c%15c%18-c%14c%19c%20c%13c(c%12c1c%104)c%21c2c%22c%23c3c6c%24c%23c%25c%26c%22c%21c%20c%27C%19%28C%29(c%18c%30c%17c(C%168)c-%31c%32c%30c%29c(=c%27%26)c%25c%32c%24c7%31)CC=CC%28
For reference, the original (Kekule) input to CDK.
C12=C3C4=C5C6=C7C8C5=C9C5=C%10C%11=C%12C%13=C%14C%10=C9C9=C%10C%14=C%14C%13=C%13C%15=C%12C(=C%11C1=C54)C1=C2C2=C4C3=C6C3=C4C4=C5C2=C1C%15=C1C%132C6(C%14=C%11C%10=C(C98)C8=C9C%11=C6C(=C15)C4=C9C3=C78)CC=CC2