It looks like RSA broke somewhere along the line on OS X 10.9 with Apple Clang 6.0 on a Core2 Duo MacBook (x86_64). The issue is present with libcryptopp.dylib but not with libcryptopp.a.
Using the static archive is OK. Crypto++ 5.6.5 is OK. MacPort compilers on the same machine are OK. Other machines, like Linux and Solaris, are OK.
This may be related: Explicit specialization of member function template in source file. We had a lot of trouble trying to squash Clang warnings for padding decoration functions and their definitions. It was one of the first things we worked on after 5.6.5 was released.
Here's what it looks like running the test vector in question.
It looks like RSA broke somewhere along the line on OS X 10.9 with Apple Clang 6.0 on a Core2 Duo MacBook (x86_64). The issue is present with
libcryptopp.dylib
but not withlibcryptopp.a
.I went back to Issue 517,Add SHA3 OIDs for signature schemes (actually, 1 commit prior), which seems like the natural place to break it. The issue was present there, too.
Using the static archive is OK. Crypto++ 5.6.5 is OK. MacPort compilers on the same machine are OK. Other machines, like Linux and Solaris, are OK.
This may be related: Explicit specialization of member function template in source file. We had a lot of trouble trying to squash Clang warnings for padding decoration functions and their definitions. It was one of the first things we worked on after 5.6.5 was released.
Here's what it looks like running the test vector in question.