On Sat, Oct 22, 2016 at 14:43:08 +0200, László Böszörményi wrote: > Hi Emilio, > > On Sat, Oct 22, 2016 at 1:06 PM, Emilio Pozuelo Monfort > <[email protected]> wrote: > > On 21/10/16 18:20, Laszlo Boszormenyi (GCS) wrote: > >> I'd like to update libcrypto++ from 5.6.4 to 5.6.5; which is a > >> semi-transition. Packages I've tried works with both version, > >> however without binNMUs those will print this: > >> Symbol `_ZTVN8CryptoPP23FilterWithBufferedInputE' has different size in > >> shared object, consider re-linking > >> Symbol `_ZTVN8CryptoPP10HexEncoderE' has different size in shared object, > >> consider re-linking > >> Symbol `_ZTVN8CryptoPP11ProxyFilterE' has different size in shared object, > >> consider re-linking > >> > >> This matches upstream recommendation[1]: > >> "maintenance release, recompile of programs recommended" > > > > Does this bump the SONAME, or is it an ABI break without a SONAME bump? > No, the SONAME is the same. ABI should be the same, but I've found > one (more may exist) case where one (probably internal) symbol can't > be found anymore: > Generating secure encryption key. This might take some time..done > cryfs: symbol lookup error: cryfs: undefined symbol: > _ZN8CryptoPP10RandomPool34GenerateIntoBufferedTransformationERNS_22BufferedTransformationERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEEy > If it causes programs built against the older version to no longer work, then it's not internal, it is an ABI break and needs SONAME bump and package name change.
Cheers, Julien
