On Wed, Aug 17, 2011 at 11:46:49 +0100, Simon McVittie wrote: > On Wed, 17 Aug 2011 at 06:46:02 +0900, Charles Plessy wrote: > > I tested python-m2crypto version 0.21.1-1 on my local system, which is a > > mixture of Squeeze, Wheezy and Sid where libssl1.0.0 and libssl0.9.8 are > > installed. > [...] > > Have I misunderstood how dynamic libraries are used, or is it indeed a bug > > that a > > package depending on libssl1.0.0 seems to be using libssl0.9.8 when > > they are co-installed ? > > Based on objdump, it shouldn't have worked for you: 0.21.1-1 and 0.21.1-2 both > have a NEEDED entry for libssl.so.1.0.0 and no NEEDED entry for > libssl.so.0.9.8, and each of libssl.so.1.0.0 and libssl.so.0.9.8 has the > appropriate thing as its SONAME, so __m2crypto.so should have picked up > libssl.so.1.0.0 and nothing else. > > The only reason I can think of why it might have worked for you is if some > other library in the same process was still linked against libssl.so.0.9.8, > and the missing symbol was picked up from there? Versioned symbols are > meant to prevent that, though, and both libssl0.9.8 and libssl1.0.0 have > versioned symbols (OPENSSL_0.9.8 and OPENSSL_1.0.0 respectively), so > __m2crypto.so should have been specifically looking for an OPENSSL_1.0.0 > implementation of each dynamic symbol it uses. (And according to objdump, > it does that correctly...) > It wouldn't be looking for a OPENSSL_1.0.0 version of a symbol that's not in libssl1.0.0, though.
Cheers, Julien -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
