Marco d'Itri <[email protected]> writes: > On Jan 20, Russ Allbery <[email protected]> wrote:
>> This also implies that there is arguably an SONAME issue with this library >> given that two versions of the library with the same SONAME don't provide >> the same symbols, but I suspect there were really, really good reasons to >> not change the SONAME. > The upstream maintainers choose to provide backward compatibility to old > binaries but not forward compatibility from old libraries. Oh, yes, that makes sense and is entirey normal. I was thinking about it the wrong way around. So the root problem is that the dependency was satisfied for the binary but there was a stray copy of the old library with the same SONAME in an earlier directory on the search path, which shadowed the correct library. -- Russ Allbery ([email protected]) <https://www.eyrie.org/~eagle/>
