> Excuse me if I'm missing something obvious or this has been discussed > elsewhere: Why will implementing pthreads require a soname change? > Won't it simply augment the existing ABI, or will it force some other > interface to break? If pthreads goes in a separate library as is the > case for Linux, I don't see any need for binary incompatibility, > either forwards or backwards.
The problem is that there are a few nits in the existing hurd libc ABI that involve cthreads and the current (pre-pthreads) multithreading support in libc. Those may be different or absent with pthreads. It might be avoidable, but it's not worth a lot of effort; with debian, it should not be so big a deal to recompile everything for a new ABI when no source code has to be changed. (Also, we are going to switch to libio not long after we have pthreads, and that has numerous ABI changes.)
