[pcp] ABI preservation vs. internal functions, was Re: pcp updates: log I/O, logmv, dumplog longopts, qa

Frank Ch. Eigler fche at redhat.com
Tue Apr 22 20:01:47 CDT 2014


Hi, Ken -

On Wed, Apr 23, 2014 at 10:56:17AM +1000, Ken McDonell wrote:
> [...]
> Yep +1 for __pmLogPutResult2 ... I'll fix it.

Cool.

> We don't need anything different, just the old implementation ...

(Right; though reuse of the new code could make the code simpler, and
just penalize older binaries with some memory alloc/copy overheads.) 


> > We could even inoculate against such future problems to some extent.
> > "internal" utility functions could be formally identified [...]

> Sounds like more work ... if I'd done this correctly the first time, we
> wouldn't need to be having this discussion.

Yeah, it's more of a long-term suggestion.  At this time, the
effective API exported by all our .so's is way, way bigger than the
core items documented in the programmer's guide books, or even the man
pages.  (I'd find it handy to have a place for utility functions that
are available to be reused amongst current-generation pcp binaries,
but that are not meant for use across versions or by out-of-tree
tools, so would be excluded from shared library / ABI consideration.)


- FChE



More information about the pcp mailing list