"Richard W.M. Jones" <rjones(a)redhat.com> wrote:
...
There are two alternatives I can think of:
Probably the simplest is to compile the RPC bindings which are
generated on Linux and supplied with libvirt in CVS and the tarball
(ie. remote_protocol.[ch]). We need to supply only the xdr_quad type
Hi Rich,
I like the idea of distributing the generated files.
Far less maintenance hassle that way.
Of course, that means we're saying developers (or anyone
running "make dist") have to use a sufficiently featureful system.
I think that is the only sane way to go.
It's the same philosophy that says you can turn up compiler-warning-
detection to the maximum and expect no warnings only on a relatively
modern and properly configured system.
If having people run distrib-building tools on inadequate
systems starts happening too often, we can add an autoconf
test to detect the losing tool(s) and warn them about it.
and a handful of 64 bit functions. I did the same thing for Windows
but in that case built my own version of a mini-XDR library with some
contributions from glibc. What we could do is bundle this mini-XDR
library with libvirt itself (or perhaps persuade gnulib to take it --
Jim?).
Doesn't hurt to ask, but once something like that is being used
by two or more projects, it's even easier to justify.
However, I confess I don't know enough about the alternatives
you mention (below) to say if it's worth pursuing.
Another, less appealing, is to look at some of the modern XDR
library
replacements. Uli suggested one, but I've lost the link at the
moment... Of course that involves porting those libraries to Mac and
Windows, which may be a load of effort in itself.