Hi, On 23/07/10 11:24, olafbuddenha...@gmx.net wrote: > On Mon, Jul 19, 2010 at 10:57:48AM +0200, Thomas Schwinge wrote: >> On Fri, Jul 16, 2010 at 12:19:21PM +0200, Emilio Pozuelo Monfort wrote: > >>> * hurd/version.h (HURD_INTERFACE_VERSION): Bumped for the >>> recently added RPCs. >> >> I don't see a need for doing this unless you conditionalize anything on >> the increased version number. The only example that I can find where >> this is done, is [glibc]/sysdeps/mach/hurd/configure.in for the ``new >> Hurd RPC interfaces supporting 64-bit file sizes'' (ChangeLog.13). >> >> Perhaps we should get rid of this one-dimensional scalar, and only use >> real functionality probes instead? > > Indeed, that's what I suggested last time this topic came up. A linear > interface version number doesn't make much sense with a non-linear > development/deployment model... > > Of course checking individual interface changes requires coming up with > a good autoconf test for the new interfaces. I hope Emilio already > started an extra thread on this...
I asked Thomas on irc (as the autoconf expert), but I guess he missed the question. So Thomas, any idea on how to check for a Hurd RPC reliably? > Also, what to do with the existing interface version number? Update it > nevertheless for any new interfaces? Leave it there but don't ever > update it? Drop it alltogether? IMHO either we remove it, or we update it with every interface change. Regards, Emilio