On Thu, Jul 13, 2006 at 01:12:47PM +0200, michel.ponceau@xxxxxxxx wrote: > I don't understand. Libvirt users must of course compile API and > application with the same libvirt.h, after updating the value of > VIR_MAX_CPUS when necessary. And go from first info element to next one by > (info+1), which adds sizeof(virVcpuInfo) to address, including correct > length of cpumap according to: > unsigned char cpumap[(VIR_MAX_CPUS + 7) / 8]; /* Bit map of usable > real CPUs. > For alignment constraints, compiler rounds up the sizeof(virVcpuInfo)when > needed. Well if you need to recompile the library to use a larger set of CPU you don't garantee API and ABI, by definition. And you will need to recompile the library because the structures passed between the app and the library change size, this will also likely affect the proxy since I expect vCPU informations to be provided (so communication between the library and proxy must change). There is certainly solutions to the scalability problem, but suggesting a change of header and recompilation of the whole software stack is really not one I would like to propose, it breaks basically all expectations an enterprise customer would have w.r.t. using the library. Unless I really misunderstood what you were suggesting, which is possible in that case reformulating it might be a good idea, because I really understood you want to make the MAX number of CPU a compile time setting. Daniel -- Daniel Veillard | Red Hat http://redhat.com/ veillard@xxxxxxxxxx | libxml GNOME XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/