This can be considered as a v2 for hyper annotation patch [1]. Dan argued that hyper-to-long mapping is considered as legacy and hyper should be mapped to long long for all new APIs. Therefore, he voted against the longlong annotation, because this should be the default anyway. I removed the longlong annotations from the original patch. This results in version-A of the attached patch 0001. But this optional annotation makes the parser more complex. One of the reasons for annotating things in the .x is to have single place to define a new procedure. But as the long/longlong annotation only affect exiting procedures and will not be added to new ones it's much simpler to store this information in the generator itself. This information it is fixed and will not change anymore. This is implemented in version-B of the attached patch 0001. I favor version-B, what's your opinion on this? Attached patch 0002 extends the apibuild script to allow the usage of long for the same fixed set as in the remote generator only. It reports an error when new functions or structs use long instead of longlong. [1] https://www.redhat.com/archives/libvir-list/2011-May/msg01434.html Matthias