On Thu, Oct 19, 2017 at 05:36:24PM +0100, Bruce Richardson wrote: > Both Linux and BSD EALs should provide exactly the same set of functions > to the user, even if in some cases the function is a stub returning > ENOTSUP. Because of this we should only need a single version.map file for > all of EAL - not one each for Linux and FreeBSD. > > Unfortunately, there was some inconsistencies in the two EALs, specifically > 4 functions were missing from the BSD interrupts files. As part of adding > the needed stubs, I also discovered that there was nothing terribly > Linux-specific about the Linux-specific rte_interrupts.h file, which > appeared to contain a superset of the BSD contents. Therefore, this file > could be made common also. > > The net result is that there are now zero BSD-specific EAL headers, and > only a single Linux one for KNI. The version.map file is now moved to a > common location also, just inside rte_eal directory. > > [While an argument could be made to put it in common, this location works > best for integration with the existing meson build-sysetm work. It also > allows the same version file to be used for non-posix-like environments > which may not use the "common" folder as much] > > > Bruce Richardson (5): > eal/bsdapp: align interrupt include file with Linux version > eal/bsdapp: fix missing interrupt stub functions > eal: merge bsdapp and linuxapp interrupt headers > eal: mark internal interrupts file as such to doxygen > eal: use a single version map file >
Compilation tested for both static and shared builds on FreeBSD 11.0 and Fedora 26. /Bruce
