On Wed, Apr 10, 2019 at 02:24:56PM +0100, Bruce Richardson wrote: > On Wed, Apr 10, 2019 at 09:10:25AM -0400, Aaron Conole wrote: > > > > Okay - I'll look at this part more. I think I went down the path of > > explicitly setting these because the comments didn't match with what was > > occuring (for example, in the section that I changed that loops through > > all versions, only the AVX2 and Scalar were being tested on my system, > > while the comment implied SSE). > > > > I also believe that I split out the functions because of the linking > > issue (I guess the way the linker resolves the functions works properly > > when the weak versions are in a different translation unit)? I'll spend > > some time trying to get it working in a different way. > > > > Regardless, this wasn't ready for posting as 'PATCH' - I meant it as > > RFC. I don't intend to change the first two patches, though. > > > > And thank you for the all the feedback! > > > I've dug into this a bit, and I'm doing up a patch to remove the use of > weak symbols in our libraries (note, just libs, not drivers) entirely. > That's fairly easy to do, and not a big change, but should make this > problem go away. > > /Bruce
Ref: http://patches.dpdk.org/project/dpdk/list/?series=4242
