Hi Emil, On 2016-01-27, Emil Velikov wrote:
> Hi Reinette, > > Thanks for the patch. I've actually made a similar patch one some 12 > hours ago. Silly me forgot to push it - it's in master now. Thank you very much. > > On 27 January 2016 at 21:29, Chatre, Reinette <reinette.chatre at intel.com> > wrote: >> Hi Stefan, >> >> On 2016-01-27, Stefan Agner wrote: >>> [added Emil to CC] >>> Hm, I wonder why that did not happen on my system. What OS/gcc version >>> do you use? >> >> On my host I am building an image using Yocto > (http://www.yoctoproject.org/) to contain latest libdrm. In case you are not > familiar with Yocto this system will first create its own build environment > (think > sysroot) and then build this library within it. Since the build environment is > separate from host it is very sensitive to dependencies being correct and > could > explain why this issue is not encountered by folks building on and for their > hosts. > I have been doing this daily without issue since about April of last year. The > version I am using relies on gcc 4.9.2. >> > Err... sorry I'm not sure how that explains it. The missing symbol has > been part of libdrm.so since forever, so regardless which version gets > picked (the host one or not) it wouldn't error out in such a way. > > I'm leaning that it's due to different LDFLAGS - what are you using ? > The first 'offender' that comes to mind is --as-needed although I > might be mistaking. I am indeed using --as-needed Reinette
