Hi Anand, I've opened a bug for this issue [1]. My first diagnostic was incorrect. Configure detect the CPU as a 386, which is quite wrong. This in turns cause a lot of issues :
This enable the x86 compat layer in liburcu which in turns barf on these: __asm__ __volatile__ ("pushfl; popl %0" : "=r" (res) : ); [...] __asm__ __volatile__ ("pushl %0; popfl" : : "r" (newflags) : "cc"); The compiler on Mac OSX has specifically been compiled to run on x86_64 (seen in config.log, --host=x86_64-apple-darwin11). The pushfl, popfl instructions are not present on x86_64, so the assembler fails... The workaround would be to force the build/host flag at configure time to this (similar to [2]) : ./configure --build=x86_64-apple-darwin11 We will have to either fix this in the configure.ac script or document this bug. Christian [1] - http://bugs.lttng.org/issues/578 [2] - https://trac.macports.org/ticket/21011 On Wed, Jun 26, 2013 at 11:50 AM, Anand Buddhdev <ana...@ripe.net> wrote: > Hi Christian, > > Were you able to figure out from the config.log files why the > architecture detection wasn't working properly? > > Anand > > On 21/06/2013 20:40, Anand Buddhdev wrote: > >> Hi Christian, >> >> I've attached both here. >> >> Anand >> >> On 21/06/2013 18:51, Christian Babeux wrote: >> >>> Hmm that's really weird :/. For some reason configure detect an i386 >>> CPU which is certainly not right. Could you post your config.log with >>> and without the --host flag (specifying which is which)? >>> >>> There might be some slight incompatibilities on OSX and the CPU >>> detection mechanism we are using... _______________________________________________ lttng-dev mailing list lttng-dev@lists.lttng.org http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev