On 20.07.2013 09:58, Pekka Jääskeläinen wrote: > On 07/20/2013 03:41 AM, Chuck Zhao wrote: >> Did anybody ever seems to have seen such a behavior before? > > Yes. I remember this problem has occured before. Then it was a problem > with the 'xclang' script we use to figure out the alignments (config/xclang). > Then the issue, IIRC, was that we used wrong switch in clang or llc > to pass the target triplet. It had changed in some LLVM version.
The root cause was clang failing to link a trivial C program. This was due to it not being able to find some standard library (that seems to be placed in a distribution/glibc/gcc/... specific directory). Clang has a few big switch-cases in lib/Driver/ToolChains.cpp setting this path. Somehow it ended up in one of the "default" cases, giving a wrong path. Could be even that this was because of the target triple (armv7l-..) that clang reverted to armv4 that confused the clang Driver. My memory unfortunately fails me here. I remember seeing this on ARM (only). But how was this fixed? And why didn't I add a check for the alignments to 'configure'? Perhaps we should :) Well, at least you (Chuck) are on the right path now ;) kalle ------------------------------------------------------------------------------ See everything from the browser to the database with AppDynamics Get end-to-end visibility with application monitoring from AppDynamics Isolate bottlenecks and diagnose root cause in seconds. Start your free trial of AppDynamics Pro today! http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk _______________________________________________ pocl-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/pocl-devel
