On 05/24/12 08:43, Alon Levy wrote: > On Wed, May 23, 2012 at 08:59:22PM +0200, Gerd Hoffmann wrote: >> On 05/22/12 17:29, Alon Levy wrote: >>> We can't initialize QXLDevSurfaceCreate field by field because it has a >>> pa hole, and so 4 bytes remain uninitialized when building on x86-64, so >>> just memset. >> >> So you get valgrind warnings for the hole? why? nobody should ever >> access the hole, so the missing initialization should not hurt in theory ... > > Because we allocate this struct on the stack and then copy it over an fd > to spice, through the dispatcher pipe.
Ah, yea, copying will make valgrind complain of course ... I'll go queue it up. cheers, Gerd
