Regression introduced in c1a5a70b51f12dedf354102217c7cd4247ed3a4b. If double-padding is applied, the length of the mask on the wire may be smaller than libXi's mask_len. When copying, only the wire length must be copied, with the remainder set to 0. When advancing to the button labels, the wire length matters, not libXi's internal length.
Reported-by: Jeremy Huddleston <[email protected]> Signed-off-by: Peter Hutterer <[email protected]> --- I think this one should fix the issue. You're seeing this because the ifs in pad_to_double get triggered on your box, but not on mine. That's a separate issue I guess, shouldn't trigger on OS X. nonetheless, this bug would be present on MIPS N32 too. src/XExtInt.c | 11 +++++++++-- 1 files changed, 9 insertions(+), 2 deletions(-) diff --git a/src/XExtInt.c b/src/XExtInt.c index 89c0894..0c64f9a 100644 --- a/src/XExtInt.c +++ b/src/XExtInt.c @@ -1610,12 +1610,14 @@ copy_classes(XIDeviceInfo* to, xXIAnyInfo* from, int *nclasses) int struct_size; int state_size; int labels_size; + int wire_mask_size; cls_wire = (xXIButtonInfo*)any_wire; sizeXIButtonClassType(cls_wire->num_buttons, &struct_size, &state_size, &labels_size); cls_lib = next_block(&ptr_lib, struct_size); + wire_mask_size = ((cls_wire->num_buttons + 7)/8 + 3)/4 * 4; cls_lib->type = cls_wire->type; cls_lib->sourceid = cls_wire->sourceid; @@ -1623,10 +1625,14 @@ copy_classes(XIDeviceInfo* to, xXIAnyInfo* from, int *nclasses) cls_lib->state.mask_len = state_size; cls_lib->state.mask = next_block(&ptr_lib, state_size); memcpy(cls_lib->state.mask, &cls_wire[1], - cls_lib->state.mask_len); + wire_mask_size); + if (state_size != wire_mask_size) + memset(&cls_lib->state.mask[wire_mask_size], 0, + state_size - wire_mask_size); cls_lib->labels = next_block(&ptr_lib, labels_size); - atoms =(uint32_t*)((char*)&cls_wire[1] + cls_lib->state.mask_len); + + atoms =(uint32_t*)((char*)&cls_wire[1] + wire_mask_size); for (j = 0; j < cls_lib->num_buttons; j++) cls_lib->labels[j] = *atoms++; -- 1.7.7.6 _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
