On Fri, May 27, 2011 at 01:12:11PM +0100, Peter Maydell wrote: > This patch series enhances the type checking of TCG values done when > compiling with debugging enabled, so that it can detect confusion of > TCGv_ptr values with whichever of TCGv_i32 and TCGv_i64 corresponds to > the pointer-width type on the compile host. This means that such > errors will always be compile failures, rather than only failing on > one of the two kinds of compile host. > > In particular, this would have caught the recent compile error > I introduced... > > I've tested this on both 32 and 64 bit hosts, with a full compile > for all targets. Note that to get a clean debug compile for both > hosts (whether with or without this patchset) you'll need to have > applied the following recent patches which all fix compile problems: > > * http://patchwork.ozlabs.org/patch/97559/ > (target-arm: Fix compilation failure for 64 bit hosts) > * http://patchwork.ozlabs.org/patch/97418/ > (target-s390x: Fix wrong argument in call of tcg_gen_shl_i64()) > * http://patchwork.ozlabs.org/patch/96665/ > (ppc: Fix compilation for ppc64-softmmu) > > Peter Maydell (2): > tcg/tcg-op.h: Fix prototypes for ld/st functions on 64 bit hosts > tcg: If DEBUG_TCGV, distinguish TCGv_ptr from TCGv_i32/TCGv_i64 > > tcg/tcg-op.h | 48 +++++++++++++++++++++++++++--------------------- > tcg/tcg.h | 52 ++++++++++++++++++++++++++++++++++------------------ > 2 files changed, 61 insertions(+), 39 deletions(-) >
Thanks, both applied. -- Aurelien Jarno GPG: 1024D/F1BCDB73 aurel...@aurel32.net http://www.aurel32.net