Am Donnerstag 28 März 2013, 19:43:04 schrieb Peter Maydell:
> Make sure we set the buffer to something in the softusb_read_{dmem,pmem}
> error paths, since the caller will use the buffer unconditionally.
> (Newer gcc is smart enough to spot this and complain about 'may be
> used uninitialized'.)
>
> Signed-off-by: Peter Maydell <peter.mayd...@linaro.org>
> ---
> This needs to be applied before the 'drop sysbus_add_mem' patchset to
> fix a compilation failure with newer gcc. That patchset should apply
> after this without needing rebase, I think.
>
> hw/milkymist-softusb.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/hw/milkymist-softusb.c b/hw/milkymist-softusb.c
> index d911686..9f83209 100644
> --- a/hw/milkymist-softusb.c
> +++ b/hw/milkymist-softusb.c
> @@ -131,6 +131,7 @@ static inline void
> softusb_read_dmem(MilkymistSoftUsbState *s, if (offset + len >=
> s->dmem_size) {
> error_report("milkymist_softusb: read dmem out of bounds "
> "at offset 0x%x, len %d", offset, len);
> + memset(buf, 0, len);
> return;
> }
>
> @@ -155,6 +156,7 @@ static inline void
> softusb_read_pmem(MilkymistSoftUsbState *s, if (offset + len >=
> s->pmem_size) {
> error_report("milkymist_softusb: read pmem out of bounds "
> "at offset 0x%x, len %d", offset, len);
> + memset(buf, 0, len);
> return;
> }
Acked-By: Michael Walle <mich...@walle.cc>
