Hi,
On 2/10/25 4:20 PM, Cornelia Huck wrote: > On Fri, Feb 07 2025, Richard Henderson <richard.hender...@linaro.org> wrote: > >> On 2/7/25 03:02, Cornelia Huck wrote: >>> And switch to using the generated definitions. >>> >>> Generated against Linux 6.14-rc1. >>> >>> Signed-off-by: Cornelia Huck<coh...@redhat.com> >>> --- >>> target/arm/cpu-sysreg-properties.c | 716 ++++++++++++++++++++++++++++- >>> target/arm/cpu-sysregs.h | 116 +---- >>> target/arm/cpu-sysregs.h.inc | 164 +++++++ >>> 3 files changed, 860 insertions(+), 136 deletions(-) >>> create mode 100644 target/arm/cpu-sysregs.h.inc >> Why are we committing generated files and not generating them at build-time? > We'd either have to carry a copy of Linux' sysregs file, or generate a > build dependency on Linux. I think we should handle this similar to the > Linux headers update, where we do an explicit update and check for > anything unexpected that might have crept in. (Same applies if we switch > to any other external source for register definitions.) Yes this was the initial intent, I mean do a manual import from linux sysreg or any other reliable source such as the JSON dump pointed out by Marc. That way we are still able to analyze potential breakages. Eric > >
