Re: [Qemu-devel] [PATCH 04/67] target/arm: Remove offset argument to gen_exception_internal_insn

[Richard Henderson <richard.henderson@linaro.org>]

On 7/29/19 6:52 AM, Peter Maydell wrote:
> I'm not so convinced about this one -- gen_exception_insn()
> and gen_exception_internal_insn() shouldn't have the
> same pattern of function prototype but different semantics
> like this, it's confusing. It happens that both the cases
> of wanting to generate an "internal" exception happen to want
> it to be taken with the PC being for the following insn,
> not the current one, but that seems more coincidence to
> me than anything else.

I don't like "offsets", because they don't work as expected between different
modes.  Would you prefer the pc in full be passed in?  Would you prefer that
the previous patches also pass in a pc, instead of implicitly using
base.pc_next (you had rationale vs patch 2 for why it was ok as-is).

Shall we shuffle these patches later, after the Great Renaming of Things Named
PC, as discussed wrt patch 6 (pc_read and friends), so that the "offset"
parameter immediately becomes the Right Sort of PC, rather than some
intermediary confusion?


r~