On 7/29/19 6:43 AM, Peter Maydell wrote:
> On Fri, 26 Jul 2019 at 18:50, Richard Henderson
> wrote:
>>
>> With this, we can have the function return a value from the DisasContext.
>>
>> Signed-off-by: Richard Henderson
>> ---
>> scripts/decodetree.py | 5 -
>> tests/decode/succ_function.decode | 2 ++
>> 2 files changed, 6 insertions(+), 1 deletion(-)
>> create mode 100644 tests/decode/succ_function.decode
>>
>> diff --git a/scripts/decodetree.py b/scripts/decodetree.py
>> index d7a59d63ac..4259d87a95 100755
>> --- a/scripts/decodetree.py
>> +++ b/scripts/decodetree.py
>> @@ -195,7 +195,10 @@ class MultiField:
>> """Class representing a compound instruction field"""
>> def __init__(self, subs, mask):
>> self.subs = subs
>> -self.sign = subs[0].sign
>> +if len(subs):
>> +self.sign = subs[0].sign
>> +else:
>> +self.sign = 0
>> self.mask = mask
>>
>> def __str__(self):
>> diff --git a/tests/decode/succ_function.decode
>> b/tests/decode/succ_function.decode
>> new file mode 100644
>> index 00..632a9de252
>> --- /dev/null
>> +++ b/tests/decode/succ_function.decode
>> @@ -0,0 +1,2 @@
>> +%foo !function=foo
>> +foo %foo
>> --
>> 2.17.1
>
> Could you also update the documentation in docs/devel/decodetree.rst ?
>
> This code change looks like it also now allows definitions
> of fields that specify nothing at all (ie there's no check
> that a field definition with no "unnamed_field" parts has
> a !function specifier) -- what do they do, or should they
> be made syntax errors ?
Ah good point. Should be syntax errors.
> Is one of these functions which just returns a constant
> from no input bits still a "static int func(DisasContext *s, int x)"
> taking a pointless input argument, or is it now a
> "static int func(DisasContext *s)" ? I guess from the fact
> this code doesn't change the way a call is output that it
> is the former, but would the latter be cleaner ?
Right on both counts. Because of how the loop in MultiField is set up, x will
always be given 0. I'll see about cleaning this up.
In the meantime, fyi, this is used for setting the S bit for thumb1 insns,
where S=0 iff the insn is within an IT block.
r~