On 09/01/15 04:30, Christopher Felton wrote:
> On 1/8/15, 10:26 AM, Henry Gomersall wrote:
>> >On 08/01/15 15:54, Christopher Felton wrote:
>>> >>On 1/8/2015 9:07 AM, Henry Gomersall wrote:
>>>>> >>>>On 08/01/15 15:01, Christopher Felton wrote:
>>>>>>> >>>>>>On 1/8/2015 8:44 AM, Henry Gomersall wrote:
>>>>>>>>>>> >>>>>>>>>>On 08/01/15 14:00, Christopher Felton wrote:
> <snip>
>>>>>>> >>>>>>Is this what you currently have in your repo?
>>>>> >>>>Yup
>>>>> >>>>
>>>>> >>>>https://bitbucket.org/heng/myhdl/raw/641c450dac65cfa16edd42c45a43c101f3aa2db1/doc/source/manual/rtl.rst
>>>>> >>>>
>>> >>A quick look this looks good for a PR. I will fork
>>> >>tonight and test on my system(s) as well. I am a
>>> >>little confused why one of the expected output tables
>>> >>had to be changed (worried it is system dependent
>>> >>based on the random values even though it is seeded).
>> >
>> >Do you mean changed from the original source you had up? Yeah, that
>> >concerned me slightly too - sorry, I should have noted that.
>> >
>> >This would suggest it should be stable...
>> >https://docs.python.org/3/library/random.html#notes-on-reproducibility
>> >
>> >though this is worth noting too...
>> >http://stackoverflow.com/questions/11929701/why-is-seeding-the-random-generator-not-stable-between-versions-of-python
>> >
>> >Note that the always_comb code is replicated as it stands.
>> >
> I forked and tested and no errors. I think the
> duplicate code should be removed and change to
> the other test.
I think the problem is the introspection method done in always_comb.
func.func_globals.items() just returns the wrong thing under doctest
when there is a break in the code and no amount of playing around seems
to solve this.
I've implemented a workaround in which mux_1 and the signals are created
directly below the Mux code, these are then used in the test bench. I
think it's pretty good and doesn't substantially detract from the
information presentation.
Cheers,
Henry
|
