Re: [pcp] qa/859 and python problem in fetchgroup wrapper?

[Nathan Scott <nathans@xxxxxxxxxx>]

----- Original Message -----
> [...]
> Yep ... tv_usec here is of type suseconds_t ...
> 
> /usr/include/sys/_types.h:typedef __int32_t   __darwin_suseconds_t;   /* [???]
> microseconds */
> /usr/include/sys/time.h:typedef __darwin_suseconds_t  suseconds_t;
> 
> Is there an "approved" way of dealing with this in Python wrapper code?
> [...]

Nothing clean - have a look at pmUnits which deals with HAVE_BITFIELDS_LTOR.
We could add a symbol like HAVE_32BIT_USEC (or some better name) and expose
that in python-land, then hook it up via the c_api like pmUnits does.  Would
need it to come all the way from configure.ac I guess.

> sure looks like a bug ... but I don't think that's anywhere near the place my
> test is blowing up.
> 

Yeah, I've written a little test exposing it and exercising the fix - it does
indeed look unrelated, you'd have seen this stacktrace ...

  File "/usr/lib64/python3.4/site-packages/pcp/pmapi.py", line 236, in __init__
    self.tv_nsec = usec
  NameError: name 'usec' is not defined


cheers.

--
Nathan