On Fri, Aug 06, 2010 at 09:14:20AM +0200, Andrew Beekhof wrote:
On Thu, Aug 5, 2010 at 6:04 PM, Darryl L. Pierce
<dpierce(a)redhat.com> wrote:
> I've spent the past week or so trying to get WMI implemented and
> available via MinGW so that we could take advantage of it for our
> Windows builds of Matahari.
>
> After some frustration at finding that it's been requested, but nobody's
> implemented it, since 2005,
The most important question is "why?".
What's the hurdle here? I got the impression from our meetings that
the only problem was getting patches reviewed fast enough.
Is it just the size of WMI or is there some technical reason why
mingw's method of wrapping DLL entry points doesn't work for WMI?
Size isn't an issue for us.
Unlike the mingw project, we don't need to support every WMI
function... just the 20? 30? that are needed by Matahari.
The main blocker is technical: taking the Microsoft descriptors (*.idl)
and converting them into something compilable. I've just found another
option, using Wine's widl compiler, to process the IDL files and have
made a little more progress, but still have not successfully produced
stubs that can be compiled with MinGW.
> I think for our use it might not be the
> right path to follow.
>
> I have found a lot of Win32 APIs that directly give us some of the
> information that we need, such as CPU type, archetecture and similar.
> But it's been a little more elusive to get information like the UUID of
> the mother board and CPU model.
>
> In the short term, it would probably be easier to do our Windows
> development on Windows using WMI. Then we can look toward getting WMI
> into MinGW afterward and moving our builds back to Linux at that point.
>
> Thoughts?
Not a fan to be honest.
We need to know ASAP if using WMI with mingw is possible viable.
If its not, then we don't want to be maintaining a bunch of mingw
packages in Fedora.
This is far too great a major technology shift to defer to the end of
the project - right at the point where we'll be under the most
pressure to get things stablilzed and out the door.
What I'd like to see for Monday's scrum is two WMI calls being used.
Pick one that looks/does something trivial and one that looks more complex.
Use whatever means necessary - it doesn't have to involve automatic
generation of header and .def files.
We just need to see that its possible.
Okay, so we just want a single source that can compile under MinGW and make
two WMI calls. Is that correct?
--
Darryl L. Pierce, Sr. Software Engineer @ Red Hat, Inc.
Delivering value year after year.
Red Hat ranks #1 in value among software vendors.
http://www.redhat.com/promo/vendor/