Success! was: still problems with libofx/clientuuid

[Jack]

On 2016.10.05 11:10, jeffjl at outlook.com wrote:
> On 10/4/2016 4:43 PM, Jack wrote:
> > On 2016.10.04 15:49, jeffjl at outlook.com wrote:
> >> On October 4, 2016 4:07 PM, Jack wrote:
> >>
> >> >On 2016.10.03 02:20, Thomas Baumgart wrote:
> >> >> On Monday 03 October 2016 02:33:49 jeffjl at outlook.com wrote:
> >> >>
> >> >>> On Sunday, October 2, 2016 9:31 PM, Jack wrote: I could see the
> >> >>> clientuid box on the ofx details page, I  couldn't enter  
> anything
> >> >>
> >> >>> If the box cannot be edited, it seems that the KMM build did  
> not
> >> >>> recognize that libofx supports CLIENTUID, and so disabled the  
> box.
> >> >>> And if that's true, then even if you hand edit the KMM file to  
> add
> >> >>> the CLIENTUID, KMM won't pass it to libofx because it thinks  
> it's
> >> >>> not supported.
> >> >>
> >> >> Well, the detection was fine, but the build environment did not  
> tell
> >> >> the cmopiler about it. This is now fixed in the 4.8 branch.
> >>
> >> I just updated and tried this again on my Windows build
> >> (emerge/mingw) and I still get "No" support in the configure  
> results.
> >> This is with cmake version 2.8.12.1. I'm not cmake literate but I
> >> played with it some and I don't see why it doesn't work. It finds
> >> libofx, it gets to the check_struct_has_member() call (I added a
> >> message(STATUS) just before that to make sure). I thought it might  
> be
> >> a path problem so I tried putting the full path to my libofx.h in  
> the
> >> check_struct_has_member() call and it still doesn't find it. I  
> tried
> >> copying the example from the cmake documentation, then one time and
> >> one time only it printed status saying "Performing Test
> >> HAVE_TIMEVAL_TV_SEC" even though the file path in the command was
> >> invalid. Subsequent runs of the exact same command do not print  
> that
> >> status. And I never saw a status like that for libofx. So I'm
> >> thinking my cmake is broken.
> > Did you start with a completely empty build directory?  I vaguely
> > recall that cmake caches some stuff, and I never found a way to tell
> > it not to use the cache.  I suppose you could probably figure out  
> the
> > specific file(s) to delete, but I just wiped it all out and started
> > from scratch.
> I did not start from a clean folder and you're right about the cache.  
> It
> is part of my problem. I tried deleting the cache file  
> (CmakeCache.txt)
> and then I get the "Performing Test", but the test still fails. I can
> just edit the CmakeCache.txt file and set it to get a "yes", and that
> sticks. If I get the urge I may mess with it some more, or do a clean
> build (though I'm thinking the test would still fail).
I don't understand why it should be necessary, but try with a  
completely clean build dir.  A recursive diff afterwards might show  
what else was causing the problem.