Re: [Boost-cmake] Labels for known failures

Subject: Re: [Boost-cmake] Labels for known failures
From: Mike Jackson (mike.jackson_at_[hidden])
Date: 2009-05-19 08:37:02

On Tue, May 19, 2009 at 7:27 AM, David Cole <david.cole_at_[hidden]> wrote:
> On Mon, May 18, 2009 at 3:50 PM, Brad King <brad.king_at_[hidden]> wrote:
>>
>> Michael Jackson wrote:
>>>
>>>  thanks for the heads up about vxl. I took those 2 scripts and adapted
>>> them for boost. I have since been able to submit an experimental Dashboard
>>> to <http://www.cdash.org/CDashPublic/index.php?project=Boost>. Look for
>>> ferb_at_[hidden] - the one that actually ran all the tests.
>>
>> Nice.
>>
>>> Also, if you click on the "site" you get some stats about the machine
>>> that built boost.. All of which is either wrong or not filled in. Are there
>>> some configuration options for this that need to be set or is the parsing
>>> code for CTest not grepping OS X machines correctly?
>>
>> IIRC it's parsed by code in the "Source/kwsys/SystemInformation.cxx"
>> source in CMake:
>>
>>
>> http://www.cmake.org/cgi-bin/viewcvs.cgi/Source/kwsys/SystemInformation.cxx?revision=1.40&root=CMake&view=markup
>>
>
> I looked at this issue briefly a couple months ago, but was too busy to make
> a ctest fix at the time. The problem with SystemInformation on Mac OSX is
> that the command lines used to gather the information do not use a full path
> to execute the programs. We should do a FIND_PROGRAM equivalent and use the
> full path to the programs because the environment of running from a crontab
> entry does not have the same PATH setup that a typical Mac terminal shell
> does...
>
> I just opened a cmake/ctest bug and assigned it to myself so it does not get
> lost in the shuffle...
> http://public.kitware.com/Bug/view.php?id=9050
>
> HTH,
> David

I hacked on CMake last night and got most of it working. I will try to
do a patch and attach it to the bug.

 I use the function sysctlbyname(..) directly in the cmake code. All
that was needed was a #include <sys/sysctl.h> and no extra link
libraries. What I need are some more sysctl -a outputs from various
Apple machines to test. Here is the list that I had to work with.

MacBook Pro Core 2 Duo
iBook G4
Mac Mini G4

When I get to work I am going to test my Mac Pro Xeon Hyperthreaded system.

Mike