Any branch called *core*
(which should actually be *client*; "core" is deprecated)
is a client release, and the non-client code (server, API)
is in an indeterminate state.

The configurable plan class feature is only in trunk,
where it compiles.
It hasn't been tested; I'll do some basic testing of it,
then announce it.

-- David

On 05-Jun-2012 12:59 PM, Jon Sonntag wrote:
> The server_stable branch compiles without errors doesn't have
> plan_class_spec.h so I assume it doesn't support the XML plan classes
> configuration.  So,  I tried boinc_core_release_7_0a and also
> boinc_core_release_7_0_28, but both have compilation errors:
>
>
>
> plan_class_spec.cpp: In member function 'bool
> PLAN_CLASS_SPEC::check(SCHEDULER_REQUEST&, HOST_USAGE&)':
>
> plan_class_spec.cpp:212: error: 'struct CUDA_DEVICE_PROP' has no member
> named 'dtotalGlobalMem'
>
> plan_class_spec.cpp:216: error: 'struct CUDA_DEVICE_PROP' has no member
> named 'dtotalGlobalMem'
>
> plan_class_spec.cpp:293: error: 'struct CUDA_DEVICE_PROP' has no member
> named 'dtotalGlobalMem'
>
> plan_class_spec.cpp:294: error: 'struct CUDA_DEVICE_PROP' has no member
> named 'dtotalGlobalMem'
>
> plan_class_spec.cpp:352: warning: format '%lu' expects type 'long unsigned
> int', but argument 5 has type 'cl_ulong'
>
> plan_class_spec.cpp:458: warning: format '%lu' expects type 'long unsigned
> int', but argument 5 has type 'cl_ulong'
>
> make[2]: *** [plan_class_spec.o] Error 1
>
> make[2]: Leaving directory
> `/home/boincadm/boinc/boinc_core_release_7_0a/sched'
>
> make[1]: *** [all-recursive] Error 1
>
> make[1]: Leaving directory `/home/boincadm/boinc/boinc_core_release_7_0a'
>
> make: *** [all] Error 2
>
>
>
> While I can certainly fix the bugs, the question is which version should be
> used for the boinc server?  Also, if the branches are normally newer than
> server_stable, then why is the following in server_stable/version.h: #define
> PACKAGE_STRING "BOINC 7.1.0" when the current client is only 7.0.28?
>
>
>
> Also, changing the total memory to a double will lead to problems since
> availableRam in sched_customize uses that value if it isn't set which could
> lead to some real strange values as the double is cast to an int. Also,
> sched_customize has hardcoded values for the RAM required for CUDA and ATI
> apps.  What's the point of the XML config if the min ram is hardcoded there?
> Or, is sched_customize no long used with the XML config?
>
>
>
> Jon Sonntag
>
> _______________________________________________
> boinc_dev mailing list
> [email protected]
> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to