[CMake] option bug ?

[Michael Hertling]

On 07/07/2010 09:40 AM, Chris Hillery wrote:
> On Wed, Jul 7, 2010 at 12:32 AM, Michael Hertling <mhertling at online.de>wrote:
> 
>> IMO, things aren't sooo bad. ;-)
>>
>> [100 lines of explanation of how SET() behaves in 6 different ways elided]
>>
> 
> ....I think you've just proven my point. Thanks! :)

Actually, I intended to disprove your point to a certain extend. ;)

SET() works as expected except for SET(VAR "..." CACHE <TYPE> "...") and
the related OPTION() in some cases. If VAR does not have an entry in the
cache or in the current scope SET() behaves reasonable: Afterwards, both
entries exist and are equal. If VAR already has both entries SET() does
not trade off the cache for the current scope which could be considered
as reasonable, too. Nevertheless, the UNINITIALIZED thing is pretty
weird, indeed.

Regards,

Michael