[cmake-developers] Imported Locations in FindBoost.cmake

[Brad King]

Hi Folks,

For reference, there is an issue tracker entry for this problem:

 https://gitlab.kitware.com/cmake/cmake/issues/16091

Similar alternatives have been discussed there.

On 06/20/2016 01:32 AM, Andreas Weis wrote:
> On 6/19/2016 9:17 PM, Mike Gelfand wrote:
>> The suggested way to deal with this seems to be to use
>> MAP_IMPORTED_CONFIG_<CONFIG> target properties

That property is meant for client code to set when it has a different
set of configurations than those imported.  Technically that is true
here, but CMake should have better default behavior for the builtin
configurations like RelWithDebInfo.

> I'm still unsure what to make of the current
> fallback-to-first-configuration behavior. I did not find this mentioned
> anywhere in the docs, so I guess it's not a behavior one should rely upon?
> I was wondering whether it would make sense to properly define what
> CMake does in such a case. In particular since, as you mentioned,
> package configuration files might already rely on some assumptions there.

There is discussion about this in the above-linked issue.

I think switching the order of configurations is a good solution
to use until the alternatives can be considered:

 FindBoost: Make imported targets fall back to `Release`
 https://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=c9fca42f

Thanks,
-Brad