When running on CMSMS on windows server.
If you submit page content  "juq'à" it fails with 
Incorrect string value: '\xC3' for column 'word' at row 1

Reason:
When CMCMS php search module try to index words it uses 
- html_entity_decode that gives UTF8 ouptut (now by default since PHP 5.4.0)
and
- preg_split that will use default server encoding (so not UTF8 when running on
windows)

The consequence is that a string like 
x6A x75 x71 x27 xC3 xA0    (juq'à in UTF8 as output of html_entity_decode)
Cannot be splitted correctly by preg_split  on windows 

Resolution:
In search.tools.php change the following
$words = preg_split('/[\s,!.;:\?()+-\/\\\\]+', $phrase);
==>
$words = preg_split('/[\s,!.;:\?()+-\/\\\\]+/u', $phrase);

This way we enforce preg_split uses UTF8 as well.

Fix risk assesment:
- This fix will not change the behaviour on Linux where UTF8 is already the
default for preg_split
- This fix will not alter behaviour for any PHP >= 5.4.0 as

=> I would propose to add a check at install time to check default_charset to
UTF8 as well.

Testing:
I've done quite a bunch of test locally and remotly. Seems to solve it .

Thanks
James