'Verify' task fails when module *.info file contains non-UTF-8 symbols

Problem/Motivation

If one of the *.info files in a module, a theme or a profile is encoded in something other than UTF-8 or contains invalid byte-sequences, the 'Verify' task will fail with a bunch of warnings:

and it will be impossible to add new sites to the broken platform because Aegir won't have any information about the available install profiles:

Steps to reproduce

What happens is this:

1. a hosting-task Drush command will be dispatched when a user clicks 'Verify' button in the Aegir UI
2. a new Drush process will be invoked in the backend to execute provision-verify command
3. provision module will look for all available modules, profiles and themes on a platform
4. provision_parse_info_file() function will parse info-files and keep any values as-is without modifications
5. when all the work in provision-verify is done, the list of discovered packages has to be communicated between two Drush processes, so the data has to be serialized
6. to do that, _scrub_object() function is called on the list of packages and it looks like this:
   

   function _scrub_object($input) {
     return json_decode(json_encode($input), TRUE);
   }
   

7. if $inputcontains invalid UTF-8 sequences, json_encode() will return FALSE
8. as the result, Aegir does not see any packages on the platform.

Proposed resolution

provision_parse_info_file() should strip any invalid UTF-8 sequences from parsed values. There is no better place to do this.

Remaining tasks

• Find a way to check encoding when 'mbstring' extension is not available.
• Tests maybe?

User interface changes

None.

API changes

None.

Data model changes

None.