change callers of determine_charset to check for nulls, instead of calling html_supported_charset (to pre-validate charset name) and then calling this (which has to run through a list of names anyway).
The parser depends on zend_html, which means it's impossible
to write parser unit tests without compiling all of hphp. This
separates the pieces that use runtime structures into ext_string, so
the rest can be moved to util/zend. There was a raise_warning call
indirectly from string_html_decode (in determine_charset), but the
runtime use of string_html_decode already checks that it is a
supported charset and otherwise throws NotImplementedException, so I
removed this. Another diff will do the actual move to util.
This change is mostly for FB internal organizational reasons.
Building is not effected beyond the fact that the target now
lands in hphp/hhvm/hhvm rather than src/hhvm/hhvm.
steveo