I think it can't be code completion because there is no parser for this kind of file.
But even if we admit this is code completion that parses an unknown format :
- why does it monopolize cpu (99%) ?
- why once we recover cpu, the first event (a kind of editor activation I think) makes C::B consume its released 99% ?
The code-completion parser may indeed be the cause of the trouble here. Since quite some time now, it's not parsing files based on their extension (file type). It is parsing (or trying to) anything that is thrown at its way. This change was made to be able to parse the C++ headers, which have no extension (e.g. vector, map, iostream, etc).
I see two possible solutions to this:
1) make it an option (aka "the manual way")
2) parse only files (belonging to project) with known extensions (i.e. C/C++ files) but ignore this filter for #included files
I 'll have to think this thing a bit more but I see solution nr.2 is more correct...