Community
Participate
Working Groups
class Strategy{ public: enum _Ability { IDIOT, NORMAL, CHEAT }; Strategy( _Ability a ); }; int main(int argc, char **argv) { Strategy * Strategies[3] = { new Strategy( Str[CTRL+SPACE] } With the above code, doing a content assist where indicated, will actually erase the existing Str and leave nothing in its place. (Expect content assist to complete "Strategy")
When I run this under the debugger, it works fine. There is a race condition in the UI code here.
I'm going to look at a bunch of content assist bugs
If you change the code to have 2 constructors, it works fine. There are 2 runnables here, and one hides() the other's work.
I believe this is a result of the code that gives us the slightly different functionality for function proposals when you have no prefix and partial parameter information (bug 50807). The problem is that completionOnConstructorReference is not looking for the right things and is using the result.getIndexOfNextParameter() inappropriately (leading to the behaviour desired for 50807 which is not want we want here)
fixed in HEAD
Reopening. Please address in AST2 branch as well.
fix has been applied to the cdt_ast2 branch
Regression test ContentAssistTests._testBug72824() is failing.
Content assist doesn't seem to be working in the array initializer.
Created attachment 58699 [details] proposed patch Patch includes a quick change to the GNUCPPSourceParser to allow for completions in the array initializer. Also, since the J-Unit test for this bug requires it, the patch includes changes to include enumerators in class scoped reference. Example: Strategy::CH//completion here results in CHEAT
CCing an active committer :)
This one got hidden in the sands of time. Putting back on cdt-core. If Toni doesn't have time for it, I'll take a look tomorrow.
I didn't realize there were still cdt bugs assigned to me. I saw Bryan's patch go by, but have been busy lately and I guess the bug wasn't glamorous enough to look at right away :)
Sure, I'll take a look.
I'll commit after M5.
Committed. Thanks, Bryan.