Recent searches
Search options
Retro C++ quiz #25
template<int i = 3 > 4 > // Well-formed?
class X {};
Without checking:
A. Yes
B. No
C: Show answers
The answer is B
This is not well formed code, this is another variation on the maximal munch problem https://stackoverflow.com/q/28354108/1708801
This particular case is covered in http://eel.is/c++draft/temp.param#15
Adding parens fixes it
template<int i = (3 > 4) >
class X {};
I also wrote longer piece on maximal munch a while ago: https://shafik.github.io/c++/maximal%20munch/2020/12/28/maximal_munch_and_cpp.html
@shafik I feel like this is a trick question given it's named "retro"....
Retro means that there are my old quizzes I am recycling.
I have been doing this like five years now and some really good once folks have not seen especially since I moved from Twitter.
@shafik ah, but the answer may have changed depending on how old that was 
@shafik I guess the edge case I'm thinking of that changed would have required a >> in there somewhere instead of naked >s
@shafik
1) was this meant to refer to the current paragraph 19 of that section?
p19:
> When parsing a default template argument for a constant template parameter, the first non-nested > is taken as the end of the template-parameter-list rather than a greater-than operator.
2) i thought the max-munch rule was only a phase 3 thing...?
3) Without [temp.param] p19, we would just follow the grammar like so: if a template-parameter is a parameter-declaration with an initializer-clause, then a '>' in that initializer-clause would be parsed as part of an assignment-expression (as usual for assignment-expressions outside this special context).
...
@shafik
like... if there were a max-munch rule for the parsing in phase 7, wouldn't it be something like, "the next syntactic construct is the longest sequence of tokens that could constitute one of the grammar terms allowed in its context", in which case...
isn't p19 *preventing* max-munch (i.e. matching of the longest possible assignment-expression)?