Better Code with C++ Attributes - Ben Saks - CppCon 2019

CppCon.org
Discussion & Comments: / cpp
Presentation Materials: github.com/CppCon/CppCon2019
-
Better Code with C++ Attributes
Modern C++ attributes such as [[noreturn]] and [[deprecated]] can be used to improve code quality and programmer productivity. They provide valuable information to the compiler, allowing it to generate more efficient code and provide more useful warning messages. C++20 is slated to include several new attributes, including [[no_unique_address]] and the design-by-contract attributes. They also let you use compiler extensions on specific toolchains without locking you into using only those toolchains.
This session explores several C++ attributes and provides examples of how to use each one to improve your programs. Although it focuses on the standard attributes, it also includes examples of non-standard attributes from specific toolchains.
-
Ben Saks
-
Videos Filmed & Edited by Bash Films: www.BashFilms.com
*-----*
Register Now For CppCon 2022: cppcon.org/registration/
*-----*

[@seditt5146]

Just letting anyone watch know that the sound clears up around the 4 minute mark, you can keep watching and the Echo goes away.

[@stephenjames2951]

Both videos by Ben Saks have been very interesting and valuable.

[@dima1250narod0ru]

Thanks a lot. It was very useful and helpful :)
It's a pity: the standard gave us portable attributes, and then we have to wrap them in macros for portability :)

[@ninepoints5932]

Ho conditional [[fallthrough]] I didn't realize that was possible :thumbsup:

[@jimlin897]

I have imaged [[no_unique_address]] for a long time WOW

[@tourdesource]

IIRC flowing off the end of a function is actually UB. `main` might be the one exception.

[@seditt5146]

37:20 I have attempted to use [[ likely ]] in a Ternary statement and it always tosses an error and won't compile. I have found no configuration which an attribute can be used in this case.

[@freddymartinez8741]

Maybe because it will be fully supported in c++20

[@Uristqwerty]

Upon reaching the part about solutions for unrecognized attributes, I was amused at the thought that it might be, itself, solved with an attribute, like [[safe_to_ignore(list, of, known, attribute, names)]]. Ah well, guess it's boring *old* technology, without any self-referential humour.

[@FF_Fanatic]

I'm thinking that since, e.g., Clang can give you "did you mean..." suggestions when you misspell something like a function or namespace, a compiler could potentially issue a warning for a misspelled attribute only if it's close enough to a known name (e.g., through Levenshtein distance).