Hacker News Reader

Comment by john_moscow | Hacker News Reader

I use this pattern across my codebase for the "critical functionality missing" scenario:

  #if DEBUG
  #define TODO(msg)
  #else
  #define TODO(msg) #error msg
  #endif

Works like a charm. I've got another PRERELEASE() macro that lets you do an unofficial preview build, but would break an official stable release.

View on HN

Macros cannot expand to preprocessing directives. This does not result in a #error-directive, this just results in a syntax error because of an unexpected # token. A better definition would be something that highlights that message in the error output, such as static_assert(0, msg).

That doesn't sound like it would scale at all for any more than a couple of devs.

Neither do TODO comments. This is just for a short-term "don't forget this piece before calling the current job done".

For anything more long-term you want to use a proper task management system with priorities, deadlines, dependencies and so on.

Slight tweak:

    #if DEBUG
    #define TODO(msg) #warning msg
    #else
    #define TODO(msg) #error msg
    #endif

Nice! Isn't this exactly assert in C++? (the following is from https://en.cppreference.com/w/cpp/error/assert)

    #ifdef NDEBUG
    #define assert(condition) ((void)0)
    #else
    #define assert(condition) /*implementation defined*/
    #endif

    ...
    assert(("There are five lights", 2 + 2 == 5));
    ...

    test: test.cc:10: int main(): Assertion `((void)"There are five lights", 2+2==5)' failed.

Take it as a token that you are doing something right. :)

Assert's run-time. This is compile-time (i.e. all instances are guaranteed to trigger during build). C++ equivalent would be a static_assert, but it has a totally different use case like this:

  static_assert(sizeof(StructUsedInSomeBinaryProtocol) == SomeConstantExpectedByTheOtherSide, "<...>")

We used to do static asserts with a preprocessor macro that typedefed an array with size expressed as: (your condition) ? 1 : -1. Point being, static asserts have the use case you want them to have; I see no reason to not have a TODO macro that expands to static_assert(false, "...").

Reply on news.ycombinator.com