C++ coroutines: Converting among tasks that use the same promise

by Raymond Chen

If the only difference between tasks is in the awaiter, then it’s possible to convert between them without the promise even knowing what happened. We have an example of this with our simple_task and com_simple_task, which differ only in the awaiter produced by the co_await operator. This means that we can actually convert between the two by simple wrapping the promise inside the other class...

Online C++ User Group Meetings in May 2021

by Jens Weller

The monthly overview on upcoming C++ User Group meetings. A bit later then usual, seems my talk at C++now and announcing Meeting C++ 2021 has kept me a bit too busy.

Next week Tuesday: Meeting C++ online - May - Design Patterns - Facts and Misconceptions with Klaus Iglberger. If you're interested in this topic, Klaus also gives a training on Modern C++ and Design Patterns in June.

C++ coroutines: Waiting synchronously for our coroutine to complete

by Raymond Chen

Last time, we added an extension point that permitted us to respond differently to the completion of the coroutine. We’re going to put that extension point to good use by adding the ability to wait synchronously for the coroutine to complete.

C++ coroutines: Adding COM context support to our awaiter

by Raymond Chen

You may want to have awaiters that apply custom resume behavior. For example, in Windows, you are likely to want your awaiter to preserve the COM thread context. For X11 programming, you may want to the awaiter to return to the render thread if the co_await was initiated from the render thread. Today we’ll add the ability to customize the awaiter to our coroutine promise type...

C++ coroutines: Snooping in on the coroutine body

by Raymond Chen

A coroutine promise can snoop on the coroutine body by implementing a method named await_transform. Any time the coroutine body performs a co_await, the thing being awaited is passed through the await_transform method, and whatever await_transform returns is the thing that is actually awaited. This is the mysterious “We’re not ready to talk about step 1 yet” that kept reappearing in our introduction to awaitable objects...

The most popular C++ standard features

by Jens Weller

Continuing the series about the Meeting C++ survey results with a look at the standard features. Last week I compared the ISOCPP survey to the one of Meeting C++.

When I was looking into the questions I could ask in the survey tool, it came to my mind that it would be interesting to know more about the details of standards, not just asking for which standard folks use in various ways. So in this blog post, I'm going to show you the questions about standard feature usage.

C++ coroutines: How do I create a coroutine that terminates on an unhandled exception?

by Raymond Chen

Last time, we saw that declaring a coroutine as noexcept doesn’t do what you think. The noexcept specific says that production of the coroutine does not throw an exception, but it says nothing about what happens during execution of the coroutine. If an exception occurs inside the coroutine, the promise’s unhandled_exception method decides what happens...

C++ coroutines: What does it mean when I declare my coroutine as noexcept?

by Raymond Chen

Suppose you want a coroutine that terminates on unhandled exceptions, or equivalently (looking at it from the consumer side) a coroutine that never throws an exception when awaited. For regular functions, the way to say this is to put the noexcept exception specification on your function declaration...

C++ coroutines: Associating multiple task types with the same promise type

by Raymond Chen

We created two very similiar promises for hot-start and cold-start coroutines. It turns out that we can unify them...

C++ coroutines: Improving cold-start coroutines which complete synchronously

by Raymond Chen

Last time, we learned that the naïve implementation of cold-start coroutines is susceptible to stack build-up. What we want is for await_suspend to return false if the coroutine completed synchronously. This is tricky because that reintroduces a race condition where the coroutine runs asynchronously and completes at the same time we try to transition from synchronous to asynchronous behavior in the awaiter...