{"id":107191,"date":"2022-09-16T07:00:00","date_gmt":"2022-09-16T14:00:00","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/oldnewthing\/?p=107191"},"modified":"2022-09-16T06:12:40","modified_gmt":"2022-09-16T13:12:40","slug":"20220916-00","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/oldnewthing\/20220916-00\/?p=107191","title":{"rendered":"Serializing asynchronous operations in C++\/WinRT, gotchas and final assembly"},"content":{"rendered":"

Last time, we came up with a way of making asynchronous operations run in sequence<\/a>, but I noted that there were some gotchas we need to work through.<\/p>\n

One gotcha is cancellation.<\/p>\n

In the C++\/WinRT implementation of Windows Runtime asynchronous operations, cancellation of an asynchronous operation trigger the completion callback immediately<\/a>, without waiting for the coroutine to acknowledge the cancellation. If we had used the Completed<\/code> callback to detect the completion of the coroutine, we would have started running the next coroutine before the previous one cleaned up from its cancellation.<\/p>\n

Good thing we aren’t doing that. We trigger the next coroutine at the destruction of the chained_task_completer<\/code>, which we arranged to destruct last, so everything else in the coroutine has already destructed. (Well, the parameters haven’t been destructed yet, but they were all references, so there was nothing to destruct.)<\/p>\n

The other thing to worry about is coroutine destruction. That’s where you take a suspended coroutine and call destroy<\/code> on it. This basically rips the rug out of the coroutine and destructs everything in it without letting any of the coroutine body run anything. If that happens to our coroutine while suspended at the co_await lazy_start<\/code>, then the chained_task_completer<\/code> will destruct and start running the next coroutine prematurely.<\/p>\n

Fortunately, C++\/WinRT coroutines are never abandoned. They always run to completion (possibly by throwing an exception). So at least for C++\/WinRT coroutines, we don’t have to worry about this.\u00b9<\/p>\n

What if the task_sequencer<\/code> is destructed while there are still pending coroutines?<\/p>\n

When the task_sequencer<\/code> destructs, the m_latest<\/code> shared pointer destructs, which means that the last node in the chain now has only one strong reference, namely the strong reference from the last coroutine in the chain:<\/p>\n

\ncoroutine<\/code>
\nchained_task<\/code><\/p>\n\n\n\n\n\n\n\n\n\n\n
\u2192<\/td>\ncurrent<\/code><\/td>\n\u2192<\/td>\nnext<\/code><\/td>\n<\/tr>\n
 <\/td>\n <\/td>\n\u2199\ufe0e<\/td>\n<\/tr>\n
 <\/td>\ncoroutine<\/code><\/td>\n <\/td>\nchained_task<\/code><\/td>\n<\/tr>\n
 <\/td>\ncurrent<\/code><\/td>\n\u2192<\/td>\nnext<\/code><\/td>\n<\/tr>\n
 <\/td>\n <\/td>\n\u2199\ufe0e<\/td>\n<\/tr>\n
 <\/td>\ncoroutine<\/code><\/td>\n <\/td>\nchained_task<\/code><\/td>\n<\/tr>\n
 <\/td>\ncurrent<\/code><\/td>\n\u2192<\/td>\nnext<\/code><\/td>\n\u2192<\/td>\nnullptr<\/code><\/td>\n<\/tr>\n
 <\/td>\n <\/td>\n <\/td>\n\u00d7
\n\u2191
\nm_latest\n <\/code><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

As the coroutines complete, the nodes come off the head of the linked list, and when the last one completes, the last chained_task<\/code> destructs. Everything is cleaned up.<\/p>\n

Now we can take our task_sequencer<\/code> for a spin.<\/p>\n

task_sequencer sequence;\r\n\r\nwinrt::Windows::Foundation::IAsyncOperation<winrt::hstring>\r\nMessageAfterDelayAsync(\r\n    winrt::hstring message,\r\n    winrt::Windows::Foundation::TimeSpan delay)\r\n{\r\n    co_await winrt::resume_after(delay);\r\n    std::wcout << message.c_str() << std::endl;\r\n    co_return message + L\" done\";\r\n}\r\n\r\nauto AddMessageAfterDelayAsync(\r\n    winrt::hstring message,\r\n    winrt::Windows::Foundation::TimeSpan delay)\r\n{\r\n    return sequence.QueueTaskAsync(\r\n        [=] { return MessageAfterDelayAsync(message, delay); });\r\n}\r\n\r\nvoid do_in_sequence()\r\n{\r\n    auto first = AddMessageAfterDelayAsync(L\"first\", 1s);\r\n    auto second = AddMessageAfterDelayAsync(L\"second\", 1s);\r\n    auto third = AddMessageAfterDelayAsync(L\"third\", 1s);\r\n\r\n    \/\/ Cancel the second one, just for fun.\r\n    second.Cancel();\r\n\r\n    \/\/ wait for the third one.\r\n    auto third_message = third.get();\r\n\r\n    \/\/ print the results\r\n    std::wcout << first.get().c_str() << std::endl;\r\n    std::wcout << third_message.c_str() << std::endl;\r\n}\r\n<\/pre>\n
\u00b9 In general, abandonment of a coroutine via premature destroy()<\/code> is not something most coroutine libraries deal with, so at least we’re in good company.<\/p>\n","protected":false},"excerpt":{"rendered":"
Cancellation and abandonment.<\/p>\n","protected":false},"author":1069,"featured_media":111744,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[1],"tags":[25],"class_list":["post-107191","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-oldnewthing","tag-code"],"acf":[],"blog_post_summary":"
Cancellation and abandonment.<\/p>\n","_links":{"self":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts\/107191","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/users\/1069"}],"replies":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/comments?post=107191"}],"version-history":[{"count":0,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts\/107191\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/media\/111744"}],"wp:attachment":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/media?parent=107191"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/categories?post=107191"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/tags?post=107191"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}