Flaky test: `FutureAwaiterTest.testWithFutureCancel`

[c8ef]

• https://github.com/alibaba/async_simple/actions/runs/23145631498/job/67232913273
• https://github.com/alibaba/async_simple/actions/runs/22401052059/job/64930317070

[ChuanqiXu9]

CC @poor-circle

[c8ef]

no worries, I have a fix for this, will post it tonight.

[poor-circle]

no worries, I have a fix for this, will post it tonight.

Do you have any ideas for that? It's a bug of library or test case?

[poor-circle]

I reopen it because we should check and avoid sleep time overflow in async_simple::coro::Sleep, @c8ef Do you have interesting about this?

[c8ef]

I reopen it because we should check and avoid sleep time overflow in async_simple::coro::Sleep, @c8ef Do you have interesting about this?

Two directions come to mind: introducing sanitizers in CI, and adopting safe_duration_cast from a third-party library such as fmt. WDYT?

[poor-circle]

I reopen it because we should check and avoid sleep time overflow in async_simple::coro::Sleep, @c8ef Do you have interesting about this?

Two directions come to mind: introducing sanitizers in CI, and adopting safe_duration_cast from a third-party library such as fmt. WDYT?

Add a dependency for third-party library maybe too heavy. Could we just check it manually?

[c8ef]

I reopen it because we should check and avoid sleep time overflow in async_simple::coro::Sleep, @c8ef Do you have interesting about this?

Two directions come to mind: introducing sanitizers in CI, and adopting safe_duration_cast from a third-party library such as fmt. WDYT?

Add a dependency for third-party library maybe too heavy. Could we just check it manually?

By "adopt" I mean we can vendor a single header - no need to pull in a whole dependency. I'm not a chrono expert, so I'm not confident reviewing it myself; adopting a lightweight, mature implementation seems more appropriate to me.

[poor-circle]

I reopen it because we should check and avoid sleep time overflow in async_simple::coro::Sleep, @c8ef Do you have interesting about this?

Two directions come to mind: introducing sanitizers in CI, and adopting safe_duration_cast from a third-party library such as fmt. WDYT?

Add a dependency for third-party library maybe too heavy. Could we just check it manually?

By "adopt" I mean we can vendor a single header - no need to pull in a whole dependency. I'm not a chrono expert, so I'm not confident reviewing it myself; adopting a lightweight, mature implementation seems more appropriate to me.

Great, could you make a pr for this bug?

[c8ef]

I reopen it because we should check and avoid sleep time overflow in async_simple::coro::Sleep, @c8ef Do you have interesting about this?

Two directions come to mind: introducing sanitizers in CI, and adopting safe_duration_cast from a third-party library such as fmt. WDYT?

Add a dependency for third-party library maybe too heavy. Could we just check it manually?

By "adopt" I mean we can vendor a single header - no need to pull in a whole dependency. I'm not a chrono expert, so I'm not confident reviewing it myself; adopting a lightweight, mature implementation seems more appropriate to me.

Great, could you make a pr for this bug?

Will do, but not right now - I'm currently more interested in adding uthread support for windows : )