Add new Task-based socket methods for UDP and reimplement existing ones using SocketAsyncEventArgs

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528

[antonfirsov]

Closes#41502, but does not change the existing APM methods, I plan to do it in a follow-up PR.

The PR is built on @geoffkizer's branch, almost all product code changes were written by him, I just added a bunch of tests, fixed an issue with Windows cancellation (4e4add1), and moved extension methods to the Socket class as per #43901.

As such, the product code needs a third pair of yes to review. @stephentoub maybe?

[antonfirsov]

@geoffkizer can you please take another look at the tests, and give a final approval if things look OK?

I had to delete my SendTo cancellation test, since it looks like there is no reliable way to enforce sendto to return EWOULDBLOCK. Flooding the kernel works on some Linuxes, but not on others, and does not work on Mac.

I opened a bunch of issues for the corner cases discovered, and referenced them in my test workarounds. Most of them are minor compatibility cases, except the first one, which is actually concerning: #47342, #47335, #47469, #47425

[geoffkizer]

One small nit above, otherwise LGTM.

[antonfirsov]

/azp run runtime-libraries-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[antonfirsov]

OuterLoop test failures are unrelated: #46619, #43207, #40798, #47528