Explicitly set HAVE_IOCTLSOCKET_FIONBIO=1

[kendalharland]

Description

Adds a new CMake HAVE_IOCTLSOCKET_FIONBIO define that allows curl to build (link to definition). When this is not set, we hit the "no non-blocking..." error at the bottom of that file.

This is not a permanent solution but the build has been broken for several weeks and this should unblock us while we debug. It is a workaround because this define should be set automatically when building curl and this does not seem to be an issue for the upstream firebase/firebase-cpp-sdk builds which means we have a bug somewhere.

This should also help us roll out avx2 support again.

Testing

See workflow runs at:
https://github.com/thebrowsercompany/firebase-cpp-sdk/actions?query=branch:kendal/sync-this-fork

Type of Change

Place an x the applicable box:

• Bug fix. Add the issue # below if applicable.
• New feature. A non-breaking change which adds functionality.
• Other, such as a build process or documentation change.

---

[darinf]

It looks like this whole workflow is windows specific, so LGTM!

[compnerd]

Interesting; would you happen to have the logs to see why it is suddenly failing? I wonder if this is a CMake regression or something else.

[kendalharland]

Interesting; would you happen to have the logs to see why it is suddenly failing? I wonder if this is a CMake regression or something else.

Sure, pretty much any of these logs: https://github.com/thebrowsercompany/firebase-cpp-sdk/actions/workflows/bcny-firebase.yml?query=is:failure.

The failure is very consistent: D:\a\firebase-cpp-sdk\firebase-cpp-sdk\BinaryCache\firebase\external\src\curl\lib\nonblock.c(89,1): error C1189: #error: "no non-blocking method was found/used/set" [D:\a\firebase-cpp-sdk\firebase-cpp-sdk\BinaryCache\firebase\external\src\curl-build\lib\libcurl.vcxproj]

[compnerd]

Sorry, I meant the CMake configure logs that indicate why the test for the HAVE_IOCTLSOCKET_FIONBIO is failing now.

[kendalharland]

Sorry, I meant the CMake configure logs that indicate why the test for the HAVE_IOCTLSOCKET_FIONBIO is failing now.

Oh, no I do not have those. I have not been able to find any meaningful difference in the cmake logs yet when I look at the upstream builds vs our own.

[compnerd]

That is odd, I would expect it to have the full log of why the test program failed which would help us understand what is going on. Do you have copies of them on hand?

[kendalharland]

That is odd, I would expect it to have the full log of why the test program failed which would help us understand what is going on. Do you have copies of them on hand?

Possibly! But I am not sure what files you are looking for: Are these logs that are separate from what is already printed in the cmake configure step from the Github workflow run?

[compnerd]

Yes, the logs are in CMakeError.txt and CMakeOutput.txt in the output directory IIRC or something along those lines.

[kendalharland]

Yes, the logs are in CMakeError.txt and CMakeOutput.txt in the output directory IIRC or something along those lines.

Ah I see. I can try to generate these and share; We should merge this in the meantime to unblock @darinf first: Earlier I gave up trying to get this to build locally: my development machine doesn't have the necessary dependencies and some setup is needed - it fails before we get to this curl build error