Added a feature test to auto enable/disable CPPHTTPLIB_USE_NON_BLOCKI…

[TobiasWallner]

Summary

Add CMake feature detection for CPPHTTPLIB_USE_NON_BLOCKING_GETADDRINFO on Windows.

When HTTPLIB_USE_NON_BLOCKING_GETADDRINFO is enabled, CMake now checks whether GetAddrInfoExCancel is available before exporting CPPHTTPLIB_USE_NON_BLOCKING_GETADDRINFO through the httplib::httplib target. If the symbol is not available, the non-blocking getaddrinfo support is disabled automatically.

Motivation

Some Windows machines/toolchains do not provide GetAddrInfoExCancel, which causes downstream projects using cpp-httplib via CMake to fail compilation when CPPHTTPLIB_USE_NON_BLOCKING_GETADDRINFO is exported unconditionally.

Previously, consumers had to work around this by manually removing the interface compile definition from httplib::httplib. With this change, cpp-httplib configures itself correctly out of the box.

Testing

Verified on a Windows machine where GetAddrInfoExCancel is unavailable:

• CMake reports GetAddrInfoExCancel - not found
• HTTPLIB_IS_USING_NON_BLOCKING_GETADDRINFO is set to FALSE
• CPPHTTPLIB_USE_NON_BLOCKING_GETADDRINFO is not present on the consumer compile line
• compiled-library mode builds successfully

[yhirose]

@sum01 @jimmy-park @Tachi107 this PR looks good to me. Could you please take a look at it?

[yhirose]

@TobiasWallner could you take a look at the CI failures of test / windows with SSL and test /windows without SSL? (You don't need to worry about other failures.)

[sum01]

Wouldn't it probably be better to use either WARNING or even FATAL_ERROR here?

I'm not sure what people would prefer personally, but having the build stop makes some sense to me, as it's an explicitly enabled option that will not work/be supported.

[TobiasWallner]

Good point, i would use a warning, because the user might still want to compile it and provide the function by itself or link something that cmake does not see at this point.

[TobiasWallner]

Wouldn't it probably be better to use either WARNING or even FATAL_ERROR here?

I'm not sure what people would prefer personally, but having the build stop makes some sense to me, as it's an explicitly enabled option that will not work/be supported.

Changed it to a warning

[yhirose]

@TobiasWallner, since the master branch has progressed quite a bit, please run rebase master.

[TobiasWallner]

yes, sry i will, i just had not had the time and head for it lately

[yhirose]

@TobiasWallner could you please remove things unrelated to CMake such as b29d191?

[TobiasWallner]

@TobiasWallner could you please remove things unrelated to CMake such as b29d191?

Sry, sure done

[TobiasWallner]

@TobiasWallner could you take a look at the CI failures of test / windows with SSL and test /windows without SSL? (You don't need to worry about other failures.)

522b60a

Lets see if this correctly addresses that, I cannot reproduce the failure on my local machine, lets see what happens in CI.

[yhirose]

@TobiasWallner thanks for your fine contribution!