Report connection timeout as separate event

[gjasny]

No description provided.

[yhirose]

@gjasny, thanks for the pull request. Could you take a look at the problem on GitHub Actions workflow?
https://github.com/yhirose/cpp-httplib/actions/runs/1745000655

[yhirose]

Could please revert back the changes of moving the following lines in test/test.cc?

ASSERT_TRUE(res);
ASSERT_EQ(200, res->status);

[yhirose]

According to the code added in the unit test, does this new error type occur only with AF_INET? If so, it will actually create more confusion for users using AF_UNSPEC, because they won't receive the ConnectionTimeout even if it's declared in httplib.h...

cli.set_address_family(AF_INET);

[gjasny]

According to the code added in the unit test, does this new error type occur only with AF_INET? If so, it will actually create more confusion for users using AF_UNSPEC, because they won't receive the ConnectionTimeout even if it's declared in httplib.h...

cli.set_address_family(AF_INET);

It took me a while why the test was failing on Linux. What happens is:

google.com gets resolved to both: an IPv4 and an IPv6 address. The client code loops over both and for IPv4 sets the error code to "ConnectionTimeout". Then it tries IPv6 but that's not working because the test host only has an IPv4 address. Thus the error code gets overwritten and set to "Connection".

By explicitly requesting an IPv4 lookup the test avoids the second IPv6 loop and the overwriting of the error code.

If the host would have IPv6 connectivity it would also return "ConnectionTimeout" for IPv6.

[yhirose]

@gjasny, thanks for the clear explanation. I'll merge it to the master branch soon. Thanks for your contribution!