cmake: Add CURL_CA_BUNDLE/CURL_CA_FALLBACK/CURL_CA_PATH

[webmaster128]

No description provided.

[mention-bot]

@webmaster128, thanks for your PR! By analyzing the history of the files in this pull request, we identified @Sukender, @jzakrzewski and @billhoffman to be potential reviewers.

[Paxxi]

Looks good to me except one thing.

auto should probably be ignored on Windows as none of the paths are valid and there's really no concept of a system wide cert bundle for OpenSSL. I would check for the WIN32 variable and skip the test. It's true both for x86 and x64 despite it's name.

[webmaster128]

auto should probably be ignored on Windows as none of the paths are valid and there's really no concept of a system wide cert bundle for OpenSSL. I would check for the WIN32 variable and skip the test. It's true both for x86 and x64 despite it's name.

Do I understand correctly that this a matter of taste with no impact on the behaviour of the script?

Whenever possible, I'd prefer a general solution over a platform specific solution to avoid extra complexity that comes with case distinctions. When we set a bunch of possible paths, that do not exist on Windows, fine. But why not add a Windows path into the mix? The current code would allow that (but as you said, there is no such path).

One thing that is let's me tent to a general solution is that "being on Windows" is not a black or white question. With Cygwin and MinGW in the mix, we need to support systems that are more or less like Windows, which makes feature testing much easier than OS distinctions. When you search the Internet for CMAKE_LEGACY_CYGWIN_WIN32 you see that there was a lot of confusion about if Cygwin is a WIN32 for cmake or not (impacting our 2.8 minimum version as well). So I'd rather not be in the business of detection how much Windows-like a system is.

As far as I can see, there is no harm in finding and using given paths on a Windows-like system with a Unix-like file system.

[Paxxi]

I think you're right. I generally try to avoid checks not needed on the platform out of habit.

To make sure I did test this PR on Windows now and the tests doesn't generate any message/warnings/errors so should be perfectly fine as is.

[bagder]

Thanks a lot for the extra set of eyes and comments. I'll merge this set!

@webmaster128: just a request for the future: please check the commit message style we use and try to write the first line in that style. I've edited the commits now to adhere.