Fall back to fcntl if ioctl(FIOCLEX, ..) is unsupported

[tbu-]

Fixes #41347.

[rust-highfive]

r? @BurntSushi

(rust_highfive has picked a reviewer for you, use r? to override)

[eddyb]

cc @rust-lang/libs

[Mark-Simulacrum]

Thanks for the PR! We’ll periodically check in on it to make sure that @BurntSushi or someone else from the team reviews it soon.

[BurntSushi]

@tbu- Sorry, I'm missing a lot of context here. I don't have any idea what this change actually does or what problem it solves. Could you perhaps say more about what's happening here?

[eddyb]

@BurntSushi If you see the linked issue, this change is trying to provide support for faux-linux systems that don't support the ioctl (I think it's similar to Linux 2.4 but I'm not sure).

[alexcrichton]

Could you expand the description in the PR to describe the problem? And can you also leave a comment as to why the code is structured the way it is?

I'd personally prefer to not go too far down the road of supporting pre-2.6.18 Linux kernels, so is this literally intended to support Linux 2.4 or is it portability to Esxi?

[arielb1]

@tbu-

Could you describe which kernels are you trying to support with this PR?

[tbu-]

This was just trying to fix the issue that was encountered in #41437. I don't know if it needs other stuff as well.

I'm not affected by this, so I'm not too attached to this patch.

[alexcrichton]

Ok, in that case I'm personally inclined to hold off for now. We don't know if the target in #41347 actually warrants an entirely new compiler target or if this change is enough to get the program working. In light of that it seems like it may be premature to do this?

[tbu-]

OK.