Add naettGetTotalBytesRead

[hrydgard]

This lets you implement a progress bar for downloads safely and efficiently by calling naettGetTotalBytesRead() after checking naettComplete() and finding out it's not done yet.

The first commit makes sure that res->headers only gets written to once by the various backends. This is not actually needed for the two later commits, and still doesn't technically make it safe to read headers before the full request is complete (although in practice it'll be), but it's still good practice and probably a slight performance improvement. Not implemented in linux.c though because it's less convenient to do there.

Fixes #16

[erkkah]

Thanks. The only issue with this is that it extends the non-thread-safe handling of the completed field with the progress field. I'll add an issue about that.

[erkkah]

Of course some conflicts. If you rebase on main, I'll merge!

[hrydgard]

will do!

also, accesses to 32-bit fields are effectively atomic - although unlike actual atomics not with respect to surrounding operations since the compiler can move stuff around, but progress only needs to be approximate anyway, so...

[hrydgard]

Rebased on main.

[hrydgard]

Hm, looks like a linking error with the "min" function. I don't get that error myself, but I'll try to do something.

[erkkah]

Yep, it's just the reordering that can be an issue.

Yes, the definition of min is different on the mingw pipeline used in tests. That's why I had the define earlier, but I was a bit too eager.

[erkkah]

BTW, I've fixed the warnings in the test rig locally.

[hrydgard]

Took the easy way out.

By the way, thanks for this convenient little library. Exactly what I needed, and I'll probably end up using it in more projects in the future.

[erkkah]

Still fails on the old uses of min. I don't have an MSVC setup available.
Wouldn't it work to keep the old #define min(a,b) (((a)<(b))?(a):(b)) define, and the NOMINMAX define?

You're welcome. This little library was just what I needed as well :)

[hrydgard]

Alright, I'll move back to that. Although, I don't see where those old uses are - I don't see any in the code.

[hrydgard]

OK the windows test runner is clearly not even using my latest changes, it seems to be running an older build. Not sure what's going on.

[erkkah]

Weird. I'll fix!

[erkkah]

Moved this to a new PR #23 For branching / workflow reasons.