zstd system library suppport

[etbe]

Please make the CMakeLists.txt file check for /usr/include/zstd.h and /usr/lib/x86_64-linux-gnu/libzstd.a and use them instead of building an inline copy of zstd if available.

[richgel999]

If we do this we're now introducing a dependency, making the library more brittle.

[etbe]

I think that having multiple copies of zstd being used on the same system makes it more brittle. Then if there's an issue with zstd it has to be solved in multiple places. If there's an issue that's not obvious (EG occasional data corruption) then things will fail in unexpected ways.

Also I'm not suggesting a hard dependency, just a cmake check for it and use the inline code if it's not installed.

[richgel999]

We're going to be working on our cmake situation more and can add this as an option.

[jamesdolan]

Hi Rich, any word on this? 2.0 regresses this even further. Previously we could just set BASISD_SUPPORT_KTX2_ZSTD=0, but now basisu_astc_ldr_encode.cpp unconditionally includes a custom zstd include via a relative path. The ../zstd/zstd.h hard coded path really forces us to patch/fork to fix, and seems like Basisu could just give each external dependency its own search path that users could override co cleanly avoid this.
Looks like a few PRs have been submitted in the past for this (e.g. #344) but not merged.

Also, for use cases where assets already have a lossless compression applied (or in the case of web deployment where gzip is typically applied by server already), its also nice to be able to turn this off internally to avoid that redundancy.

[richgel999]

I hear you, and it is a problem we're going to fix. From our perspective, it's not a regression, it's our previous/current default, which is no dependencies on external libraries to simplify building. Our default will never require system Zstd or any other 3rd party system libraries. (I've wasted probably weeks of my time trying to get projects to build over the years.)
We're going to address this using a CMake option. I'll look at the PR and make sure it doesn't break anything and decide where to go from here. It was higher priority to ship v2.0.