Scas 1.0 tracker

[pixelherodev]

• Check hashes against sass-built kernel for each platform

This means using a static KERNEL_VERSION with sass and noting down the hashes of every produced binary. Then, we need to use that same static KERNEL_VERSION with scas and ensure the hashes match. If they don't, looking at the symbol file and debug output should be enough to narrow down the cause.

• Ensure symbol files are accurate

Symbol files will differ from sass', if only because they don't have the 'produced by sass' line. Furthermore, sass produces symbol files with all-lowercase symbol names, while scas preserves the casing.

For 1.1:

• Actually implement .echo
• Fix memory leaks in paths not taken by the kernel
• Generating listing files
• Fuzz?

[pixelherodev]

This should be a week away, tops. It's entirely possible scas is already good enough (hence the release candidate), but it's also possible more work will be needed.

[pixelherodev]

Okay, I've been doing work on libscas, hence the delay. I'm narrowing the scope of this to just have the binary good enough for the kernel for all supported platforms. Anything else is pushed off to 1.1

[pixelherodev]

Update: after testing on the TI84+, I fixed a few more bugs (hence scas is now on 1.0-rc4). Pages 0 and 1 are 100% correct, but there is a minor difference in the output of the floating point math routines in page 2.

[pixelherodev]

The kernel is clearly not as representative as I'd thought; castle is disturbingly broken under 1.0-rc4. A full test of KnightOS/KnightOS is needed before marking this as final, and a solution to catch regressions needs to be in place for 1.1.