Flint v1.7.1

[lucaas-d3v]

Release v1.7.1: POSIX I/O Engine, Zero-Copy, and Memory Shielding
This release consolidates Flint as a systems automation language focused on raw performance. Inefficient User Space abstractions have been replaced by direct Linux Kernel calls, ensuring scalability for servers under high load.
Performance and Disk I/O (Kernel Space):

• perf(fs): Implemented Zero-Copy in copy() using sendfile and posix_fadvise. File copying now occurs entirely within Kernel Ring 0, achieving 5x lower CPU consumption (User Time) than GNU cp.
• perf(fs): Introduced the FLINT_C_PATH macro to force path allocation on the Stack ($O(1)$). Metadata operations (is_dir, file_size, rm, mkdir) now execute with zero impact and zero fragmentation in the Virtual Arena.

Process and Memory Management:

• refactor(spawn): Eradicated the fork() syscall. Adopted posix_spawnp for subprocess creation, eliminating Page Table cloning overhead on servers with high RAM usage.
• fix(mem): Resolved memory leaks in ls() and spawn(). Implemented a "Hybrid Allocation" pattern: isolated dynamic malloc/realloc usage during unpredictable data collection, followed by an exact commit to the Virtual Arena (flint_alloc_raw) and a free() of the scratchpad memory.

Syntax and Error Handling:

• fix(parser): Structural stabilization of the AST for nested if/else blocks.
• feat(runtime): Integrated the catch operator with C POSIX errno, allowing for the safe capture of hardware, permission, and network failures.

Validation Metrics (vs. Python 3.12 and Bash):

• Raw I/O (5GB): Flint is 1.2x faster than Bash (cp) and 1.08x faster than Python, with a negligible CPU fraction.
• Mass Stat (10,000 files): Flint executed in a constant 7.1ms (zero temporal degradation), making it 2.03x faster than Python and 25% faster than pure Bash loops.