DBYTE 13.0.0
13.0 is done. The ZIP is fixed. QEMU passes G1-G5. The BIOSTAR box has not earned a green light yet, and real disks are still hands-off.
NEW TO DBYTE? Start with the first-time guide. Need only the compiler? Use install + verify. Reviewing release truth? Go to release proof.
00 release identity
VERSION 13.0.0 CLASS VIRTUAL_QUALIFIED_SOFTWARE_FINAL TAG v13.0.0 TAG OBJECT 81f36fa28a8caa343145b01632e7563c089a87cc COMMIT 3b68536dc280dade3afde1620eeffbeb0309ec97 PACKAGE dbyte-v13.0.0-windows-x64.zip BYTES 4016724 ENTRIES 536 SHA-256 40468ef823543b23638d633fbfc5e7e82b6da23ce4659238a3c427f2916337d9 RUNTIME 25 entries
These numbers are the release. Change one and you are talking about different bytes. The published ZIP was reproduced twice from the exact tag and both copies matched byte-for-byte.
10 what v13 adds
v13 keeps the language, owned runtime, ABI, flat VM, collections and Native MIR foundations from v12. It adds a BIOS-first freestanding x86_64 machine path: checked Multiboot2 handoff, versioned BootInfo, bounded memory-map normalization, ACPI table validation, APIC/IOAPIC/timer discovery, read-only PCI mechanism #1 enumeration, legacy diagnostics, deterministic ELF/ISO packaging and bounded virtual acceptance.
The important rule is separation. Compiler behavior, QEMU execution, a physical motherboard boot and a physical disk write are four different claims. v13.0 seals the software and virtual-machine claim. It does not smuggle the other three into the word “final.”
20 where to read next
| Question | Owner |
|---|---|
| Install and verify the package | 13.0 install |
| Source to native bytes | compiler path + 12.4 Native MIR |
| Freestanding machine proof | machine foundation |
| Tag/package reproducibility | release proof |
| Current post-release work | 13.1 development snapshot |
30 what is still not done
V13 SOFTWARE STABLE / SEALED G1-G5 VIRTUAL PASS REAL_BIOSTAR PENDING REAL_STORAGE DISARMED FULL_OS PENDING
QEMU remains a useful virtual oracle. It cannot certify the spare physical PC, media durability, storage controller selection, GUI, networking or a future full operating-system claim.
Stable is not a magic spell. It only covers the thing we actually tested.
40 working model for developers
Think of DBYTE as a chain of contracts rather than one compiler executable. Source syntax is parsed, canonical semantic analysis decides meaning, Native MIR makes machine-oriented intent explicit, target legalization applies ABI and hardware constraints, and the writer emits the final object or executable. A failure is easier to isolate when you ask which contract first diverged from the expected state.
For language work, start above Native MIR. For register allocation or instruction selection, stay below semantic analysis. For freestanding failures, separate compiler bytes from firmware topology and physical-device evidence. This separation is deliberate: one subsystem should not be "fixed" merely because another subsystem produced a bad observation.