chore: prepare release v0.33.0
Release / build (amd64, linux) (push) Successful in 49s
Release / build (arm64, linux) (push) Successful in 43s
Release / build (loong64, linux) (push) Successful in 46s
Release / build (riscv64, linux) (push) Successful in 45s
Test / vet (push) Successful in 47s
Release / release (push) Successful in 18s
Test / test (push) Successful in 2m39s
Test / build (push) Successful in 43s

This commit is contained in:
2026-09-14 23:36:19 +02:00
parent 40476546df
commit fff9f75595
6 changed files with 101 additions and 6 deletions
+12
View File
@@ -237,6 +237,18 @@ STP+SUB for large frames) and SB/global symbol references (ADRP+ADD pairs with
R_ADDRARM64 relocations). Like the other encoders it is validated
byte-for-byte against `GOARCH=arm64 go tool asm`.
On top of the per-architecture encoders, every framed function carries the
**stack-split guard**: the prologue check against `g.stackguard0` (small,
medium and large frame classes, the medium and large classes materialising
their offset through the architecture's temporary register and the large
class adding the SP-underflow branch) and the trailing morestack block
(save the link register, `CALL runtime.morestack_noctxt`, jump back to the
function entry). The auto-NOSPLIT rule, the frame classes, the large-frame
prologue and epilogue forms and the tail calls match the toolchain's
`stacksplit` and `preprocess` output byte for byte; a parity suite
assembles kernel files with gasm and the installed `go tool asm` and diffs
the bytes on all four architectures.
On top of the encoder, `Assemble` walks a parsed `TEXT` body, converts each
operand to an encoder operand, and lays the instructions out so local labels
resolve to relative jump offsets: jumps start in the short (rel8) form and