kit

commit 51d3beb1b6239099fe5d1500fdaecaa2971e76ff
parent b930f6ac494f8af3d271859633d313b6729238b6
Author: Ryan Sepassi <rsepassi@gmail.com>
Date:   Fri, 29 May 2026 14:01:18 -0700

asm: implement .comm/.lcomm and .uleb128/.sleb128 (were silently skipped)

The shared assembler driver lumped .comm/.lcomm/.uleb128/.sleb128 into the
accepted-but-ignored skip-to-EOL bucket. That was silent miscompilation, not a
missing-feature error: .comm produced no symbol and reserved no space (a
program relying on '.comm buf,1024' got a missing/zero-size symbol), and
.uleb128/.sleb128 emitted zero bytes, shifting every following byte in
hand-written DWARF/exception tables.

Implement them:
  - .comm/.lcomm NAME, SIZE[, ALIGN] mint an SK_COMMON symbol (global / local)
    with the given size and alignment; the linker allocates .bss space.
  - .uleb128/.sleb128 VALUE[, ...] emit proper LEB128 byte sequences.

Verified: .comm gbuf -> global COMMON (align 8), .lcomm lbuf -> local COMMON;
LEB128 output matches llvm-mc byte-for-byte (00 7f 80 01 e5 8e 26 7e ff 7e 3f).
Adds an aa64 LEB128 encode corpus case to the default suite.

Diffstat:
M
src/asm/asm.c
 | 
64
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
A
test/asm/encode/aa64_data_leb.expected.hex
 | 
1
+
A
test/asm/encode/aa64_data_leb.s
 | 
9
+++++++++
A
test/asm/encode/aa64_data_leb.targets
 | 
2
++

4 files changed, 73 insertions(+), 3 deletions(-)
diff --git a/src/asm/asm.c b/src/asm/asm.c
@@ -897,15 +897,73 @@ static void do_directive(AsmDriver* d, Sym name) {
     return;
   }
 
+  /* .comm/.lcomm NAME, SIZE[, ALIGN] — declare a common symbol. Previously
+   * skipped, which silently produced no symbol and reserved no space. Model
+   * both as SK_COMMON (the linker allocates .bss space); .comm is global,
+   * .lcomm local. */
+  if (sym_eq(d, name, "comm") || sym_eq(d, name, "lcomm")) {
+    int is_local = sym_eq(d, name, "lcomm");
+    Sym nm = expect_ident(d, ".comm");
+    i64 size = 0, align = 1;
+    if (!asm_driver_eat_comma(d)) d_panicf(d, "asm: .comm: expected ','");
+    size = asm_driver_parse_const(d);
+    if (asm_driver_eat_comma(d)) align = asm_driver_parse_const(d);
+    if (size < 0) size = 0;
+    if (align < 1) align = 1;
+    {
+      ObjSym* s = sym_mut(d, intern_sym(d, nm));
+      s->kind = (u16)SK_COMMON;
+      s->bind = (u16)(is_local ? SB_LOCAL : SB_GLOBAL);
+      s->size = (u64)size;
+      s->common_align = (u64)align;
+    }
+    d_skip_to_eol(d);
+    return;
+  }
+  /* .uleb128/.sleb128 VALUE[, VALUE...] — emit LEB128-encoded bytes.
+   * Previously skipped, which emitted nothing and corrupted any hand-written
+   * DWARF / exception tables that follow. */
+  if (sym_eq(d, name, "uleb128") || sym_eq(d, name, "sleb128")) {
+    int sgn = sym_eq(d, name, "sleb128");
+    (void)asm_driver_cur_section(d);
+    for (;;) {
+      i64 v = asm_driver_parse_const(d);
+      u8 buf[16];
+      u32 n = 0;
+      if (sgn) {
+        int more = 1;
+        while (more) {
+          u8 b = (u8)((u64)v & 0x7fu);
+          v >>= 7; /* arithmetic right shift keeps the sign */
+          if ((v == 0 && !(b & 0x40u)) || (v == -1 && (b & 0x40u)))
+            more = 0;
+          else
+            b |= 0x80u;
+          buf[n++] = b;
+        }
+      } else {
+        u64 uv = (u64)v;
+        do {
+          u8 b = (u8)(uv & 0x7fu);
+          uv >>= 7;
+          if (uv) b |= 0x80u;
+          buf[n++] = b;
+        } while (uv);
+      }
+      d->mc->emit_bytes(d->mc, buf, n);
+      if (!asm_driver_eat_comma(d)) break;
+    }
+    d_skip_to_eol(d);
+    return;
+  }
+
   /* CFI block + accepted-but-ignored directives.  Keep parser
    * forward-progress without aborting the whole TU. */
   if (starts_with(d, name, "cfi_") || sym_eq(d, name, "file") ||
       sym_eq(d, name, "loc") || sym_eq(d, name, "ident") ||
       sym_eq(d, name, "popsection") || sym_eq(d, name, "pushsection") ||
       sym_eq(d, name, "previous") ||
-      sym_eq(d, name, "subsections_via_symbols") || sym_eq(d, name, "comm") ||
-      sym_eq(d, name, "lcomm") || sym_eq(d, name, "uleb128") ||
-      sym_eq(d, name, "sleb128") || sym_eq(d, name, "macro") ||
+      sym_eq(d, name, "subsections_via_symbols") || sym_eq(d, name, "macro") ||
       sym_eq(d, name, "endm") || sym_eq(d, name, "if") ||
       sym_eq(d, name, "endif") || sym_eq(d, name, "else") ||
       sym_eq(d, name, "include")) {
diff --git a/test/asm/encode/aa64_data_leb.expected.hex b/test/asm/encode/aa64_data_leb.expected.hex
@@ -0,0 +1 @@
+007f8001e58e267eff7e3faa
diff --git a/test/asm/encode/aa64_data_leb.s b/test/asm/encode/aa64_data_leb.s
@@ -0,0 +1,9 @@
+.text
+    .uleb128 0
+    .uleb128 127
+    .uleb128 128
+    .uleb128 624485
+    .sleb128 -2
+    .sleb128 -129
+    .sleb128 63
+    .byte 0xaa
diff --git a/test/asm/encode/aa64_data_leb.targets b/test/asm/encode/aa64_data_leb.targets
@@ -0,0 +1 @@
+aa64
+\ No newline at end of file