From a20a82c41732ecbdd9fe9f4359844bfea4698bf6 Mon Sep 17 00:00:00 2001
From: Graeme Geldenhuys <graemeg@gmail.com>
Date: Fri, 26 Jun 2026 18:30:53 +0100
Subject: [PATCH] fix(asm): brace-comment strip must skip quoted strings

The asm brace-comment strip (added with the threads leaf to remove {...}
comments the lexer leaves inside asm blocks) ran over the WHOLE line, including
inside .ascii string literals.  The native backend emits string-const data as
.ascii directives, and the compiler's own codegen contains template strings with
literal braces (e.g. codegen.qbe.pas's 'data $__s%d = { w -1, w %d, w %d, b
"%s", b 0 }').  When the compiler was assembled by the internal assembler, the
strip ate the '{ ... }' span out of those .ascii literals, corrupting the
template strings.

The damage only showed when the compiler ran --emit-ir / GenerateIR: the
corrupted template produced garbage data lines (data $__s0 = <garbage>), so the
~1240 IR-assertion tests failed (TestHelloWorld_HasStrLitData etc.).  Compiled
user programs were unaffected (their own Format/strings come from freshly built
RTL), and the QBE-built compiler was unaffected (gcc assembles its .ascii), which
is why local QBE fixpoints stayed green while the CI native-internal TestRunner
went red.

Fix: skip over quoted strings in the brace strip, exactly as the adjacent
'#'-comment strip already does, so a brace inside a .ascii literal is left
intact.

Root-caused from the CI failure on ci/linux-direct-syscalls; the corruption is
visible as -clean but -garbled string data from a
native-internal-assembled compiler.
---
 .../main/pascal/blaise.assembler.x86_64.pas   | 23 +++++++++++++++++--
 1 file changed, 21 insertions(+), 2 deletions(-)

diff --git a/compiler/src/main/pascal/blaise.assembler.x86_64.pas b/compiler/src/main/pascal/blaise.assembler.x86_64.pas
index 93edb81..35f5872 100644
--- a/compiler/src/main/pascal/blaise.assembler.x86_64.pas
+++ b/compiler/src/main/pascal/blaise.assembler.x86_64.pas
@@ -670,11 +670,30 @@ begin
   { Strip a brace comment (open/close = ASCII 123/125) that survived the lexer:
     a whole-line or trailing comment inside an asm block.  Single line only;
     remove the span from the open brace to the matching close brace, or to end
-    of line when unterminated. }
+    of line when unterminated.  Skip over quoted strings so that a brace INSIDE
+    a .ascii string literal (the string-const data the native backend emits) is
+    left intact rather than mistaken for a comment. }
   P := 0;
   while P < Length(S) do
   begin
-    if S[P] = 123 then
+    if S[P] = Ord('"') then
+    begin
+      { Skip the whole quoted string, honouring \" escapes. }
+      P := P + 1;
+      while P < Length(S) do
+      begin
+        if S[P] = Ord('\') then
+          P := P + 2
+        else if S[P] = Ord('"') then
+        begin
+          P := P + 1;
+          Break;
+        end
+        else
+          P := P + 1;
+      end;
+    end
+    else if S[P] = 123 then
     begin
       Q := P + 1;
       while (Q < Length(S)) and (S[Q] <> 125) do