find_replace: cover o3 function prologues on every backend

arm64, llvm and the region-interpreter rdrop path now watch the two
prologue words the way the x86-64 jit already did, so icon.replace_function's
lit+jmp patch retires a compiled callee (and clears fn_nat) instead of leaving
a stale stub. three new icon regressions: repeated o3 replace, failed
publication keeps the old entry, and a direct call after warming the region.

1 files changed, 44 insertions, 7 deletions

diff --git a/src/core/core.c b/src/core/core.c
index 748c508..3133c55 100644
--- a/src/core/core.c
+++ b/src/core/core.c
@@ -4000,8 +4000,21 @@ static void rforget(Rg *R)
 	if ((unsigned)R->fn < (unsigned)code_words) fn_nat[R->fn].stub = 0;
 	R->nat = 0;
 }
-static void rmark(Rg *R) { R->marked = 1; cinc(cov, R->lo, R->hi); }
-static void runmark(Rg *R) { cdec(cov, R->lo, R->hi); }
+static void rmark(Rg *R)
+{
+	R->marked = 1; cinc(cov, R->lo, R->hi);
+	/* cover the function prologue: icon.replace_function patches it with lit+jmp */
+	if ((unsigned)R->fn < (unsigned)code_words) {
+		cinc(cov, R->fn, R->fn + 2);
+		tab[R->fn] = 0;
+		if ((unsigned)(R->fn + 1) < (unsigned)code_words) tab[R->fn + 1] = 0;
+	}
+}
+static void runmark(Rg *R)
+{
+	cdec(cov, R->lo, R->hi);
+	if ((unsigned)R->fn < (unsigned)code_words) cdec(cov, R->fn, R->fn + 2);
+}
 static int ralloc(Rg *R, RB *B)
 {
 	int i, j, k, n = B->n;
@@ -5232,7 +5245,11 @@ static int rdrop(int a)		/* a store landed on region code: drop those regions */
 {
 	int j, hit = 0;
 	for (j = 0; j < nrg; j++)
-		if (rgs[j]->lifted && rgs[j]->lo <= a && a < rgs[j]->hi) {
+		if (rgs[j]->lifted && ((rgs[j]->lo <= a && a < rgs[j]->hi) ||
+		    ((unsigned)rgs[j]->fn < (unsigned)code_words && a >= rgs[j]->fn && a < rgs[j]->fn + 2))) {
+			/* prologue patch (icon.replace_function) retires the region; a body store only clears it */
+			if ((unsigned)rgs[j]->fn < (unsigned)code_words && a >= rgs[j]->fn && a < rgs[j]->fn + 2)
+				rgs[j]->dead = 1;
 			if (trace) fprintf(stderr, "o3: dropped %d..%d (a store to %d: %d)\n", rgs[j]->lo, rgs[j]->hi - 1, a, m[a]);
 			rclear(rgs[j]); hit = 1;
 		}
@@ -5346,8 +5363,16 @@ static void runmark(Rg *R)
 }
 #else
 static void rforget(Rg *R) { (void)R; }
-static void rmark(Rg *R) { R->marked = 1; cinc(rcov, R->lo, R->hi); }
-static void runmark(Rg *R) { cdec(rcov, R->lo, R->hi); }
+static void rmark(Rg *R)
+{
+	R->marked = 1; cinc(rcov, R->lo, R->hi);
+	if ((unsigned)R->fn < (unsigned)code_words) cinc(rcov, R->fn, R->fn + 2);
+}
+static void runmark(Rg *R)
+{
+	cdec(rcov, R->lo, R->hi);
+	if ((unsigned)R->fn < (unsigned)code_words) cdec(rcov, R->fn, R->fn + 2);
+}
 static void rcompile(Rg *R, int eager) { (void)R; (void)eager; }
 #endif
 #endif
@@ -5359,8 +5384,20 @@ static void rcompile(Rg *R, int eager) { (void)R; (void)eager; }
 typedef Cell (*RFn)(Cell *mem, Cell *regs, int ent);
 static VM_TLS int nregions;
 static void rforget(Rg *R) { rt_queue(R->rt); R->rt = 0; R->nat = 0; R->dnat = 0; }
-static void rmark(Rg *R) { R->marked = 1; cinc(rcov, R->lo, R->hi); cinc(lcov, R->lo, R->hi); }
-static void runmark(Rg *R) { cdec(rcov, R->lo, R->hi); cdec(lcov, R->lo, R->hi); }
+static void rmark(Rg *R)
+{
+	R->marked = 1; cinc(rcov, R->lo, R->hi); cinc(lcov, R->lo, R->hi);
+	if ((unsigned)R->fn < (unsigned)code_words) {
+		cinc(rcov, R->fn, R->fn + 2); cinc(lcov, R->fn, R->fn + 2);
+	}
+}
+static void runmark(Rg *R)
+{
+	cdec(rcov, R->lo, R->hi); cdec(lcov, R->lo, R->hi);
+	if ((unsigned)R->fn < (unsigned)code_words) {
+		cdec(rcov, R->fn, R->fn + 2); cdec(lcov, R->fn, R->fn + 2);
+	}
+}
 static LLVMValueRef as_f64(LLVMValueRef v, int fp) { return fp ? v : LLVMBuildBitCast(bd, v, f64t, ""); }
 static LLVMValueRef as_i64(LLVMValueRef v, int fp) { return fp ? LLVMBuildBitCast(bd, v, i64t, "") : v; }
 /* the variables go back to their frame slots. before a call from inside the