diff --git a/src/hb-ot-cff-common.hh b/src/hb-ot-cff-common.hh
index 2c4b1c6c0..a544fea1b 100644
--- a/src/hb-ot-cff-common.hh
+++ b/src/hb-ot-cff-common.hh
@@ -68,11 +68,15 @@ struct CFFIndex
 	    hb_requires (hb_is_iterable (Iterable))>
   bool serialize (hb_serialize_context_t *c,
 		  const Iterable &iterable,
-		  unsigned data_size = (unsigned) -1)
+		  const unsigned *p_data_size = nullptr)
   {
     TRACE_SERIALIZE (this);
-    if (data_size == (unsigned) -1)
+    unsigned data_size;
+    if (p_data_size)
+      data_size = *p_data_size;
+    else
       total_size (iterable, &data_size);
+
     auto it = hb_iter (iterable);
     if (unlikely (!serialize_header (c, +it, data_size))) return_trace (false);
     unsigned char *ret = c->allocate_size<unsigned char> (data_size, false);
diff --git a/src/hb-subset-cff1.cc b/src/hb-subset-cff1.cc
index 1541b542f..fb93448dc 100644
--- a/src/hb-subset-cff1.cc
+++ b/src/hb-subset-cff1.cc
@@ -831,7 +831,7 @@ OT::cff1::accelerator_subset_t::serialize (hb_serialize_context_t *c,
        return false;
 
     auto *cs = c->start_embed<CFF1CharStrings> ();
-    if (likely (cs->serialize (c, plan.subset_charstrings, data_size)))
+    if (likely (cs->serialize (c, plan.subset_charstrings, &data_size)))
       plan.info.char_strings_link = c->pop_pack (false);
     else
     {
diff --git a/src/hb-subset-cff2.cc b/src/hb-subset-cff2.cc
index bdae27c76..3c52fb9c2 100644
--- a/src/hb-subset-cff2.cc
+++ b/src/hb-subset-cff2.cc
@@ -566,7 +566,7 @@ OT::cff2::accelerator_subset_t::serialize (hb_serialize_context_t *c,
        return false;
 
     auto *cs = c->start_embed<CFF2CharStrings> ();
-    if (likely (cs->serialize (c, plan.subset_charstrings, data_size)))
+    if (likely (cs->serialize (c, plan.subset_charstrings, &data_size)))
       plan.info.char_strings_link = c->pop_pack (false);
     else
     {