diff --git a/include/jemalloc/internal/activity_callback.h b/include/jemalloc/internal/activity_callback.h
index 6c2e84e3..0f4f3962 100644
--- a/include/jemalloc/internal/activity_callback.h
+++ b/include/jemalloc/internal/activity_callback.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_ACTIVITY_CALLBACK_H
 #define JEMALLOC_INTERNAL_ACTIVITY_CALLBACK_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 /*
  * The callback to be executed "periodically", in response to some amount of
  * allocator activity.
diff --git a/include/jemalloc/internal/arena_externs.h b/include/jemalloc/internal/arena_externs.h
index 22d7fff7..d79b607a 100644
--- a/include/jemalloc/internal/arena_externs.h
+++ b/include/jemalloc/internal/arena_externs.h
@@ -1,8 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_ARENA_EXTERNS_H
 #define JEMALLOC_INTERNAL_ARENA_EXTERNS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_stats.h"
 #include "jemalloc/internal/bin.h"
 #include "jemalloc/internal/div.h"
+#include "jemalloc/internal/emap.h"
 #include "jemalloc/internal/extent_dss.h"
 #include "jemalloc/internal/hook.h"
 #include "jemalloc/internal/pages.h"
diff --git a/include/jemalloc/internal/arena_inlines_a.h b/include/jemalloc/internal/arena_inlines_a.h
index 8568358c..214ce80b 100644
--- a/include/jemalloc/internal/arena_inlines_a.h
+++ b/include/jemalloc/internal/arena_inlines_a.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_ARENA_INLINES_A_H
 #define JEMALLOC_INTERNAL_ARENA_INLINES_A_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_structs.h"
+
 static inline unsigned
 arena_ind_get(const arena_t *arena) {
 	return arena->ind;
diff --git a/include/jemalloc/internal/arena_inlines_b.h b/include/jemalloc/internal/arena_inlines_b.h
index 420a62b2..c7d08227 100644
--- a/include/jemalloc/internal/arena_inlines_b.h
+++ b/include/jemalloc/internal/arena_inlines_b.h
@@ -1,14 +1,22 @@
 #ifndef JEMALLOC_INTERNAL_ARENA_INLINES_B_H
 #define JEMALLOC_INTERNAL_ARENA_INLINES_B_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_externs.h"
+#include "jemalloc/internal/arena_structs.h"
 #include "jemalloc/internal/div.h"
 #include "jemalloc/internal/emap.h"
+#include "jemalloc/internal/jemalloc_internal_inlines_b.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/large_externs.h"
 #include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/prof_externs.h"
+#include "jemalloc/internal/prof_structs.h"
 #include "jemalloc/internal/rtree.h"
 #include "jemalloc/internal/safety_check.h"
 #include "jemalloc/internal/sc.h"
 #include "jemalloc/internal/sz.h"
+#include "jemalloc/internal/tcache_inlines.h"
 #include "jemalloc/internal/ticker.h"
 
 static inline arena_t *
diff --git a/include/jemalloc/internal/arena_stats.h b/include/jemalloc/internal/arena_stats.h
index 15f1d345..3407b023 100644
--- a/include/jemalloc/internal/arena_stats.h
+++ b/include/jemalloc/internal/arena_stats.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_ARENA_STATS_H
 #define JEMALLOC_INTERNAL_ARENA_STATS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/lockedint.h"
 #include "jemalloc/internal/mutex.h"
diff --git a/include/jemalloc/internal/arena_structs.h b/include/jemalloc/internal/arena_structs.h
index e6868fce..0fffa7eb 100644
--- a/include/jemalloc/internal/arena_structs.h
+++ b/include/jemalloc/internal/arena_structs.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_ARENA_STRUCTS_H
 #define JEMALLOC_INTERNAL_ARENA_STRUCTS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/arena_stats.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/bin.h"
diff --git a/include/jemalloc/internal/arena_types.h b/include/jemalloc/internal/arena_types.h
index 45eec69f..a1fc8926 100644
--- a/include/jemalloc/internal/arena_types.h
+++ b/include/jemalloc/internal/arena_types.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_ARENA_TYPES_H
 #define JEMALLOC_INTERNAL_ARENA_TYPES_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/sc.h"
 
 /* Default decay times in milliseconds. */
diff --git a/include/jemalloc/internal/assert.h b/include/jemalloc/internal/assert.h
index be4d45b3..38eb2a2c 100644
--- a/include/jemalloc/internal/assert.h
+++ b/include/jemalloc/internal/assert.h
@@ -1,3 +1,4 @@
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/malloc_io.h"
 #include "jemalloc/internal/util.h"
 
diff --git a/include/jemalloc/internal/atomic.h b/include/jemalloc/internal/atomic.h
index c0f73122..6dd2a7c6 100644
--- a/include/jemalloc/internal/atomic.h
+++ b/include/jemalloc/internal/atomic.h
@@ -1,7 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_ATOMIC_H
 #define JEMALLOC_INTERNAL_ATOMIC_H
 
-#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
+#include "jemalloc/internal/jemalloc_preamble.h"
 
 #define JEMALLOC_U8_ATOMICS
 #if defined(JEMALLOC_GCC_ATOMIC_ATOMICS)
@@ -22,6 +22,8 @@
 #  error "Don't have atomics implemented on this platform."
 #endif
 
+#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
+
 /*
  * This header gives more or less a backport of C11 atomics. The user can write
  * JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_sizeof_type); to generate
diff --git a/include/jemalloc/internal/atomic_c11.h b/include/jemalloc/internal/atomic_c11.h
index a5f9313a..74173b03 100644
--- a/include/jemalloc/internal/atomic_c11.h
+++ b/include/jemalloc/internal/atomic_c11.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_ATOMIC_C11_H
 #define JEMALLOC_INTERNAL_ATOMIC_C11_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include <stdatomic.h>
 
 #define ATOMIC_INIT(...) ATOMIC_VAR_INIT(__VA_ARGS__)
diff --git a/include/jemalloc/internal/atomic_gcc_atomic.h b/include/jemalloc/internal/atomic_gcc_atomic.h
index 471515e8..0819fde1 100644
--- a/include/jemalloc/internal/atomic_gcc_atomic.h
+++ b/include/jemalloc/internal/atomic_gcc_atomic.h
@@ -1,8 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H
 #define JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/assert.h"
 
+#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
+
 #define ATOMIC_INIT(...) {__VA_ARGS__}
 
 typedef enum {
@@ -126,4 +129,6 @@ atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, type val,	\
 	    atomic_enum_to_builtin(mo));				\
 }
 
+#undef ATOMIC_INLINE
+
 #endif /* JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H */
diff --git a/include/jemalloc/internal/atomic_gcc_sync.h b/include/jemalloc/internal/atomic_gcc_sync.h
index e02b7cbe..21136bd0 100644
--- a/include/jemalloc/internal/atomic_gcc_sync.h
+++ b/include/jemalloc/internal/atomic_gcc_sync.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H
 #define JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
+#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
+
 #define ATOMIC_INIT(...) {__VA_ARGS__}
 
 typedef enum {
@@ -192,4 +196,6 @@ atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, type val,	\
 	return __sync_fetch_and_xor(&a->repr, val);			\
 }
 
+#undef ATOMIC_INLINE
+
 #endif /* JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H */
diff --git a/include/jemalloc/internal/atomic_msvc.h b/include/jemalloc/internal/atomic_msvc.h
index 67057ce5..a429f1ab 100644
--- a/include/jemalloc/internal/atomic_msvc.h
+++ b/include/jemalloc/internal/atomic_msvc.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_ATOMIC_MSVC_H
 #define JEMALLOC_INTERNAL_ATOMIC_MSVC_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
+#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
+
 #define ATOMIC_INIT(...) {__VA_ARGS__}
 
 typedef enum {
@@ -155,4 +159,6 @@ atomic_fetch_xor_##short_type(atomic_##short_type##_t *a,		\
 	    &a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val);		\
 }
 
+#undef ATOMIC_INLINE
+
 #endif /* JEMALLOC_INTERNAL_ATOMIC_MSVC_H */
diff --git a/include/jemalloc/internal/background_thread_externs.h b/include/jemalloc/internal/background_thread_externs.h
index 6ae3c8d8..0d34ee55 100644
--- a/include/jemalloc/internal/background_thread_externs.h
+++ b/include/jemalloc/internal/background_thread_externs.h
@@ -1,6 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
 #define JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/background_thread_structs.h"
+#include "jemalloc/internal/base.h"
+#include "jemalloc/internal/mutex.h"
+
 extern bool opt_background_thread;
 extern size_t opt_max_background_threads;
 extern malloc_mutex_t background_thread_lock;
diff --git a/include/jemalloc/internal/background_thread_inlines.h b/include/jemalloc/internal/background_thread_inlines.h
index 92c5febe..4ed05d1b 100644
--- a/include/jemalloc/internal/background_thread_inlines.h
+++ b/include/jemalloc/internal/background_thread_inlines.h
@@ -1,6 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_INLINES_H
 #define JEMALLOC_INTERNAL_BACKGROUND_THREAD_INLINES_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_inlines_a.h"
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/background_thread_externs.h"
+
 JEMALLOC_ALWAYS_INLINE bool
 background_thread_enabled(void) {
 	return atomic_load_b(&background_thread_enabled_state, ATOMIC_RELAXED);
diff --git a/include/jemalloc/internal/background_thread_structs.h b/include/jemalloc/internal/background_thread_structs.h
index 83a91984..67b68797 100644
--- a/include/jemalloc/internal/background_thread_structs.h
+++ b/include/jemalloc/internal/background_thread_structs.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
 #define JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/mutex.h"
+
 /* This file really combines "structs" and "types", but only transitionally. */
 
 #if defined(JEMALLOC_BACKGROUND_THREAD) || defined(JEMALLOC_LAZY_LOCK)
diff --git a/include/jemalloc/internal/base.h b/include/jemalloc/internal/base.h
index 23207563..6b41aa6f 100644
--- a/include/jemalloc/internal/base.h
+++ b/include/jemalloc/internal/base.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_BASE_H
 #define JEMALLOC_INTERNAL_BASE_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/edata.h"
 #include "jemalloc/internal/ehooks.h"
 #include "jemalloc/internal/mutex.h"
diff --git a/include/jemalloc/internal/bin.h b/include/jemalloc/internal/bin.h
index 027af088..ed27c18f 100644
--- a/include/jemalloc/internal/bin.h
+++ b/include/jemalloc/internal/bin.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_BIN_H
 #define JEMALLOC_INTERNAL_BIN_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/bin_stats.h"
 #include "jemalloc/internal/bin_types.h"
 #include "jemalloc/internal/edata.h"
diff --git a/include/jemalloc/internal/bin_info.h b/include/jemalloc/internal/bin_info.h
index 7fe65c86..b6175550 100644
--- a/include/jemalloc/internal/bin_info.h
+++ b/include/jemalloc/internal/bin_info.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_BIN_INFO_H
 #define JEMALLOC_INTERNAL_BIN_INFO_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/bitmap.h"
 
 /*
diff --git a/include/jemalloc/internal/bin_stats.h b/include/jemalloc/internal/bin_stats.h
index 0b99297c..f95b9e9c 100644
--- a/include/jemalloc/internal/bin_stats.h
+++ b/include/jemalloc/internal/bin_stats.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_BIN_STATS_H
 #define JEMALLOC_INTERNAL_BIN_STATS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/mutex_prof.h"
 
 typedef struct bin_stats_s bin_stats_t;
diff --git a/include/jemalloc/internal/bin_types.h b/include/jemalloc/internal/bin_types.h
index 945e8326..5ec22dfd 100644
--- a/include/jemalloc/internal/bin_types.h
+++ b/include/jemalloc/internal/bin_types.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_BIN_TYPES_H
 #define JEMALLOC_INTERNAL_BIN_TYPES_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/sc.h"
 
 #define BIN_SHARDS_MAX (1 << EDATA_BITS_BINSHARD_WIDTH)
diff --git a/include/jemalloc/internal/bit_util.h b/include/jemalloc/internal/bit_util.h
index 70fa4bc9..c413a75d 100644
--- a/include/jemalloc/internal/bit_util.h
+++ b/include/jemalloc/internal/bit_util.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_BIT_UTIL_H
 #define JEMALLOC_INTERNAL_BIT_UTIL_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/assert.h"
 
 /* Sanity check. */
diff --git a/include/jemalloc/internal/bitmap.h b/include/jemalloc/internal/bitmap.h
index dc19454d..e501da47 100644
--- a/include/jemalloc/internal/bitmap.h
+++ b/include/jemalloc/internal/bitmap.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_BITMAP_H
 #define JEMALLOC_INTERNAL_BITMAP_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/bit_util.h"
 #include "jemalloc/internal/sc.h"
 
diff --git a/include/jemalloc/internal/buf_writer.h b/include/jemalloc/internal/buf_writer.h
index 37aa6de5..fa0ac99c 100644
--- a/include/jemalloc/internal/buf_writer.h
+++ b/include/jemalloc/internal/buf_writer.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_BUF_WRITER_H
 #define JEMALLOC_INTERNAL_BUF_WRITER_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/tsd_types.h"
+
 /*
  * Note: when using the buffered writer, cbopaque is passed to write_cb only
  * when the buffer is flushed.  It would make a difference if cbopaque points
diff --git a/include/jemalloc/internal/cache_bin.h b/include/jemalloc/internal/cache_bin.h
index c9c8f865..218e368e 100644
--- a/include/jemalloc/internal/cache_bin.h
+++ b/include/jemalloc/internal/cache_bin.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_CACHE_BIN_H
 #define JEMALLOC_INTERNAL_CACHE_BIN_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/jemalloc_internal_externs.h"
 #include "jemalloc/internal/ql.h"
 #include "jemalloc/internal/safety_check.h"
 #include "jemalloc/internal/sz.h"
diff --git a/include/jemalloc/internal/ckh.h b/include/jemalloc/internal/ckh.h
index 7b3850bc..8e9d7fed 100644
--- a/include/jemalloc/internal/ckh.h
+++ b/include/jemalloc/internal/ckh.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_CKH_H
 #define JEMALLOC_INTERNAL_CKH_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/tsd.h"
 
 /* Cuckoo hashing implementation.  Skip to the end for the interface. */
diff --git a/include/jemalloc/internal/counter.h b/include/jemalloc/internal/counter.h
index 79abf064..74e30701 100644
--- a/include/jemalloc/internal/counter.h
+++ b/include/jemalloc/internal/counter.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_COUNTER_H
 #define JEMALLOC_INTERNAL_COUNTER_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/lockedint.h"
 #include "jemalloc/internal/mutex.h"
 
 typedef struct counter_accum_s {
diff --git a/include/jemalloc/internal/ctl.h b/include/jemalloc/internal/ctl.h
index 63d27f8a..1d3e6140 100644
--- a/include/jemalloc/internal/ctl.h
+++ b/include/jemalloc/internal/ctl.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_CTL_H
 #define JEMALLOC_INTERNAL_CTL_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_stats.h"
+#include "jemalloc/internal/background_thread_structs.h"
+#include "jemalloc/internal/bin_stats.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
 #include "jemalloc/internal/malloc_io.h"
 #include "jemalloc/internal/mutex_prof.h"
diff --git a/include/jemalloc/internal/decay.h b/include/jemalloc/internal/decay.h
index cf6a9d22..74be55da 100644
--- a/include/jemalloc/internal/decay.h
+++ b/include/jemalloc/internal/decay.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_DECAY_H
 #define JEMALLOC_INTERNAL_DECAY_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/smoothstep.h"
 
 #define DECAY_UNBOUNDED_TIME_TO_PURGE ((uint64_t)-1)
diff --git a/include/jemalloc/internal/div.h b/include/jemalloc/internal/div.h
index aebae939..56d5f463 100644
--- a/include/jemalloc/internal/div.h
+++ b/include/jemalloc/internal/div.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_DIV_H
 #define JEMALLOC_INTERNAL_DIV_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/assert.h"
 
 /*
diff --git a/include/jemalloc/internal/ecache.h b/include/jemalloc/internal/ecache.h
index 71cae3e3..2bd74fde 100644
--- a/include/jemalloc/internal/ecache.h
+++ b/include/jemalloc/internal/ecache.h
@@ -1,9 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_ECACHE_H
 #define JEMALLOC_INTERNAL_ECACHE_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/eset.h"
-#include "jemalloc/internal/san.h"
 #include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/san.h"
 
 typedef struct ecache_s ecache_t;
 struct ecache_s {
diff --git a/include/jemalloc/internal/edata.h b/include/jemalloc/internal/edata.h
index d2d16c46..5fe4e14d 100644
--- a/include/jemalloc/internal/edata.h
+++ b/include/jemalloc/internal/edata.h
@@ -1,12 +1,14 @@
 #ifndef JEMALLOC_INTERNAL_EDATA_H
 #define JEMALLOC_INTERNAL_EDATA_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/bin_info.h"
 #include "jemalloc/internal/bit_util.h"
 #include "jemalloc/internal/hpdata.h"
 #include "jemalloc/internal/nstime.h"
 #include "jemalloc/internal/ph.h"
+#include "jemalloc/internal/prof_types.h"
 #include "jemalloc/internal/ql.h"
 #include "jemalloc/internal/sc.h"
 #include "jemalloc/internal/slab_data.h"
diff --git a/include/jemalloc/internal/edata_cache.h b/include/jemalloc/internal/edata_cache.h
index 8b6c0ef7..b2c7b4f1 100644
--- a/include/jemalloc/internal/edata_cache.h
+++ b/include/jemalloc/internal/edata_cache.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_EDATA_CACHE_H
 #define JEMALLOC_INTERNAL_EDATA_CACHE_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/base.h"
 
 /* For tests only. */
diff --git a/include/jemalloc/internal/ehooks.h b/include/jemalloc/internal/ehooks.h
index d583c521..947e056c 100644
--- a/include/jemalloc/internal/ehooks.h
+++ b/include/jemalloc/internal/ehooks.h
@@ -1,8 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_EHOOKS_H
 #define JEMALLOC_INTERNAL_EHOOKS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/extent_mmap.h"
+#include "jemalloc/internal/tsd.h"
+#include "jemalloc/internal/tsd_types.h"
 
 /*
  * This module is the internal interface to the extent hooks (both
diff --git a/include/jemalloc/internal/emap.h b/include/jemalloc/internal/emap.h
index 847af327..08262f1f 100644
--- a/include/jemalloc/internal/emap.h
+++ b/include/jemalloc/internal/emap.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_EMAP_H
 #define JEMALLOC_INTERNAL_EMAP_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/base.h"
 #include "jemalloc/internal/rtree.h"
 
diff --git a/include/jemalloc/internal/emitter.h b/include/jemalloc/internal/emitter.h
index 9482f68b..bc12fe92 100644
--- a/include/jemalloc/internal/emitter.h
+++ b/include/jemalloc/internal/emitter.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_EMITTER_H
 #define JEMALLOC_INTERNAL_EMITTER_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/assert.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/malloc_io.h"
 #include "jemalloc/internal/ql.h"
 
 typedef enum emitter_output_e emitter_output_t;
diff --git a/include/jemalloc/internal/eset.h b/include/jemalloc/internal/eset.h
index 4f689b47..9b7c4a89 100644
--- a/include/jemalloc/internal/eset.h
+++ b/include/jemalloc/internal/eset.h
@@ -1,9 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_ESET_H
 #define JEMALLOC_INTERNAL_ESET_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
-#include "jemalloc/internal/fb.h"
 #include "jemalloc/internal/edata.h"
+#include "jemalloc/internal/fb.h"
 #include "jemalloc/internal/mutex.h"
 
 /*
diff --git a/include/jemalloc/internal/exp_grow.h b/include/jemalloc/internal/exp_grow.h
index 8566b8a4..40a1add0 100644
--- a/include/jemalloc/internal/exp_grow.h
+++ b/include/jemalloc/internal/exp_grow.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_EXP_GROW_H
 #define JEMALLOC_INTERNAL_EXP_GROW_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/sz.h"
 typedef struct exp_grow_s exp_grow_t;
 struct exp_grow_s {
 	/*
diff --git a/include/jemalloc/internal/extent.h b/include/jemalloc/internal/extent.h
index 367793db..17feb703 100644
--- a/include/jemalloc/internal/extent.h
+++ b/include/jemalloc/internal/extent.h
@@ -1,8 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_EXTENT_H
 #define JEMALLOC_INTERNAL_EXTENT_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/ecache.h"
 #include "jemalloc/internal/ehooks.h"
+#include "jemalloc/internal/pac.h"
 #include "jemalloc/internal/ph.h"
 #include "jemalloc/internal/rtree.h"
 
diff --git a/include/jemalloc/internal/extent_dss.h b/include/jemalloc/internal/extent_dss.h
index 38f04340..c8e71e82 100644
--- a/include/jemalloc/internal/extent_dss.h
+++ b/include/jemalloc/internal/extent_dss.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_EXTENT_DSS_H
 #define JEMALLOC_INTERNAL_EXTENT_DSS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_types.h"
+#include "jemalloc/internal/tsd_types.h"
+
 typedef enum {
 	dss_prec_disabled  = 0,
 	dss_prec_primary   = 1,
diff --git a/include/jemalloc/internal/extent_mmap.h b/include/jemalloc/internal/extent_mmap.h
index 55f17ee4..e6a4649e 100644
--- a/include/jemalloc/internal/extent_mmap.h
+++ b/include/jemalloc/internal/extent_mmap.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H
 #define JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 extern bool opt_retain;
 
 void *extent_alloc_mmap(void *new_addr, size_t size, size_t alignment,
diff --git a/include/jemalloc/internal/fb.h b/include/jemalloc/internal/fb.h
index 90c4091f..e38095af 100644
--- a/include/jemalloc/internal/fb.h
+++ b/include/jemalloc/internal/fb.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_FB_H
 #define JEMALLOC_INTERNAL_FB_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/assert.h"
+#include "jemalloc/internal/bit_util.h"
+
 /*
  * The flat bitmap module.  This has a larger API relative to the bitmap module
  * (supporting things like backwards searches, and searching for both set and
diff --git a/include/jemalloc/internal/fxp.h b/include/jemalloc/internal/fxp.h
index 415a9828..e42425f9 100644
--- a/include/jemalloc/internal/fxp.h
+++ b/include/jemalloc/internal/fxp.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_FXP_H
 #define JEMALLOC_INTERNAL_FXP_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/assert.h"
+
 /*
  * A simple fixed-point math implementation, supporting only unsigned values
  * (with overflow being an error).
diff --git a/include/jemalloc/internal/hash.h b/include/jemalloc/internal/hash.h
index 7f945679..15162b94 100644
--- a/include/jemalloc/internal/hash.h
+++ b/include/jemalloc/internal/hash.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_HASH_H
 #define JEMALLOC_INTERNAL_HASH_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/assert.h"
 
 /*
diff --git a/include/jemalloc/internal/hook.h b/include/jemalloc/internal/hook.h
index 27f94841..76b9130d 100644
--- a/include/jemalloc/internal/hook.h
+++ b/include/jemalloc/internal/hook.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_HOOK_H
 #define JEMALLOC_INTERNAL_HOOK_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/tsd.h"
 
 /*
diff --git a/include/jemalloc/internal/hpa.h b/include/jemalloc/internal/hpa.h
index 01fe3166..4805efaf 100644
--- a/include/jemalloc/internal/hpa.h
+++ b/include/jemalloc/internal/hpa.h
@@ -1,9 +1,14 @@
 #ifndef JEMALLOC_INTERNAL_HPA_H
 #define JEMALLOC_INTERNAL_HPA_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/base.h"
+#include "jemalloc/internal/edata_cache.h"
+#include "jemalloc/internal/emap.h"
 #include "jemalloc/internal/exp_grow.h"
 #include "jemalloc/internal/hpa_hooks.h"
 #include "jemalloc/internal/hpa_opts.h"
+#include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/pai.h"
 #include "jemalloc/internal/psset.h"
 
diff --git a/include/jemalloc/internal/hpa_hooks.h b/include/jemalloc/internal/hpa_hooks.h
index 841f529e..72f3a43c 100644
--- a/include/jemalloc/internal/hpa_hooks.h
+++ b/include/jemalloc/internal/hpa_hooks.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_HPA_HOOKS_H
 #define JEMALLOC_INTERNAL_HPA_HOOKS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/nstime.h"
+
 typedef struct hpa_hooks_s hpa_hooks_t;
 struct hpa_hooks_s {
 	void *(*map)(size_t size);
diff --git a/include/jemalloc/internal/hpa_opts.h b/include/jemalloc/internal/hpa_opts.h
index ee84fea1..6e58c86b 100644
--- a/include/jemalloc/internal/hpa_opts.h
+++ b/include/jemalloc/internal/hpa_opts.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_HPA_OPTS_H
 #define JEMALLOC_INTERNAL_HPA_OPTS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/fxp.h"
 
 /*
diff --git a/include/jemalloc/internal/hpdata.h b/include/jemalloc/internal/hpdata.h
index 36918258..7ba92112 100644
--- a/include/jemalloc/internal/hpdata.h
+++ b/include/jemalloc/internal/hpdata.h
@@ -1,7 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_HPDATA_H
 #define JEMALLOC_INTERNAL_HPDATA_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/fb.h"
+#include "jemalloc/internal/nstime.h"
+#include "jemalloc/internal/pages.h"
 #include "jemalloc/internal/ph.h"
 #include "jemalloc/internal/ql.h"
 #include "jemalloc/internal/typed_list.h"
diff --git a/include/jemalloc/internal/inspect.h b/include/jemalloc/internal/inspect.h
index 65fef51d..0da920ca 100644
--- a/include/jemalloc/internal/inspect.h
+++ b/include/jemalloc/internal/inspect.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_INSPECT_H
 #define JEMALLOC_INTERNAL_INSPECT_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/tsd_types.h"
+
 /*
  * This module contains the heap introspection capabilities.  For now they are
  * exposed purely through mallctl APIs in the experimental namespace, but this
diff --git a/include/jemalloc/internal/jemalloc_internal_externs.h b/include/jemalloc/internal/jemalloc_internal_externs.h
index b1e5bde9..ae03c644 100644
--- a/include/jemalloc/internal/jemalloc_internal_externs.h
+++ b/include/jemalloc/internal/jemalloc_internal_externs.h
@@ -1,11 +1,12 @@
 #ifndef JEMALLOC_INTERNAL_EXTERNS_H
 #define JEMALLOC_INTERNAL_EXTERNS_H
 
+#include "jemalloc/internal/arena_types.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/hpa_opts.h"
+#include "jemalloc/internal/nstime.h"
 #include "jemalloc/internal/sec_opts.h"
 #include "jemalloc/internal/tsd_types.h"
-#include "jemalloc/internal/nstime.h"
 
 /* TSD checks this to set thread local slow state accordingly. */
 extern bool malloc_slow;
diff --git a/include/jemalloc/internal/jemalloc_internal_inlines_a.h b/include/jemalloc/internal/jemalloc_internal_inlines_a.h
index cb6d78fa..8d5e22fd 100644
--- a/include/jemalloc/internal/jemalloc_internal_inlines_a.h
+++ b/include/jemalloc/internal/jemalloc_internal_inlines_a.h
@@ -1,10 +1,14 @@
 #ifndef JEMALLOC_INTERNAL_INLINES_A_H
 #define JEMALLOC_INTERNAL_INLINES_A_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_externs.h"
+#include "jemalloc/internal/arena_types.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/bit_util.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
 #include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/tcache_externs.h"
 #include "jemalloc/internal/ticker.h"
 
 JEMALLOC_ALWAYS_INLINE malloc_cpuid_t
diff --git a/include/jemalloc/internal/jemalloc_internal_inlines_b.h b/include/jemalloc/internal/jemalloc_internal_inlines_b.h
index 152f8a03..b2cab228 100644
--- a/include/jemalloc/internal/jemalloc_internal_inlines_b.h
+++ b/include/jemalloc/internal/jemalloc_internal_inlines_b.h
@@ -1,7 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_INLINES_B_H
 #define JEMALLOC_INTERNAL_INLINES_B_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_inlines_a.h"
 #include "jemalloc/internal/extent.h"
+#include "jemalloc/internal/jemalloc_internal_inlines_a.h"
 
 static inline void
 percpu_arena_update(tsd_t *tsd, unsigned cpu) {
diff --git a/include/jemalloc/internal/jemalloc_internal_inlines_c.h b/include/jemalloc/internal/jemalloc_internal_inlines_c.h
index ae9cb0c2..1dac668a 100644
--- a/include/jemalloc/internal/jemalloc_internal_inlines_c.h
+++ b/include/jemalloc/internal/jemalloc_internal_inlines_c.h
@@ -1,14 +1,16 @@
 #ifndef JEMALLOC_INTERNAL_INLINES_C_H
 #define JEMALLOC_INTERNAL_INLINES_C_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_externs.h"
+#include "jemalloc/internal/arena_inlines_b.h"
+#include "jemalloc/internal/emap.h"
 #include "jemalloc/internal/hook.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
 #include "jemalloc/internal/log.h"
 #include "jemalloc/internal/sz.h"
 #include "jemalloc/internal/thread_event.h"
 #include "jemalloc/internal/witness.h"
-#include "jemalloc/internal/arena_externs.h"
-#include "jemalloc/internal/emap.h"
 
 /*
  * These correspond to the macros in jemalloc/jemalloc_macros.h.  Broadly, we
diff --git a/include/jemalloc/internal/jemalloc_preamble.h.in b/include/jemalloc/internal/jemalloc_preamble.h.in
index d7086302..6b55e47f 100644
--- a/include/jemalloc/internal/jemalloc_preamble.h.in
+++ b/include/jemalloc/internal/jemalloc_preamble.h.in
@@ -1,7 +1,7 @@
 #ifndef JEMALLOC_PREAMBLE_H
 #define JEMALLOC_PREAMBLE_H
 
-#include "jemalloc_internal_defs.h"
+#include "jemalloc/internal/jemalloc_internal_defs.h"
 #include "jemalloc/internal/jemalloc_internal_decls.h"
 
 #if defined(JEMALLOC_UTRACE) || defined(JEMALLOC_UTRACE_LABEL)
diff --git a/include/jemalloc/internal/large_externs.h b/include/jemalloc/internal/large_externs.h
index 8e09122d..ce9c8689 100644
--- a/include/jemalloc/internal/large_externs.h
+++ b/include/jemalloc/internal/large_externs.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_LARGE_EXTERNS_H
 #define JEMALLOC_INTERNAL_LARGE_EXTERNS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/edata.h"
 #include "jemalloc/internal/hook.h"
 
 void *large_malloc(tsdn_t *tsdn, arena_t *arena, size_t usize, bool zero);
diff --git a/include/jemalloc/internal/lockedint.h b/include/jemalloc/internal/lockedint.h
index d020ebec..062dedbf 100644
--- a/include/jemalloc/internal/lockedint.h
+++ b/include/jemalloc/internal/lockedint.h
@@ -1,6 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_LOCKEDINT_H
 #define JEMALLOC_INTERNAL_LOCKEDINT_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/tsd_types.h"
+
 /*
  * In those architectures that support 64-bit atomics, we use atomic updates for
  * our 64-bit values.  Otherwise, we use a plain uint64_t and synchronize
diff --git a/include/jemalloc/internal/log.h b/include/jemalloc/internal/log.h
index f39c598a..921985c8 100644
--- a/include/jemalloc/internal/log.h
+++ b/include/jemalloc/internal/log.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_LOG_H
 #define JEMALLOC_INTERNAL_LOG_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/malloc_io.h"
 #include "jemalloc/internal/mutex.h"
diff --git a/include/jemalloc/internal/malloc_io.h b/include/jemalloc/internal/malloc_io.h
index a375bdae..0afb0429 100644
--- a/include/jemalloc/internal/malloc_io.h
+++ b/include/jemalloc/internal/malloc_io.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_MALLOC_IO_H
 #define JEMALLOC_INTERNAL_MALLOC_IO_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
 
 #ifdef _WIN32
diff --git a/include/jemalloc/internal/mpsc_queue.h b/include/jemalloc/internal/mpsc_queue.h
index 316ea9b1..d8aa624b 100644
--- a/include/jemalloc/internal/mpsc_queue.h
+++ b/include/jemalloc/internal/mpsc_queue.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_MPSC_QUEUE_H
 #define JEMALLOC_INTERNAL_MPSC_QUEUE_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 
 /*
diff --git a/include/jemalloc/internal/mutex.h b/include/jemalloc/internal/mutex.h
index 03d3557b..46f22aec 100644
--- a/include/jemalloc/internal/mutex.h
+++ b/include/jemalloc/internal/mutex.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_MUTEX_H
 #define JEMALLOC_INTERNAL_MUTEX_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/mutex_prof.h"
 #include "jemalloc/internal/tsd.h"
diff --git a/include/jemalloc/internal/mutex_prof.h b/include/jemalloc/internal/mutex_prof.h
index 4a526a5a..14e4340b 100644
--- a/include/jemalloc/internal/mutex_prof.h
+++ b/include/jemalloc/internal/mutex_prof.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_MUTEX_PROF_H
 #define JEMALLOC_INTERNAL_MUTEX_PROF_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 #include "jemalloc/internal/nstime.h"
 #include "jemalloc/internal/tsd_types.h"
diff --git a/include/jemalloc/internal/nstime.h b/include/jemalloc/internal/nstime.h
index ad1ae532..440a4d15 100644
--- a/include/jemalloc/internal/nstime.h
+++ b/include/jemalloc/internal/nstime.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_NSTIME_H
 #define JEMALLOC_INTERNAL_NSTIME_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/assert.h"
+
 /* Maximum supported number of seconds (~584 years). */
 #define NSTIME_SEC_MAX KQU(18446744072)
 
diff --git a/include/jemalloc/internal/pa.h b/include/jemalloc/internal/pa.h
index c8aed932..5f43244d 100644
--- a/include/jemalloc/internal/pa.h
+++ b/include/jemalloc/internal/pa.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_PA_H
 #define JEMALLOC_INTERNAL_PA_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/base.h"
 #include "jemalloc/internal/decay.h"
 #include "jemalloc/internal/ecache.h"
diff --git a/include/jemalloc/internal/pac.h b/include/jemalloc/internal/pac.h
index 01c4e6af..0b173a58 100644
--- a/include/jemalloc/internal/pac.h
+++ b/include/jemalloc/internal/pac.h
@@ -1,11 +1,15 @@
 #ifndef JEMALLOC_INTERNAL_PAC_H
 #define JEMALLOC_INTERNAL_PAC_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/decay.h"
+#include "jemalloc/internal/ecache.h"
+#include "jemalloc/internal/edata_cache.h"
 #include "jemalloc/internal/exp_grow.h"
+#include "jemalloc/internal/lockedint.h"
 #include "jemalloc/internal/pai.h"
 #include "san_bump.h"
 
-
 /*
  * Page allocator classic; an implementation of the PAI interface that:
  * - Can be used for arenas with custom extent hooks.
diff --git a/include/jemalloc/internal/pages.h b/include/jemalloc/internal/pages.h
index c9d10ce2..0ecc2cd0 100644
--- a/include/jemalloc/internal/pages.h
+++ b/include/jemalloc/internal/pages.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_PAGES_EXTERNS_H
 #define JEMALLOC_INTERNAL_PAGES_EXTERNS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 /* Actual operating system page size, detected during bootstrap, <= PAGE. */
 extern size_t	os_page;
 
diff --git a/include/jemalloc/internal/pai.h b/include/jemalloc/internal/pai.h
index d978cd7d..dd64ee59 100644
--- a/include/jemalloc/internal/pai.h
+++ b/include/jemalloc/internal/pai.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_PAI_H
 #define JEMALLOC_INTERNAL_PAI_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/edata.h"
+#include "jemalloc/internal/tsd_types.h"
+
 /* An interface for page allocation. */
 
 typedef struct pai_s pai_t;
diff --git a/include/jemalloc/internal/peak.h b/include/jemalloc/internal/peak.h
index 59da3e41..2a973cb8 100644
--- a/include/jemalloc/internal/peak.h
+++ b/include/jemalloc/internal/peak.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_PEAK_H
 #define JEMALLOC_INTERNAL_PEAK_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 typedef struct peak_s peak_t;
 struct peak_s {
 	/* The highest recorded peak value, after adjustment (see below). */
diff --git a/include/jemalloc/internal/peak_event.h b/include/jemalloc/internal/peak_event.h
index b808ce04..cc2a1401 100644
--- a/include/jemalloc/internal/peak_event.h
+++ b/include/jemalloc/internal/peak_event.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_PEAK_EVENT_H
 #define JEMALLOC_INTERNAL_PEAK_EVENT_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/tsd_types.h"
+
 /*
  * While peak.h contains the simple helper struct that tracks state, this
  * contains the allocator tie-ins (and knows about tsd, the event module, etc.).
diff --git a/include/jemalloc/internal/ph.h b/include/jemalloc/internal/ph.h
index 89de8663..3ae38710 100644
--- a/include/jemalloc/internal/ph.h
+++ b/include/jemalloc/internal/ph.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_PH_H
 #define JEMALLOC_INTERNAL_PH_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/assert.h"
+#include "jemalloc/internal/bit_util.h"
+
 /*
  * A Pairing Heap implementation.
  *
diff --git a/include/jemalloc/internal/prng.h b/include/jemalloc/internal/prng.h
index 14542aa1..81060d32 100644
--- a/include/jemalloc/internal/prng.h
+++ b/include/jemalloc/internal/prng.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_PRNG_H
 #define JEMALLOC_INTERNAL_PRNG_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/bit_util.h"
 
 /*
diff --git a/include/jemalloc/internal/prof_data.h b/include/jemalloc/internal/prof_data.h
index 016b6507..43e8d7e7 100644
--- a/include/jemalloc/internal/prof_data.h
+++ b/include/jemalloc/internal/prof_data.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_PROF_DATA_H
 #define JEMALLOC_INTERNAL_PROF_DATA_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/mutex.h"
 
 extern malloc_mutex_t bt2gctx_mtx;
diff --git a/include/jemalloc/internal/prof_externs.h b/include/jemalloc/internal/prof_externs.h
index cce5c8f5..514c5804 100644
--- a/include/jemalloc/internal/prof_externs.h
+++ b/include/jemalloc/internal/prof_externs.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_PROF_EXTERNS_H
 #define JEMALLOC_INTERNAL_PROF_EXTERNS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/base.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/prof_hook.h"
 
diff --git a/include/jemalloc/internal/prof_hook.h b/include/jemalloc/internal/prof_hook.h
index 8615dc53..3c5ff8bf 100644
--- a/include/jemalloc/internal/prof_hook.h
+++ b/include/jemalloc/internal/prof_hook.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_PROF_HOOK_H
 #define JEMALLOC_INTERNAL_PROF_HOOK_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 /*
  * The hooks types of which are declared in this file are experimental and
  * undocumented, thus the typedefs are located in an 'internal' header.
diff --git a/include/jemalloc/internal/prof_inlines.h b/include/jemalloc/internal/prof_inlines.h
index c0783fc1..6cb73735 100644
--- a/include/jemalloc/internal/prof_inlines.h
+++ b/include/jemalloc/internal/prof_inlines.h
@@ -1,10 +1,14 @@
 #ifndef JEMALLOC_INTERNAL_PROF_INLINES_H
 #define JEMALLOC_INTERNAL_PROF_INLINES_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_inlines_b.h"
+#include "jemalloc/internal/jemalloc_internal_inlines_c.h"
+#include "jemalloc/internal/prof_externs.h"
+#include "jemalloc/internal/prof_structs.h"
 #include "jemalloc/internal/safety_check.h"
 #include "jemalloc/internal/sz.h"
 #include "jemalloc/internal/thread_event.h"
-#include "jemalloc/internal/jemalloc_internal_inlines_c.h"
 
 JEMALLOC_ALWAYS_INLINE void
 prof_active_assert(void) {
diff --git a/include/jemalloc/internal/prof_log.h b/include/jemalloc/internal/prof_log.h
index ccb557dd..0b1271c8 100644
--- a/include/jemalloc/internal/prof_log.h
+++ b/include/jemalloc/internal/prof_log.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_PROF_LOG_H
 #define JEMALLOC_INTERNAL_PROF_LOG_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/mutex.h"
 
 extern malloc_mutex_t log_mtx;
diff --git a/include/jemalloc/internal/prof_recent.h b/include/jemalloc/internal/prof_recent.h
index 959e336b..33649e6d 100644
--- a/include/jemalloc/internal/prof_recent.h
+++ b/include/jemalloc/internal/prof_recent.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_PROF_RECENT_H
 #define JEMALLOC_INTERNAL_PROF_RECENT_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/edata.h"
+#include "jemalloc/internal/mutex.h"
+
 extern malloc_mutex_t prof_recent_alloc_mtx;
 extern malloc_mutex_t prof_recent_dump_mtx;
 
diff --git a/include/jemalloc/internal/prof_stats.h b/include/jemalloc/internal/prof_stats.h
index 7954e82d..c4d269e5 100644
--- a/include/jemalloc/internal/prof_stats.h
+++ b/include/jemalloc/internal/prof_stats.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_PROF_STATS_H
 #define JEMALLOC_INTERNAL_PROF_STATS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/mutex.h"
+
 typedef struct prof_stats_s prof_stats_t;
 struct prof_stats_s {
 	uint64_t req_sum;
diff --git a/include/jemalloc/internal/prof_structs.h b/include/jemalloc/internal/prof_structs.h
index 49061f02..084a549d 100644
--- a/include/jemalloc/internal/prof_structs.h
+++ b/include/jemalloc/internal/prof_structs.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_PROF_STRUCTS_H
 #define JEMALLOC_INTERNAL_PROF_STRUCTS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/ckh.h"
 #include "jemalloc/internal/edata.h"
 #include "jemalloc/internal/mutex.h"
diff --git a/include/jemalloc/internal/prof_sys.h b/include/jemalloc/internal/prof_sys.h
index 0eb50788..e6e7f06f 100644
--- a/include/jemalloc/internal/prof_sys.h
+++ b/include/jemalloc/internal/prof_sys.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_PROF_SYS_H
 #define JEMALLOC_INTERNAL_PROF_SYS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/base.h"
+#include "jemalloc/internal/mutex.h"
+
 extern malloc_mutex_t prof_dump_filename_mtx;
 extern base_t *prof_base;
 
diff --git a/include/jemalloc/internal/psset.h b/include/jemalloc/internal/psset.h
index e1d64970..7e510b7f 100644
--- a/include/jemalloc/internal/psset.h
+++ b/include/jemalloc/internal/psset.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_PSSET_H
 #define JEMALLOC_INTERNAL_PSSET_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/hpdata.h"
 
 /*
diff --git a/include/jemalloc/internal/ql.h b/include/jemalloc/internal/ql.h
index c7f52f86..ebe69988 100644
--- a/include/jemalloc/internal/ql.h
+++ b/include/jemalloc/internal/ql.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_QL_H
 #define JEMALLOC_INTERNAL_QL_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/qr.h"
 
 /*
diff --git a/include/jemalloc/internal/rb.h b/include/jemalloc/internal/rb.h
index 343e7c13..5f2771a9 100644
--- a/include/jemalloc/internal/rb.h
+++ b/include/jemalloc/internal/rb.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_RB_H
 #define JEMALLOC_INTERNAL_RB_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/safety_check.h"
+
 /*-
  *******************************************************************************
  *
diff --git a/include/jemalloc/internal/rtree.h b/include/jemalloc/internal/rtree.h
index 22f5f9dc..f559c94f 100644
--- a/include/jemalloc/internal/rtree.h
+++ b/include/jemalloc/internal/rtree.h
@@ -1,7 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_RTREE_H
 #define JEMALLOC_INTERNAL_RTREE_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/base.h"
 #include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/edata.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/rtree_tsd.h"
 #include "jemalloc/internal/sc.h"
diff --git a/include/jemalloc/internal/rtree_tsd.h b/include/jemalloc/internal/rtree_tsd.h
index e45525c5..59f18570 100644
--- a/include/jemalloc/internal/rtree_tsd.h
+++ b/include/jemalloc/internal/rtree_tsd.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_RTREE_CTX_H
 #define JEMALLOC_INTERNAL_RTREE_CTX_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 /*
  * Number of leafkey/leaf pairs to cache in L1 and L2 level respectively.  Each
  * entry supports an entire leaf, so the cache hit rate is typically high even
diff --git a/include/jemalloc/internal/safety_check.h b/include/jemalloc/internal/safety_check.h
index 7854c1bf..ef778dae 100644
--- a/include/jemalloc/internal/safety_check.h
+++ b/include/jemalloc/internal/safety_check.h
@@ -1,10 +1,12 @@
 #ifndef JEMALLOC_INTERNAL_SAFETY_CHECK_H
 #define JEMALLOC_INTERNAL_SAFETY_CHECK_H
 
-#define SAFETY_CHECK_DOUBLE_FREE_MAX_SCAN_DEFAULT 32
-
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/assert.h"
 #include "jemalloc/internal/pages.h"
 
+#define SAFETY_CHECK_DOUBLE_FREE_MAX_SCAN_DEFAULT 32
+
 void safety_check_fail_sized_dealloc(bool current_dealloc, const void *ptr,
     size_t true_size, size_t input_size);
 void safety_check_fail(const char *format, ...);
diff --git a/include/jemalloc/internal/san.h b/include/jemalloc/internal/san.h
index 8813d6bb..79723965 100644
--- a/include/jemalloc/internal/san.h
+++ b/include/jemalloc/internal/san.h
@@ -1,8 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_GUARD_H
 #define JEMALLOC_INTERNAL_GUARD_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/ehooks.h"
 #include "jemalloc/internal/emap.h"
+#include "jemalloc/internal/jemalloc_internal_externs.h"
+#include "jemalloc/internal/tsd.h"
 
 #define SAN_PAGE_GUARD PAGE
 #define SAN_PAGE_GUARDS_SIZE (SAN_PAGE_GUARD * 2)
diff --git a/include/jemalloc/internal/san_bump.h b/include/jemalloc/internal/san_bump.h
index 0a8e76e9..d6e9cfc5 100644
--- a/include/jemalloc/internal/san_bump.h
+++ b/include/jemalloc/internal/san_bump.h
@@ -1,9 +1,11 @@
 #ifndef JEMALLOC_INTERNAL_SAN_BUMP_H
 #define JEMALLOC_INTERNAL_SAN_BUMP_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/edata.h"
 #include "jemalloc/internal/exp_grow.h"
 #include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/witness.h"
 
 #define SBA_RETAINED_ALLOC_SIZE ((size_t)4 << 20)
 
diff --git a/include/jemalloc/internal/sc.h b/include/jemalloc/internal/sc.h
index 9bab347b..770835cc 100644
--- a/include/jemalloc/internal/sc.h
+++ b/include/jemalloc/internal/sc.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_SC_H
 #define JEMALLOC_INTERNAL_SC_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
 
 /*
diff --git a/include/jemalloc/internal/sec.h b/include/jemalloc/internal/sec.h
index fa863382..8ef1e9fb 100644
--- a/include/jemalloc/internal/sec.h
+++ b/include/jemalloc/internal/sec.h
@@ -1,8 +1,12 @@
 #ifndef JEMALLOC_INTERNAL_SEC_H
 #define JEMALLOC_INTERNAL_SEC_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/base.h"
 #include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/pai.h"
+#include "jemalloc/internal/sec_opts.h"
 
 /*
  * Small extent cache.
diff --git a/include/jemalloc/internal/sec_opts.h b/include/jemalloc/internal/sec_opts.h
index a3ad72fb..19ed1492 100644
--- a/include/jemalloc/internal/sec_opts.h
+++ b/include/jemalloc/internal/sec_opts.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_SEC_OPTS_H
 #define JEMALLOC_INTERNAL_SEC_OPTS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 /*
  * The configuration settings used by an sec_t.  Morally, this is part of the
  * SEC interface, but we put it here for header-ordering reasons.
diff --git a/include/jemalloc/internal/seq.h b/include/jemalloc/internal/seq.h
index ef2df4c6..9bb6b235 100644
--- a/include/jemalloc/internal/seq.h
+++ b/include/jemalloc/internal/seq.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_SEQ_H
 #define JEMALLOC_INTERNAL_SEQ_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/atomic.h"
 
 /*
diff --git a/include/jemalloc/internal/slab_data.h b/include/jemalloc/internal/slab_data.h
index e821863d..724c71e3 100644
--- a/include/jemalloc/internal/slab_data.h
+++ b/include/jemalloc/internal/slab_data.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_SLAB_DATA_H
 #define JEMALLOC_INTERNAL_SLAB_DATA_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/bitmap.h"
 
 typedef struct slab_data_s slab_data_t;
diff --git a/include/jemalloc/internal/spin.h b/include/jemalloc/internal/spin.h
index 6940f15e..87c400d5 100644
--- a/include/jemalloc/internal/spin.h
+++ b/include/jemalloc/internal/spin.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_SPIN_H
 #define JEMALLOC_INTERNAL_SPIN_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 #define SPIN_INITIALIZER {0U}
 
 typedef struct {
diff --git a/include/jemalloc/internal/stats.h b/include/jemalloc/internal/stats.h
index 727f7dcb..310178ea 100644
--- a/include/jemalloc/internal/stats.h
+++ b/include/jemalloc/internal/stats.h
@@ -1,6 +1,10 @@
 #ifndef JEMALLOC_INTERNAL_STATS_H
 #define JEMALLOC_INTERNAL_STATS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/tsd_types.h"
+
 /*  OPTION(opt,		var_name,	default,	set_value_to) */
 #define STATS_PRINT_OPTIONS						\
     OPTION('J',		json,		false,		true)		\
diff --git a/include/jemalloc/internal/sz.h b/include/jemalloc/internal/sz.h
index a799cea9..955d8ec0 100644
--- a/include/jemalloc/internal/sz.h
+++ b/include/jemalloc/internal/sz.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_SIZE_H
 #define JEMALLOC_INTERNAL_SIZE_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/bit_util.h"
 #include "jemalloc/internal/pages.h"
 #include "jemalloc/internal/sc.h"
diff --git a/include/jemalloc/internal/tcache_externs.h b/include/jemalloc/internal/tcache_externs.h
index 37f61646..af6fd970 100644
--- a/include/jemalloc/internal/tcache_externs.h
+++ b/include/jemalloc/internal/tcache_externs.h
@@ -1,6 +1,12 @@
 #ifndef JEMALLOC_INTERNAL_TCACHE_EXTERNS_H
 #define JEMALLOC_INTERNAL_TCACHE_EXTERNS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/base.h"
+#include "jemalloc/internal/cache_bin.h"
+#include "jemalloc/internal/sz.h"
+#include "jemalloc/internal/tcache_types.h"
+
 extern bool opt_tcache;
 extern size_t opt_tcache_max;
 extern ssize_t	opt_lg_tcache_nslots_mul;
diff --git a/include/jemalloc/internal/tcache_inlines.h b/include/jemalloc/internal/tcache_inlines.h
index 2b8db0a3..b69d89ad 100644
--- a/include/jemalloc/internal/tcache_inlines.h
+++ b/include/jemalloc/internal/tcache_inlines.h
@@ -1,11 +1,16 @@
 #ifndef JEMALLOC_INTERNAL_TCACHE_INLINES_H
 #define JEMALLOC_INTERNAL_TCACHE_INLINES_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/arena_externs.h"
 #include "jemalloc/internal/bin.h"
+#include "jemalloc/internal/jemalloc_internal_inlines_b.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
+#include "jemalloc/internal/large_externs.h"
 #include "jemalloc/internal/san.h"
 #include "jemalloc/internal/sc.h"
 #include "jemalloc/internal/sz.h"
+#include "jemalloc/internal/tcache_externs.h"
 #include "jemalloc/internal/util.h"
 
 static inline bool
diff --git a/include/jemalloc/internal/tcache_structs.h b/include/jemalloc/internal/tcache_structs.h
index 176d73de..75918158 100644
--- a/include/jemalloc/internal/tcache_structs.h
+++ b/include/jemalloc/internal/tcache_structs.h
@@ -1,11 +1,12 @@
 #ifndef JEMALLOC_INTERNAL_TCACHE_STRUCTS_H
 #define JEMALLOC_INTERNAL_TCACHE_STRUCTS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/cache_bin.h"
 #include "jemalloc/internal/ql.h"
 #include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/tcache_types.h"
 #include "jemalloc/internal/ticker.h"
-#include "jemalloc/internal/tsd_types.h"
 
 /*
  * The tcache state is split into the slow and hot path data.  Each has a
diff --git a/include/jemalloc/internal/tcache_types.h b/include/jemalloc/internal/tcache_types.h
index 583677ea..cea86fb7 100644
--- a/include/jemalloc/internal/tcache_types.h
+++ b/include/jemalloc/internal/tcache_types.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_TCACHE_TYPES_H
 #define JEMALLOC_INTERNAL_TCACHE_TYPES_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/sc.h"
 
 typedef struct tcache_slow_s tcache_slow_t;
diff --git a/include/jemalloc/internal/test_hooks.h b/include/jemalloc/internal/test_hooks.h
index 2b90afe1..af3f2755 100644
--- a/include/jemalloc/internal/test_hooks.h
+++ b/include/jemalloc/internal/test_hooks.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_TEST_HOOKS_H
 #define JEMALLOC_INTERNAL_TEST_HOOKS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 extern JEMALLOC_EXPORT void (*test_hooks_arena_new_hook)(void);
 extern JEMALLOC_EXPORT void (*test_hooks_libc_hook)(void);
 
diff --git a/include/jemalloc/internal/thread_event.h b/include/jemalloc/internal/thread_event.h
index 2f4e1b39..46c57ed5 100644
--- a/include/jemalloc/internal/thread_event.h
+++ b/include/jemalloc/internal/thread_event.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_THREAD_EVENT_H
 #define JEMALLOC_INTERNAL_THREAD_EVENT_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/tsd.h"
 
 /* "te" is short for "thread_event" */
diff --git a/include/jemalloc/internal/ticker.h b/include/jemalloc/internal/ticker.h
index de034995..dca9bd10 100644
--- a/include/jemalloc/internal/ticker.h
+++ b/include/jemalloc/internal/ticker.h
@@ -1,6 +1,7 @@
 #ifndef JEMALLOC_INTERNAL_TICKER_H
 #define JEMALLOC_INTERNAL_TICKER_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/prng.h"
 #include "jemalloc/internal/util.h"
 
diff --git a/include/jemalloc/internal/tsd.h b/include/jemalloc/internal/tsd.h
index e36ffc6d..4f22dcff 100644
--- a/include/jemalloc/internal/tsd.h
+++ b/include/jemalloc/internal/tsd.h
@@ -7,6 +7,7 @@
  * tsd_boot1, tsd_boot, tsd_booted_get, tsd_get_allocates, tsd_get, and tsd_set.
  */
 #ifdef JEMALLOC_MALLOC_THREAD_CLEANUP
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/tsd_malloc_thread_cleanup.h"
 #elif (defined(JEMALLOC_TLS))
 #include "jemalloc/internal/tsd_tls.h"
diff --git a/include/jemalloc/internal/tsd_generic.h b/include/jemalloc/internal/tsd_generic.h
index c5648f63..aa8042a4 100644
--- a/include/jemalloc/internal/tsd_generic.h
+++ b/include/jemalloc/internal/tsd_generic.h
@@ -3,7 +3,10 @@
 #endif
 #define JEMALLOC_INTERNAL_TSD_GENERIC_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/ql.h"
 #include "jemalloc/internal/tsd_internals.h"
+#include "jemalloc/internal/tsd_types.h"
 
 typedef struct tsd_init_block_s tsd_init_block_t;
 struct tsd_init_block_s {
diff --git a/include/jemalloc/internal/tsd_internals.h b/include/jemalloc/internal/tsd_internals.h
index 813580c0..439f1d10 100644
--- a/include/jemalloc/internal/tsd_internals.h
+++ b/include/jemalloc/internal/tsd_internals.h
@@ -3,6 +3,7 @@
 #endif
 #define JEMALLOC_INTERNAL_TSD_INTERNALS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/activity_callback.h"
 #include "jemalloc/internal/arena_types.h"
 #include "jemalloc/internal/assert.h"
@@ -12,11 +13,11 @@
 #include "jemalloc/internal/prof_types.h"
 #include "jemalloc/internal/ql.h"
 #include "jemalloc/internal/rtree_tsd.h"
-#include "jemalloc/internal/tcache_types.h"
 #include "jemalloc/internal/tcache_structs.h"
+#include "jemalloc/internal/tcache_types.h"
+#include "jemalloc/internal/tsd_types.h"
 #include "jemalloc/internal/util.h"
 #include "jemalloc/internal/witness.h"
-#include "jemalloc/internal/tsd_types.h"
 
 /*
  * Thread-Specific-Data layout
diff --git a/include/jemalloc/internal/tsd_malloc_thread_cleanup.h b/include/jemalloc/internal/tsd_malloc_thread_cleanup.h
index a6bd3f58..fb9ea1b4 100644
--- a/include/jemalloc/internal/tsd_malloc_thread_cleanup.h
+++ b/include/jemalloc/internal/tsd_malloc_thread_cleanup.h
@@ -3,7 +3,9 @@
 #endif
 #define JEMALLOC_INTERNAL_TSD_MALLOC_THREAD_CLEANUP_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/tsd_internals.h"
+#include "jemalloc/internal/tsd_types.h"
 
 #define JEMALLOC_TSD_TYPE_ATTR(type) __thread type JEMALLOC_TLS_MODEL
 
diff --git a/include/jemalloc/internal/tsd_tls.h b/include/jemalloc/internal/tsd_tls.h
index cd0ddbd1..5e5a6e5e 100644
--- a/include/jemalloc/internal/tsd_tls.h
+++ b/include/jemalloc/internal/tsd_tls.h
@@ -3,7 +3,9 @@
 #endif
 #define JEMALLOC_INTERNAL_TSD_TLS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/tsd_internals.h"
+#include "jemalloc/internal/tsd_types.h"
 
 #define JEMALLOC_TSD_TYPE_ATTR(type) __thread type JEMALLOC_TLS_MODEL
 
diff --git a/include/jemalloc/internal/tsd_types.h b/include/jemalloc/internal/tsd_types.h
index a6ae37da..73bbe486 100644
--- a/include/jemalloc/internal/tsd_types.h
+++ b/include/jemalloc/internal/tsd_types.h
@@ -3,6 +3,8 @@
 
 #define MALLOC_TSD_CLEANUPS_MAX	4
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+
 typedef struct tsd_s tsd_t;
 typedef struct tsdn_s tsdn_t;
 typedef bool (*malloc_tsd_cleanup_t)(void);
diff --git a/include/jemalloc/internal/tsd_win.h b/include/jemalloc/internal/tsd_win.h
index 4efeffb6..8ec7eda7 100644
--- a/include/jemalloc/internal/tsd_win.h
+++ b/include/jemalloc/internal/tsd_win.h
@@ -3,7 +3,9 @@
 #endif
 #define JEMALLOC_INTERNAL_TSD_WIN_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/tsd_internals.h"
+#include "jemalloc/internal/tsd_types.h"
 
 typedef struct {
 	bool initialized;
diff --git a/include/jemalloc/internal/util.h b/include/jemalloc/internal/util.h
index dcb1c0a5..939f3891 100644
--- a/include/jemalloc/internal/util.h
+++ b/include/jemalloc/internal/util.h
@@ -1,6 +1,9 @@
 #ifndef JEMALLOC_INTERNAL_UTIL_H
 #define JEMALLOC_INTERNAL_UTIL_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/jemalloc_internal_types.h"
+
 #define UTIL_INLINE static inline
 
 /* Junk fill patterns. */
diff --git a/include/jemalloc/internal/witness.h b/include/jemalloc/internal/witness.h
index fbe5f943..937ca2d5 100644
--- a/include/jemalloc/internal/witness.h
+++ b/include/jemalloc/internal/witness.h
@@ -1,6 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_WITNESS_H
 #define JEMALLOC_INTERNAL_WITNESS_H
 
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/assert.h"
 #include "jemalloc/internal/ql.h"
 
 /******************************************************************************/
diff --git a/include/jemalloc/jemalloc_protos.h.in b/include/jemalloc/jemalloc_protos.h.in
index 3f9fc848..aff2d88f 100644
--- a/include/jemalloc/jemalloc_protos.h.in
+++ b/include/jemalloc/jemalloc_protos.h.in
@@ -1,3 +1,6 @@
+#include "jemalloc/jemalloc_defs.h"
+#include "jemalloc/jemalloc_macros.h"
+
 /*
  * The @je_@ prefix on the following public symbol declarations is an artifact
  * of namespace management, and should be omitted in application code unless
diff --git a/test/include/test/jemalloc_test.h.in b/test/include/test/jemalloc_test.h.in
index 600d993c..f9c506da 100644
--- a/test/include/test/jemalloc_test.h.in
+++ b/test/include/test/jemalloc_test.h.in
@@ -58,14 +58,7 @@ extern "C" {
 #  include "jemalloc/jemalloc@install_suffix@.h"
 #  include "jemalloc/internal/jemalloc_internal_defs.h"
 #  include "jemalloc/internal/jemalloc_internal_macros.h"
-
-static const bool config_debug =
-#ifdef JEMALLOC_DEBUG
-    true
-#else
-    false
-#endif
-    ;
+#  include "jemalloc/internal/jemalloc_preamble.h"
 
 #  define JEMALLOC_N(n) @private_namespace@##n
 #  include "jemalloc/internal/private_namespace.h"
diff --git a/test/integration/MALLOCX_ARENA.c b/test/integration/MALLOCX_ARENA.c
index 7e61df08..440ad9ef 100644
--- a/test/integration/MALLOCX_ARENA.c
+++ b/test/integration/MALLOCX_ARENA.c
@@ -2,14 +2,6 @@
 
 #define NTHREADS 10
 
-static bool have_dss =
-#ifdef JEMALLOC_DSS
-    true
-#else
-    false
-#endif
-    ;
-
 void *
 thd_start(void *arg) {
 	unsigned thread_ind = (unsigned)(uintptr_t)arg;
diff --git a/test/integration/allocated.c b/test/integration/allocated.c
index 0c64272c..967e0108 100644
--- a/test/integration/allocated.c
+++ b/test/integration/allocated.c
@@ -1,13 +1,5 @@
 #include "test/jemalloc_test.h"
 
-static const bool config_stats =
-#ifdef JEMALLOC_STATS
-    true
-#else
-    false
-#endif
-    ;
-
 void *
 thd_start(void *arg) {
 	int err;