mdbx: add fallthrough for modern gcc.

Change-Id: I948adc0e534dd9ba42b7d64e8105870cc77f82e6

.circleci/config.yml

@@ -0,0 +1,9 @@
+version: 2
+jobs:
+  build:
+    docker:
+      - image: circleci/buildpack-deps:artful
+    steps:
+      - checkout
+      - run: make all lmdb
+      - run: make check

Makefile

@@ -1,5 +1,5 @@
 # GNU Makefile for libmdbx (reliable lightning memory-mapped DB library for Linux).
-# https://github.com/ReOpen/libmdbx
+# https://github.com/leo-yuriev/libmdbx
 
 ########################################################################
 # Configuration. The compiler options must enable threaded compilation.
@@ -71,7 +71,7 @@ clean:
 
 tests:	$(TESTS)
 
-check:	tests
+check:	tests mdbx_chk
 	[ -d testdb ] || mkdir testdb && rm -f testdb/* \
 		&& echo "*** LMDB-TEST-0" && ./mtest0 && ./mdbx_chk -v testdb \
 		&& echo "*** LMDB-TEST-1" && ./mtest1 && ./mdbx_chk -v testdb \

README.md

@@ -3,10 +3,18 @@ libmdbx
 Extended LMDB, aka "Расширенная LMDB".
 
 *The Future will Positive. Всё будет хорошо.*
-[![Build Status](https://travis-ci.org/ReOpen/libmdbx.svg?branch=master)](https://travis-ci.org/ReOpen/libmdbx)
+[![Build Status](https://travis-ci.org/leo-yuriev/libmdbx.svg?branch=stable%2F0.0)](https://travis-ci.org/leo-yuriev/libmdbx)
 
-English version by Google [is here](https://translate.googleusercontent.com/translate_c?act=url&ie=UTF8&sl=ru&tl=en&u=https://github.com/ReOpen/libmdbx/tree/master).
+## Project Status for now
 
+ - The stable versions (the _stable/*_ branches) of  are frozen, i.e. no new features or API changes, but only bug fixes.
+ - The next (the _devel_ branch) version **is under active development**, i.e. current API and set of features are extreme volatile.
+ - The immediate goal of development is formation of the stable API and the stable internal database format, which allows realise all planned features.
+ - Planned features: Integrity check by Merkle tree, Support for raw block devices, Separate place for large data items, Using "roaring bitmaps" for garbage collector, Non-linear page reclaiming, Asynchronous lazy data flush to disk(s), etc.
+
+-----
+
+English version by Google [is here](https://translate.googleusercontent.com/translate_c?act=url&ie=UTF8&sl=ru&tl=en&u=https://github.com/leo-yuriev/libmdbx/tree/stable%2F0.0).
 
 ## Кратко
 
@@ -28,7 +36,7 @@ _libmdbx_ является потомком "Lightning Memory-Mapped Database",
 известной под аббревиатурой
 [LMDB](https://en.wikipedia.org/wiki/Lightning_Memory-Mapped_Database).
 Изначально доработка производилась в составе проекта
-[ReOpenLDAP](https://github.com/ReOpen/ReOpenLDAP). Примерно за год
+[ReOpenLDAP](https://github.com/leo-yuriev/ReOpenLDAP). Примерно за год
 работы внесенные изменения приобрели самостоятельную ценность. Осенью
 2015 доработанный движок был выделен в отдельный проект, который был
 [представлен на конференции Highload++

circle.yml

@@ -1,14 +0,0 @@
-machine:
-  timezone:
-    Europe/Moscow
-
-database:
-  override:
-
-compile:
-  override:
-    - make all lmdb
-
-test:
-  override:
-    - make check

lmdb.h

@@ -212,7 +212,7 @@ typedef int mdb_filehandle_t;
 #define MDB_VERSION_DATE	"2017-02-17"
 
 /** A stringifier for the version info */
-#define MDB_VERSTR(a,b,c,d)	"MDBX " #a "." #b "." #c ": (" d ", https://github.com/ReOpen/libmdbx)"
+#define MDB_VERSTR(a,b,c,d)	"MDBX " #a "." #b "." #c ": (" d ", https://github.com/leo-yuriev/libmdbx)"
 
 /** A helper for the stringifier macro */
 #define MDB_VERFOO(a,b,c,d)	MDB_VERSTR(a,b,c,d)

mdb.c

@@ -44,10 +44,8 @@
 #	define _GNU_SOURCE
 #endif
 
-/* LY: Please do not ask us for Windows support, just never!
- * But you can make a fork for Windows, or become maintainer for FreeBSD... */
 #ifndef __gnu_linux__
-#	warning "This version of ReOpenMDBX supports only GNU Linux"
+#	warning "This version of libmdbx supports only GNU Linux"
 #endif
 
 #include <stddef.h>
@@ -56,21 +54,21 @@
 #include "./defs.h"
 
 #if !__GNUC_PREREQ(4,2)
-	/* LY: Actualy ReOpenMDBX was not tested with compilers
+	/* LY: Actualy libmdbx was not tested with compilers
 	 *     older than GCC 4.4 (from RHEL6).
 	 * But you could remove this #error and try to continue at your own risk.
 	 * In such case please don't rise up an issues related ONLY to old compilers.
 	 */
-#	warning "ReOpenMDBX required at least GCC 4.2 compatible C/C++ compiler."
+#	warning "libmdbx required at least GCC 4.2 compatible C/C++ compiler."
 #endif
 
 #if !__GLIBC_PREREQ(2,12)
-	/* LY: Actualy ReOpenMDBX was not tested with something
+	/* LY: Actualy libmdbx was not tested with something
 	 *     older than glibc 2.12 (from RHEL6).
 	 * But you could remove this #error and try to continue at your own risk.
 	 * In such case please don't rise up an issues related ONLY to old systems.
 	 */
-#	warning "ReOpenMDBX required at least GLIBC 2.12."
+#	warning "libmdbx required at least GLIBC 2.12."
 #endif
 
 #if MDB_DEBUG
@@ -1026,17 +1024,19 @@ typedef struct MDB_xcursor {
 	unsigned char mx_dbflag;
 } MDB_xcursor;
 
-	/** Check if there is an inited xcursor, so #XCURSOR_REFRESH() is proper */
+	/** Check if there is an inited xcursor */
 #define XCURSOR_INITED(mc) \
 	((mc)->mc_xcursor && ((mc)->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED))
 
-	/** Update sub-page pointer, if any, in \b mc->mc_xcursor.  Needed
+	/** Update the xcursor's sub-page pointer, if any, in \b mc.  Needed
 	 *	when the node which contains the sub-page may have moved.  Called
-	 *	with \b mp = mc->mc_pg[mc->mc_top], \b ki = mc->mc_ki[mc->mc_top].
+	 *	with leaf page \b mp = mc->mc_pg[\b top].
 	 */
-#define XCURSOR_REFRESH(mc, mp, ki) do { \
+#define XCURSOR_REFRESH(mc, top, mp) do { \
 	MDB_page *xr_pg = (mp); \
-	MDB_node *xr_node = NODEPTR(xr_pg, ki); \
+	MDB_node *xr_node; \
+	if (!XCURSOR_INITED(mc) || (mc)->mc_ki[top] >= NUMKEYS(xr_pg)) break; \
+	xr_node = NODEPTR(xr_pg, (mc)->mc_ki[top]); \
 	if ((xr_node->mn_flags & (F_DUPDATA|F_SUBDATA)) == F_DUPDATA) \
 		(mc)->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(xr_node); \
 } while (0)
@@ -1048,7 +1048,7 @@ typedef struct MDB_pgstate {
 } MDB_pgstate;
 
 	/** Context for deferred cleanup of reader's threads.
-	 *  to avoid https://github.com/ReOpen/ReOpenLDAP/issues/48 */
+	 *  to avoid https://github.com/leo-yuriev/ReOpenLDAP/issues/48 */
 typedef struct MDBX_rthc {
 	struct MDBX_rthc *rc_next;
 	pthread_t rc_thread;
@@ -2138,6 +2138,10 @@ mdb_page_alloc(MDB_cursor *mc, int num, MDB_page **mp, int flags)
 			/* If mc is updating the freeDB, then the freelist cannot play
 			 * catch-up with itself by growing while trying to save it. */
 			flags &= ~(MDBX_ALLOC_GC | MDBX_ALLOC_KICK | MDBX_COALESCE | MDBX_LIFORECLAIM);
+		} else if (unlikely(txn->mt_dbs[FREE_DBI].md_entries == 0)) {
+			/* avoid (recursive) search inside empty tree and while tree is updating,
+			 * https://github.com/leo-yuriev/libmdbx/issues/31 */
+			flags &= ~MDBX_ALLOC_GC;
 		}
 	}
 
@@ -2608,8 +2612,8 @@ done:
 			if (m2 == mc) continue;
 			if (m2->mc_pg[mc->mc_top] == mp) {
 				m2->mc_pg[mc->mc_top] = np;
-				if (XCURSOR_INITED(m2) && IS_LEAF(np))
-					XCURSOR_REFRESH(m2, np, m2->mc_ki[mc->mc_top]);
+				if (IS_LEAF(np))
+					XCURSOR_REFRESH(m2, mc->mc_top, np);
 			}
 		}
 	}
@@ -4341,7 +4345,7 @@ mdb_env_create(MDB_env **env)
 }
 
 static int __cold
-mdb_env_map(MDB_env *env, void *addr, size_t usedsize)
+mdb_env_map(MDB_env *env, void *addr)
 {
 	unsigned flags = env->me_flags;
 
@@ -4381,12 +4385,6 @@ mdb_env_map(MDB_env *env, void *addr, size_t usedsize)
 	}
 #endif
 
-#ifdef MADV_REMOVE
-	if (flags & MDB_WRITEMAP) {
-		(void) madvise(env->me_map + usedsize, env->me_mapsize - usedsize, MADV_REMOVE);
-	}
-#endif
-
 	/* Turn on/off readahead. It's harmful when the DB is larger than RAM. */
 	if (madvise(env->me_map, env->me_mapsize, (flags & MDB_NORDAHEAD) ? MADV_RANDOM : MADV_WILLNEED))
 		return errno;
@@ -4439,7 +4437,7 @@ mdb_env_set_mapsize(MDB_env *env, size_t size)
 #endif
 		env->me_mapsize = size;
 		old = (env->me_flags & MDB_FIXEDMAP) ? env->me_map : NULL;
-		rc = mdb_env_map(env, old, usedsize);
+		rc = mdb_env_map(env, old);
 		if (rc)
 			return rc;
 	}
@@ -4557,8 +4555,7 @@ mdb_env_open2(MDB_env *env, MDB_meta *meta)
 		newenv = 0;
 	}
 
-	const size_t usedsize = (meta->mm_last_pg + 1) * env->me_psize;
-	rc = mdb_env_map(env, (flags & MDB_FIXEDMAP) ? meta->mm_address : NULL, usedsize);
+	rc = mdb_env_map(env, (flags & MDB_FIXEDMAP) ? meta->mm_address : NULL);
 	if (rc)
 		return rc;
 
@@ -4631,7 +4628,7 @@ void mdbx_rthc_dtor(void)
 	 *    TSD-деструкторах и поэтому может выгрузить lib.so до того как
 	 *    отработали все деструкторы.
 	 *  - Исходное проявление проблемы было зафиксировано
-	 *    в https://github.com/ReOpen/ReOpenLDAP/issues/48
+	 *    в https://github.com/leo-yuriev/ReOpenLDAP/issues/48
 	 *
 	 * Предыдущее решение посредством выделяемого динамически MDB_rthc
 	 * было не удачным, так как порождало либо утечку памяти,
@@ -6680,6 +6677,11 @@ fetchm:
 			rc = MDB_INCOMPATIBLE;
 			break;
 		}
+		if (mc->mc_ki[mc->mc_top] >= NUMKEYS(mc->mc_pg[mc->mc_top])) {
+			mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]);
+			rc = MDB_NOTFOUND;
+			break;
+		}
 		{
 			MDB_node *leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]);
 			if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) {
@@ -7030,7 +7032,8 @@ more:
 						offset *= 4; /* space for 4 more */
 						break;
 					}
-					/* FALLTHRU: Big enough MDB_DUPFIXED sub-page */
+					/* Big enough MDB_DUPFIXED sub-page */
+					/* fallthrough */
 				case MDB_CURRENT | MDB_NODUPDATA:
 				case MDB_CURRENT:
 					fp->mp_flags |= P_DIRTY;
@@ -7081,8 +7084,9 @@ prep_subDB:
 				} else {
 					memcpy((char *)mp + mp->mp_upper + PAGEBASE, (char *)fp + fp->mp_upper + PAGEBASE,
 						olddata.mv_size - fp->mp_upper - PAGEBASE);
+					memcpy((char *)(&mp->mp_ptrs), (char *)(&fp->mp_ptrs), NUMKEYS(fp) * sizeof(mp->mp_ptrs[0]));
 					for (i=0; i<NUMKEYS(fp); i++)
-						mp->mp_ptrs[i] = fp->mp_ptrs[i] + offset;
+						mp->mp_ptrs[i] += offset;
 				}
 			}
 
@@ -7210,8 +7214,7 @@ new_sub:
 				if (m3->mc_ki[i] >= mc->mc_ki[i] && insert_key) {
 					m3->mc_ki[i]++;
 				}
-				if (XCURSOR_INITED(m3))
-					XCURSOR_REFRESH(m3, mp, m3->mc_ki[i]);
+				XCURSOR_REFRESH(m3, i, mp);
 			}
 		}
 	}
@@ -7252,7 +7255,6 @@ put_sub:
 				MDB_xcursor *mx = mc->mc_xcursor;
 				unsigned i = mc->mc_top;
 				MDB_page *mp = mc->mc_pg[i];
-				int nkeys = NUMKEYS(mp);
 
 				for (m2 = mc->mc_txn->mt_cursors[mc->mc_dbi]; m2; m2=m2->mc_next) {
 					if (m2 == mc || m2->mc_snum < mc->mc_snum) continue;
@@ -7260,8 +7262,8 @@ put_sub:
 					if (m2->mc_pg[i] == mp) {
 						if (m2->mc_ki[i] == mc->mc_ki[i]) {
 							mdb_xcursor_init2(m2, mx, dupdata_flag);
-						} else if (!insert_key && m2->mc_ki[i] < nkeys) {
-							XCURSOR_REFRESH(m2, mp, m2->mc_ki[i]);
+						} else if (!insert_key) {
+							XCURSOR_REFRESH(m2, i, mp);
 						}
 					}
 				}
@@ -7371,12 +7373,7 @@ mdb_cursor_del(MDB_cursor *mc, unsigned flags)
 						if (m2 == mc || m2->mc_snum < mc->mc_snum) continue;
 						if (!(m2->mc_flags & C_INITIALIZED)) continue;
 						if (m2->mc_pg[mc->mc_top] == mp) {
-							MDB_node *n2 = leaf;
-							if (m2->mc_ki[mc->mc_top] != mc->mc_ki[mc->mc_top]) {
-								n2 = NODEPTR(mp, m2->mc_ki[mc->mc_top]);
-								if (n2->mn_flags & F_SUBDATA) continue;
-							}
-							m2->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(n2);
+							XCURSOR_REFRESH(m2, mc->mc_top, mp);
 						}
 					}
 				}
@@ -8280,8 +8277,8 @@ mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst, int fromleft)
 					m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top];
 					m3->mc_ki[csrc->mc_top-1]++;
 				}
-				if (XCURSOR_INITED(m3) && IS_LEAF(mps))
-					XCURSOR_REFRESH(m3, m3->mc_pg[csrc->mc_top], m3->mc_ki[csrc->mc_top]);
+				if (IS_LEAF(mps))
+					XCURSOR_REFRESH(m3, csrc->mc_top, m3->mc_pg[csrc->mc_top]);
 			}
 		} else
 		/* Adding on the right, bump others down */
@@ -8302,8 +8299,8 @@ mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst, int fromleft)
 					} else {
 						m3->mc_ki[csrc->mc_top]--;
 					}
-					if (XCURSOR_INITED(m3) && IS_LEAF(mps))
-						XCURSOR_REFRESH(m3, m3->mc_pg[csrc->mc_top], m3->mc_ki[csrc->mc_top]);
+					if (IS_LEAF(mps))
+						XCURSOR_REFRESH(m3, csrc->mc_top, m3->mc_pg[csrc->mc_top]);
 				}
 			}
 		}
@@ -8501,8 +8498,8 @@ mdb_page_merge(MDB_cursor *csrc, MDB_cursor *cdst)
 				m3->mc_ki[top-1] > csrc->mc_ki[top-1]) {
 				m3->mc_ki[top-1]--;
 			}
-			if (XCURSOR_INITED(m3) && IS_LEAF(psrc))
-				XCURSOR_REFRESH(m3, m3->mc_pg[top], m3->mc_ki[top]);
+			if (IS_LEAF(psrc))
+				XCURSOR_REFRESH(m3, top, m3->mc_pg[top]);
 		}
 	}
 	{
@@ -8764,8 +8761,7 @@ mdb_cursor_del0(MDB_cursor *mc)
 				} else if (m3->mc_ki[mc->mc_top] > ki) {
 					m3->mc_ki[mc->mc_top]--;
 				}
-				if (XCURSOR_INITED(m3))
-					XCURSOR_REFRESH(m3, m3->mc_pg[mc->mc_top], m3->mc_ki[mc->mc_top]);
+				XCURSOR_REFRESH(m3, mc->mc_top, mp);
 			}
 		}
 	}
@@ -8812,8 +8808,10 @@ mdb_cursor_del0(MDB_cursor *mc)
 							if (m3->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) {
 								if (!(node->mn_flags & F_SUBDATA))
 									m3->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(node);
-							} else
+							} else {
 								mdb_xcursor_init1(m3, node);
+								m3->mc_xcursor->mx_cursor.mc_flags |= C_DEL;
+							}
 						}
 					}
 				}
@@ -9304,8 +9302,8 @@ mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, pgno_t newpgno
 				m3->mc_ki[ptop] >= mc->mc_ki[ptop]) {
 				m3->mc_ki[ptop]++;
 			}
-			if (XCURSOR_INITED(m3) && IS_LEAF(mp))
-				XCURSOR_REFRESH(m3, m3->mc_pg[mc->mc_top], m3->mc_ki[mc->mc_top]);
+			if (IS_LEAF(mp))
+				XCURSOR_REFRESH(m3, mc->mc_top, m3->mc_pg[mc->mc_top]);
 		}
 	}
 	mdb_debug("mp left: %d, rp left: %d", SIZELEFT(mp), SIZELEFT(rp));
@@ -10210,8 +10208,11 @@ int mdb_dbi_open(MDB_txn *txn, const char *name, unsigned flags, MDB_dbi *dbi)
 		MDB_node *node = NODEPTR(mc.mc_pg[mc.mc_top], mc.mc_ki[mc.mc_top]);
 		if (unlikely((node->mn_flags & (F_DUPDATA|F_SUBDATA)) != F_SUBDATA))
 			return MDB_INCOMPATIBLE;
-	} else if (! (rc == MDB_NOTFOUND && (flags & MDB_CREATE))) {
-		return rc;
+	} else {
+		if (rc != MDB_NOTFOUND || !(flags & MDB_CREATE))
+			return rc;
+		if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY))
+			return EACCES;
 	}
 
 	/* Done here so we cannot fail after creating a new DB */

mdb_dump.c

@@ -186,7 +186,7 @@ int main(int argc, char *argv[])
 			break;
 		case 'l':
 			list = 1;
-			/*FALLTHROUGH*/;
+			/* fallthrough */
 		case 'a':
 			if (subname)
 				usage(prog);

mdbx.c

@@ -864,13 +864,14 @@ int mdbx_set_attr(MDB_txn *txn, MDB_dbi dbi,
 		return rc;
 	}
 
+	old_attr = 0;
 	rc = mdbx_attr_peek(&old_data, &old_attr);
 	if (unlikely(rc != MDB_SUCCESS))
 		return rc;
 
 	if (old_attr == attr && (!data ||
 			(data->mv_size == old_data.mv_size
-				&& memcpy(data->mv_data, old_data.mv_data, old_data.mv_size) == 0)))
+				&& memcmp(data->mv_data, old_data.mv_data, old_data.mv_size) == 0)))
 		return MDB_SUCCESS;
 
 	mc.mc_next = txn->mt_cursors[dbi];

mtest0.c

@@ -21,6 +21,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #include <pthread.h>

mtest1.c

@@ -20,6 +20,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)

mtest2.c

@@ -23,6 +23,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)

mtest3.c

@@ -23,6 +23,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)

mtest4.c

@@ -23,6 +23,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)

mtest5.c

@@ -23,6 +23,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)

mtest6.c

@@ -23,6 +23,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)

mtest7.c

@@ -18,6 +18,7 @@
 #include <string.h>
 #include <time.h>
 #include <sys/time.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)

mtest8.c

@@ -18,6 +18,7 @@
 #include <string.h>
 #include <time.h>
 #include <sys/time.h>
+#include <sys/sysmacros.h>
 #include "mdbx.h"
 
 #define E(expr) CHECK((rc = (expr)) == MDB_SUCCESS, #expr)