mdbx-doc: import - mdb_cursor_del does not invalidate the cursor (ITS#8857).

Unnecessary since these are DUPs, the key will always be the same

.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

README.md

@@ -5,8 +5,16 @@ Extended LMDB, aka "Расширенная LMDB".
 *The Future will Positive. Всё будет хорошо.*
 [![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/leo-yuriev/libmdbx/tree/stable%2F0.0).
+## 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
 

circle.yml

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

lmdb.h

@@ -390,7 +390,7 @@ typedef enum MDB_cursor_op {
 	MDB_GET_BOTH,			/**< Position at key/data pair. Only for #MDB_DUPSORT */
 	MDB_GET_BOTH_RANGE,		/**< position at key, nearest data. Only for #MDB_DUPSORT */
 	MDB_GET_CURRENT,		/**< Return key/data at current cursor position */
-	MDB_GET_MULTIPLE,		/**< Return key and up to a page of duplicate data items
+	MDB_GET_MULTIPLE,		/**< Return up to a page of duplicate data items
 								from current cursor position. Move cursor to prepare
 								for #MDB_NEXT_MULTIPLE. Only for #MDB_DUPFIXED */
 	MDB_LAST,				/**< Position at last key/data item */
@@ -399,7 +399,7 @@ typedef enum MDB_cursor_op {
 	MDB_NEXT,				/**< Position at next data item */
 	MDB_NEXT_DUP,			/**< Position at next data item of current key.
 								Only for #MDB_DUPSORT */
-	MDB_NEXT_MULTIPLE,		/**< Return key and up to a page of duplicate data items
+	MDB_NEXT_MULTIPLE,		/**< Return up to a page of duplicate data items
 								from next cursor position. Move cursor to prepare
 								for #MDB_NEXT_MULTIPLE. Only for #MDB_DUPFIXED */
 	MDB_NEXT_NODUP,			/**< Position at first data item of next key */
@@ -410,7 +410,7 @@ typedef enum MDB_cursor_op {
 	MDB_SET,				/**< Position at specified key */
 	MDB_SET_KEY,			/**< Position at specified key, return key + data */
 	MDB_SET_RANGE,			/**< Position at first key greater than or equal to specified key. */
-	MDB_PREV_MULTIPLE		/**< Position at previous page and return key and up to
+	MDB_PREV_MULTIPLE		/**< Position at previous page and return up to
 								a page of duplicate data items. Only for #MDB_DUPFIXED */
 } MDB_cursor_op;
 
@@ -1589,6 +1589,10 @@ int  mdb_cursor_put(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
 	/** @brief Delete current key/data pair
 	 *
 	 * This function deletes the key/data pair to which the cursor refers.
+	 * This does not invalidate the cursor, so operations such as MDB_NEXT
+	 * can still be used on it.
+	 * Both MDB_NEXT and MDB_GET_CURRENT will return the same record after
+	 * this operation.
 	 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
 	 * @param[in] flags Options for this operation. This parameter
 	 * must be set to 0 or one of the values described here.

mdb.c

@@ -602,6 +602,8 @@ typedef struct MDB_page {
 	indx_t		mp_ptrs[1];		/**< dynamic size */
 } MDB_page;
 
+#define PAGETYPE(p)	 ((p)->mp_flags & (P_BRANCH | P_LEAF | P_LEAF2 | P_OVERFLOW))
+
 	/** Size of the page header, excluding dynamic data at the end */
 #define PAGEHDRSZ	 ((unsigned) offsetof(MDB_page, mp_ptrs))
 
@@ -1304,9 +1306,6 @@ static void mdb_debug_log(int type, const char *function, int line, const char *
 #	define mdb_assert_enabled() \
 		unlikely(mdb_runtime_flags & MDBX_DBG_ASSERT)
 
-#	define mdb_audit_enabled() \
-		unlikely(mdb_runtime_flags & MDBX_DBG_AUDIT)
-
 #	define mdb_debug_enabled(type) \
 		unlikely(mdb_runtime_flags & \
 			(type & (MDBX_DBG_TRACE | MDBX_DBG_EXTRA)))
@@ -1317,7 +1316,6 @@ static void mdb_debug_log(int type, const char *function, int line, const char *
 #	else
 #		define mdb_debug_enabled(type) (0)
 #	endif
-#	define mdb_audit_enabled() (0)
 #	define mdb_assert_enabled() (0)
 #	define mdb_assert_fail(env, msg, func, line) \
 		__assert_fail(msg, __FILE__, line, func)
@@ -1427,161 +1425,6 @@ mdb_dkey(MDB_val *key, char *buf)
 	return buf;
 }
 
-#if 0 /* LY: debug stuff */
-static const char *
-mdb_leafnode_type(MDB_node *n)
-{
-	static char *const tp[2][2] = {{"", ": DB"}, {": sub-page", ": sub-DB"}};
-	return F_ISSET(n->mn_flags, F_BIGDATA) ? ": overflow page" :
-		tp[F_ISSET(n->mn_flags, F_DUPDATA)][F_ISSET(n->mn_flags, F_SUBDATA)];
-}
-
-/** Display all the keys in the page. */
-static void
-mdb_page_list(MDB_page *mp)
-{
-	pgno_t pgno = mdb_dbg_pgno(mp);
-	const char *type, *state = (mp->mp_flags & P_DIRTY) ? ", dirty" : "";
-	MDB_node *node;
-	unsigned i, nkeys, nsize, total = 0;
-	MDB_val key;
-	DKBUF;
-
-	switch (mp->mp_flags & (P_BRANCH|P_LEAF|P_LEAF2|P_META|P_OVERFLOW|P_SUBP)) {
-	case P_BRANCH:              type = "Branch page";		break;
-	case P_LEAF:                type = "Leaf page";			break;
-	case P_LEAF|P_SUBP:         type = "Sub-page";			break;
-	case P_LEAF|P_LEAF2:        type = "LEAF2 page";		break;
-	case P_LEAF|P_LEAF2|P_SUBP: type = "LEAF2 sub-page";	break;
-	case P_OVERFLOW:
-		mdb_print("Overflow page %zu pages %u%s\n",
-			pgno, mp->mp_pages, state);
-		return;
-	case P_META:
-		mdb_print("Meta-page %zu txnid %zu\n",
-			pgno, ((MDB_meta *)PAGEDATA(mp))->mm_txnid);
-		return;
-	default:
-		mdb_print("Bad page %zu flags 0x%X\n", pgno, mp->mp_flags);
-		return;
-	}
-
-	nkeys = NUMKEYS(mp);
-	mdb_print("%s %zu numkeys %u%s\n", type, pgno, nkeys, state);
-
-	for (i=0; i<nkeys; i++) {
-		if (IS_LEAF2(mp)) {	/* LEAF2 pages have no mp_ptrs[] or node headers */
-			key.mv_size = nsize = mp->mp_leaf2_ksize;
-			key.mv_data = LEAF2KEY(mp, i, nsize);
-			total += nsize;
-			mdb_print("key %u: nsize %u, %s\n", i, nsize, DKEY(&key));
-			continue;
-		}
-		node = NODEPTR(mp, i);
-		key.mv_size = node->mn_ksize;
-		key.mv_data = node->mn_data;
-		nsize = NODESIZE + key.mv_size;
-		if (IS_BRANCH(mp)) {
-			mdb_print("key %u: page %zu, %s\n", i, NODEPGNO(node), DKEY(&key));
-			total += nsize;
-		} else {
-			if (F_ISSET(node->mn_flags, F_BIGDATA))
-				nsize += sizeof(pgno_t);
-			else
-				nsize += NODEDSZ(node);
-			total += nsize;
-			nsize += sizeof(indx_t);
-			mdb_print("key %u: nsize %u, %s%s\n",
-				i, nsize, DKEY(&key), mdb_leafnode_type(node));
-		}
-		total = EVEN(total);
-	}
-	mdb_print("Total: header %u + contents %u + unused %u\n",
-		IS_LEAF2(mp) ? PAGEHDRSZ : PAGEBASE + mp->mp_lower, total, SIZELEFT(mp));
-}
-
-static void
-mdb_cursor_chk(MDB_cursor *mc)
-{
-	unsigned i;
-	MDB_node *node;
-	MDB_page *mp;
-
-	if (!mc->mc_snum || !(mc->mc_flags & C_INITIALIZED)) return;
-	for (i=0; i<mc->mc_top; i++) {
-		mp = mc->mc_pg[i];
-		node = NODEPTR(mp, mc->mc_ki[i]);
-		if (unlikely(NODEPGNO(node) != mc->mc_pg[i+1]->mp_pgno))
-			mdb_print("oops!\n");
-	}
-	if (unlikely(mc->mc_ki[i] >= NUMKEYS(mc->mc_pg[i])))
-		mdb_print("ack!\n");
-	if (XCURSOR_INITED(mc)) {
-		node = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]);
-		if (((node->mn_flags & (F_DUPDATA|F_SUBDATA)) == F_DUPDATA) &&
-			mc->mc_xcursor->mx_cursor.mc_pg[0] != NODEDATA(node)) {
-			mdb_print("blah!\n");
-		}
-	}
-}
-#endif /* 0 */
-
-/** Count all the pages in each DB and in the freelist
- *  and make sure it matches the actual number of pages
- *  being used.
- *  All named DBs must be open for a correct count.
- */
-static void mdb_audit(MDB_txn *txn)
-{
-	MDB_cursor mc;
-	MDB_val key, data;
-	MDB_ID freecount, count;
-	MDB_dbi i;
-	int rc;
-
-	freecount = 0;
-	mdb_cursor_init(&mc, txn, FREE_DBI, NULL);
-	while ((rc = mdb_cursor_get(&mc, &key, &data, MDB_NEXT)) == 0)
-		freecount += *(MDB_ID *)data.mv_data;
-	mdb_tassert(txn, rc == MDB_NOTFOUND);
-
-	count = 0;
-	for (i = 0; i<txn->mt_numdbs; i++) {
-		MDB_xcursor mx;
-		if (!(txn->mt_dbflags[i] & DB_VALID))
-			continue;
-		mdb_cursor_init(&mc, txn, i, &mx);
-		if (txn->mt_dbs[i].md_root == P_INVALID)
-			continue;
-		count += txn->mt_dbs[i].md_branch_pages +
-			txn->mt_dbs[i].md_leaf_pages +
-			txn->mt_dbs[i].md_overflow_pages;
-		if (txn->mt_dbs[i].md_flags & MDB_DUPSORT) {
-			rc = mdb_page_search(&mc, NULL, MDB_PS_FIRST);
-			for (; rc == MDB_SUCCESS; rc = mdb_cursor_sibling(&mc, 1)) {
-				unsigned j;
-				MDB_page *mp;
-				mp = mc.mc_pg[mc.mc_top];
-				for (j=0; j<NUMKEYS(mp); j++) {
-					MDB_node *leaf = NODEPTR(mp, j);
-					if (leaf->mn_flags & F_SUBDATA) {
-						MDB_db db;
-						memcpy(&db, NODEDATA(leaf), sizeof(db));
-						count += db.md_branch_pages + db.md_leaf_pages +
-							db.md_overflow_pages;
-					}
-				}
-			}
-			mdb_tassert(txn, rc == MDB_NOTFOUND);
-		}
-	}
-	if (freecount + count + NUM_METAS != txn->mt_next_pgno) {
-		mdb_print("audit: %lu freecount: %lu count: %lu total: %lu next_pgno: %lu\n",
-			txn->mt_txnid, freecount, count+NUM_METAS,
-			freecount+count+NUM_METAS, txn->mt_next_pgno);
-	}
-}
-
 int
 mdb_cmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b)
 {
@@ -3364,6 +3207,8 @@ mdb_freelist_save(MDB_txn *txn)
 	const int lifo = (env->me_flags & MDBX_LIFORECLAIM) != 0;
 
 	mdb_cursor_init(&mc, txn, FREE_DBI, NULL);
+	mc.mc_next = txn->mt_cursors[FREE_DBI];
+	txn->mt_cursors[FREE_DBI] = &mc;
 
 	/* MDB_RESERVE cancels meminit in ovpage malloc (when no WRITEMAP) */
 	clean_limit = (env->me_flags & (MDB_NOMEMINIT|MDB_WRITEMAP))
@@ -3645,6 +3490,7 @@ bailout:
 			txn->mt_lifo_reclaimed = NULL;
 		}
 	}
+	txn->mt_cursors[FREE_DBI] = mc.mc_next;
 
 	return rc;
 }
@@ -3990,9 +3836,6 @@ mdb_txn_commit(MDB_txn *txn)
 	env->me_pghead = NULL;
 	mdb_midl_shrink(&txn->mt_free_pgs);
 
-	if (mdb_audit_enabled())
-		mdb_audit(txn);
-
 	rc = mdb_page_flush(txn, 0);
 	if (likely(rc == MDB_SUCCESS)) {
 		MDB_meta meta;
@@ -7032,7 +6875,8 @@ 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;
@@ -8165,6 +8009,12 @@ mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst, int fromleft)
 
 	DKBUF;
 
+	mdb_tassert(csrc->mc_txn, PAGETYPE(csrc->mc_pg[csrc->mc_top]) == PAGETYPE(cdst->mc_pg[cdst->mc_top]));
+	if (unlikely(PAGETYPE(csrc->mc_pg[csrc->mc_top]) != PAGETYPE(cdst->mc_pg[cdst->mc_top]))) {
+		cdst->mc_txn->mt_flags |= MDB_TXN_ERROR;
+		return MDB_PROBLEM;
+	}
+
 	/* Mark src and dst as dirty. */
 	if (unlikely((rc = mdb_page_touch(csrc)) ||
 		(rc = mdb_page_touch(cdst))))
@@ -8395,6 +8245,12 @@ mdb_page_merge(MDB_cursor *csrc, MDB_cursor *cdst)
 	mdb_cassert(csrc, csrc->mc_snum > 1);	/* can't merge root page */
 	mdb_cassert(csrc, cdst->mc_snum > 1);
 
+	mdb_tassert(csrc->mc_txn, PAGETYPE(psrc) == PAGETYPE(pdst));
+	if (unlikely(PAGETYPE(psrc) != PAGETYPE(pdst))) {
+		cdst->mc_txn->mt_flags |= MDB_TXN_ERROR;
+		return MDB_PROBLEM;
+	}
+
 	/* Mark dst as dirty. */
 	if (unlikely(rc = mdb_page_touch(cdst)))
 		return rc;
@@ -8598,7 +8454,7 @@ mdb_rebalance(MDB_cursor *mc)
 						m3 = &m2->mc_xcursor->mx_cursor;
 					else
 						m3 = m2;
-					if (!(m3->mc_flags & C_INITIALIZED) || (m3->mc_snum < mc->mc_snum))
+					if (m3 == mc || !(m3->mc_flags & C_INITIALIZED))
 						continue;
 					if (m3->mc_pg[0] == mp) {
 						m3->mc_snum = 0;

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

@@ -530,7 +530,7 @@ int mdbx_replace(MDB_txn *txn, MDB_dbi dbi,
 		rc = mdbx_cursor_get(&mc, &present_key, &present_data, MDB_SET_KEY);
 		if (unlikely(rc != MDB_SUCCESS)) {
 			old_data->iov_base = NULL;
-			old_data->iov_len = rc;
+			old_data->iov_len = 0;
 			if (rc != MDB_NOTFOUND || (flags & MDB_CURRENT))
 				goto bailout;
 		} else if (flags & MDB_NOOVERWRITE) {

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)