[PATCH 1/4 V6 nft] Simplify parser rule_spec tree

[PATCH 1/4 V6 nft] Simplify parser rule_spec tree

[Carlos Falgueras García]

This patch separates the rule identification from the rule localization, so
the logic moves from the evaluator to the parser. This allows to revert the
patch "evaluate: improve rule managment checks"
(4176c7d30c2ff1b3f52468fc9c08b8df83f979a8) and saves a lot of code.

Signed-off-by: Carlos Falgueras García <carlosfg@riseup.net>
---
 src/evaluate.c     | 68 +-----------------------------------------------------
 src/parser_bison.y | 45 +++++++++++++++++-------------------
 2 files changed, 22 insertions(+), 91 deletions(-)

diff --git a/src/evaluate.c b/src/evaluate.c
index 87f5a6d..2f94ac6 100644
--- a/src/evaluate.c
+++ b/src/evaluate.c
@@ -44,12 +44,6 @@ static const char *byteorder_names[] = {
 	__stmt_binary_error(ctx, &(s1)->location, NULL, fmt, ## args)
 #define cmd_error(ctx, fmt, args...) \
 	__stmt_binary_error(ctx, &(ctx->cmd)->location, NULL, fmt, ## args)
-#define handle_error(ctx, fmt, args...) \
-	__stmt_binary_error(ctx, &ctx->cmd->handle.handle.location, NULL, fmt, ## args)
-#define position_error(ctx, fmt, args...) \
-	__stmt_binary_error(ctx, &ctx->cmd->handle.position.location, NULL, fmt, ## args)
-#define handle_position_error(ctx, fmt, args...) \
-	__stmt_binary_error(ctx, &ctx->cmd->handle.handle.location, &ctx->cmd->handle.position.location, fmt, ## args)
 
 static int __fmtstring(3, 4) set_error(struct eval_ctx *ctx,
 				       const struct set *set,
@@ -2481,68 +2475,11 @@ static int set_evaluate(struct eval_ctx *ctx, struct set *set)
 	return 0;
 }
 
-static int rule_evaluate_cmd(struct eval_ctx *ctx)
-{
-	struct handle *handle = &ctx->cmd->handle;
-
-	/* allowed:
-	 * - insert [position] (no handle)
-	 * - add [position] (no handle)
-	 * - replace <handle> (no position)
-	 * - delete <handle> (no position)
-	 */
-
-	switch (ctx->cmd->op) {
-	case CMD_INSERT:
-		if (handle->handle.id && handle->position.id)
-			return handle_position_error(ctx, "use only `position'"
-						     " instead");
-
-		if (handle->handle.id)
-			return handle_error(ctx, "use `position' instead");
-		break;
-	case CMD_ADD:
-		if (handle->handle.id && handle->position.id)
-			return handle_position_error(ctx, "use only `position'"
-						     " instead");
-
-		if (handle->handle.id)
-			return handle_error(ctx, "use `position' instead");
-
-		break;
-	case CMD_REPLACE:
-		if (handle->handle.id && handle->position.id)
-			return handle_position_error(ctx, "use only `handle' "
-						     "instead");
-		if (handle->position.id)
-			return position_error(ctx, "use `handle' instead");
-		if (!handle->handle.id)
-			return cmd_error(ctx, "missing `handle'");
-		break;
-	case CMD_DELETE:
-		if (handle->handle.id && handle->position.id)
-			return handle_position_error(ctx, "use only `handle' "
-						     "instead");
-		if (handle->position.id)
-			return position_error(ctx, "use `handle' instead");
-		if (!handle->handle.id)
-			return cmd_error(ctx, "missing `handle'");
-		break;
-	default:
-		BUG("unkown command type %u\n", ctx->cmd->op);
-	}
-
-	return 0;
-}
-
 static int rule_evaluate(struct eval_ctx *ctx, struct rule *rule)
 {
 	struct stmt *stmt, *tstmt = NULL;
 	struct error_record *erec;
 
-	if (rule_evaluate_cmd(ctx) < 0)
-		return -1;
-
 	proto_ctx_init(&ctx->pctx, rule->handle.family);
 	memset(&ctx->ectx, 0, sizeof(ctx->ectx));
 
@@ -2723,11 +2660,8 @@ static int cmd_evaluate_delete(struct eval_ctx *ctx, struct cmd *cmd)
 			return ret;
 
 		return setelem_evaluate(ctx, &cmd->expr);
-	case CMD_OBJ_RULE:
-		if (rule_evaluate_cmd(ctx) < 0)
-			return -1;
-		/* fall through */
 	case CMD_OBJ_SET:
+	case CMD_OBJ_RULE:
 	case CMD_OBJ_CHAIN:
 	case CMD_OBJ_TABLE:
 		return 0;
diff --git a/src/parser_bison.y b/src/parser_bison.y
index e16b8a3..2ca7eea 100644
--- a/src/parser_bison.y
+++ b/src/parser_bison.y
@@ -425,15 +425,12 @@ static void location_update(struct location *loc, struct location *rhs, int n)
 %type <cmd>			base_cmd add_cmd replace_cmd create_cmd insert_cmd delete_cmd list_cmd flush_cmd rename_cmd export_cmd monitor_cmd describe_cmd
 %destructor { cmd_free($$); }	base_cmd add_cmd replace_cmd create_cmd insert_cmd delete_cmd list_cmd flush_cmd rename_cmd export_cmd monitor_cmd describe_cmd
 
-%type <handle>			table_spec chain_spec chain_identifier ruleid_spec ruleset_spec
-%destructor { handle_free(&$$); } table_spec chain_spec chain_identifier ruleid_spec ruleset_spec
+%type <handle>			table_spec chain_spec chain_identifier ruleid_spec handle_spec position_spec rule_position ruleset_spec
+%destructor { handle_free(&$$); } table_spec chain_spec chain_identifier ruleid_spec handle_spec position_spec rule_position ruleset_spec
 %type <handle>			set_spec set_identifier
 %destructor { handle_free(&$$); } set_spec set_identifier
 %type <val>			family_spec family_spec_explicit chain_policy prio_spec
 
-%type <handle_spec>		handle_spec
-%type <position_spec>		position_spec
-
 %type <string>			dev_spec
 %destructor { xfree($$); }	dev_spec
 
@@ -720,11 +717,11 @@ add_cmd			:	TABLE		table_spec
 				close_scope(state);
 				$$ = cmd_alloc(CMD_ADD, CMD_OBJ_CHAIN, &$2, &@$, $5);
 			}
-			|	RULE		ruleid_spec	rule
+			|	RULE		rule_position	rule
 			{
 				$$ = cmd_alloc(CMD_ADD, CMD_OBJ_RULE, &$2, &@$, $3);
 			}
-			|	/* empty */	ruleid_spec	rule
+			|	/* empty */	rule_position	rule
 			{
 				$$ = cmd_alloc(CMD_ADD, CMD_OBJ_RULE, &$1, &@$, $2);
 			}
@@ -779,7 +776,7 @@ create_cmd		:	TABLE		table_spec
 			}
 			;
 
-insert_cmd		:	RULE		ruleid_spec	rule
+insert_cmd		:	RULE		rule_position	rule
 			{
 				$$ = cmd_alloc(CMD_INSERT, CMD_OBJ_RULE, &$2, &@$, $3);
 			}
@@ -1252,35 +1249,35 @@ set_identifier		:	identifier
 			}
 			;
 
-handle_spec		:	/* empty */
+handle_spec		:	HANDLE		NUM
 			{
-				memset(&$$, 0, sizeof($$));
+				$$.handle.location	= @$;
+				$$.handle.id		= $2;
 			}
-			|	HANDLE		NUM
+			;
+
+position_spec		:	POSITION	NUM
 			{
-				memset(&$$, 0, sizeof($$));
-				$$.location	= @$;
-				$$.id		= $2;
+				$$.position.location	= @$;
+				$$.position.id		= $2;
 			}
 			;
 
-position_spec		:	/* empty */
+rule_position		:	chain_spec
 			{
-				memset(&$$, 0, sizeof($$));
+				$$ = $1;
 			}
-			|	POSITION	NUM
+			|	chain_spec position_spec
 			{
-				memset(&$$, 0, sizeof($$));
-				$$.location	= @$;
-				$$.id		= $2;
+				$$ = $1;
+				handle_merge(&$$, &$2);
 			}
 			;
 
-ruleid_spec		:	chain_spec	handle_spec	position_spec
+ruleid_spec		:	chain_spec handle_spec
 			{
-				$$		= $1;
-				$$.handle	= $2;
-				$$.position	= $3;
+				$$ = $1;
+				handle_merge(&$$, &$2);
 			}
 			;
 
-- 
2.8.3

[PATCH 2/4 V6 nft] Implement deleting rule by description

[Carlos Falgueras García]

This patch introduces deletion in a similar fashion as in iptables, thus,
we can delete the first rule that matches our description, for example:

	$ nft list -a ruleset
	table ip t {
		chain c {
			ip saddr 1.1.1.1 counter packets 0 bytes 0 # handle 1
			ip saddr 1.1.1.2 counter packets 0 bytes 0 # handle 2
			ip saddr 1.1.1.2 counter packets 0 bytes 0 # handle 3
			ip saddr 1.1.1.4 counter packets 0 bytes 0 # handle 4
		}
	}
	$ nft delete rule table chain ip saddr 1.1.1.2 counter
	$ nft list -a ruleset
	table ip t {
		chain c {
			ip saddr 1.1.1.1 counter packets 0 bytes 0 # handle 1
			ip saddr 1.1.1.2 counter packets 0 bytes 0 # handle 3
			ip saddr 1.1.1.4 counter packets 0 bytes 0 # handle 4
		}
	}

The parser rule 'ruleid_spec' is now of the type 'struct rule' in order to
hold a rule description. When rule is identified with its handle a dummy
'struct rule' is allocated to hold the specified handle.

Signed-off-by: Carlos Falgueras García <carlosfg@riseup.net>
---
 src/evaluate.c     |  6 ++++++
 src/parser_bison.y | 24 ++++++++++++++++--------
 src/rule.c         | 50 ++++++++++++++++++++++++++++++++++++++++++++++++--
 3 files changed, 70 insertions(+), 10 deletions(-)

diff --git a/src/evaluate.c b/src/evaluate.c
index 2f94ac6..f7b349b 100644
--- a/src/evaluate.c
+++ b/src/evaluate.c
@@ -2661,7 +2661,13 @@ static int cmd_evaluate_delete(struct eval_ctx *ctx, struct cmd *cmd)
 
 		return setelem_evaluate(ctx, &cmd->expr);
 	case CMD_OBJ_SET:
+		return 0;
 	case CMD_OBJ_RULE:
+		/* CMD_LIST force caching all ruleset */
+		ret = cache_update(CMD_LIST, ctx->msgs);
+		if (ret < 0)
+			return ret;
+		return rule_evaluate(ctx, cmd->rule);
 	case CMD_OBJ_CHAIN:
 	case CMD_OBJ_TABLE:
 		return 0;
diff --git a/src/parser_bison.y b/src/parser_bison.y
index 2ca7eea..beea38b 100644
--- a/src/parser_bison.y
+++ b/src/parser_bison.y
@@ -425,8 +425,8 @@ static void location_update(struct location *loc, struct location *rhs, int n)
 %type <cmd>			base_cmd add_cmd replace_cmd create_cmd insert_cmd delete_cmd list_cmd flush_cmd rename_cmd export_cmd monitor_cmd describe_cmd
 %destructor { cmd_free($$); }	base_cmd add_cmd replace_cmd create_cmd insert_cmd delete_cmd list_cmd flush_cmd rename_cmd export_cmd monitor_cmd describe_cmd
 
-%type <handle>			table_spec chain_spec chain_identifier ruleid_spec handle_spec position_spec rule_position ruleset_spec
-%destructor { handle_free(&$$); } table_spec chain_spec chain_identifier ruleid_spec handle_spec position_spec rule_position ruleset_spec
+%type <handle>			table_spec chain_spec chain_identifier handle_spec position_spec rule_position ruleset_spec
+%destructor { handle_free(&$$); } table_spec chain_spec chain_identifier handle_spec position_spec rule_position ruleset_spec
 %type <handle>			set_spec set_identifier
 %destructor { handle_free(&$$); } set_spec set_identifier
 %type <val>			family_spec family_spec_explicit chain_policy prio_spec
@@ -438,7 +438,7 @@ static void location_update(struct location *loc, struct location *rhs, int n)
 %destructor { close_scope(state); table_free($$); }	table_block_alloc
 %type <chain>			chain_block_alloc chain_block
 %destructor { close_scope(state); chain_free($$); }	chain_block_alloc
-%type <rule>			rule rule_alloc
+%type <rule>			rule ruleid_spec rule_alloc
 %destructor { rule_free($$); }	rule
 
 %type <val>			set_flag_list	set_flag
@@ -745,9 +745,10 @@ add_cmd			:	TABLE		table_spec
 			}
 			;
 
-replace_cmd		:	RULE		ruleid_spec	rule
+replace_cmd		:	RULE		chain_spec	handle_spec	rule
 			{
-				$$ = cmd_alloc(CMD_REPLACE, CMD_OBJ_RULE, &$2, &@$, $3);
+				handle_merge(&$2, &$3);
+				$$ = cmd_alloc(CMD_REPLACE, CMD_OBJ_RULE, &$2, &@$, $4);
 			}
 			;
 
@@ -792,7 +793,7 @@ delete_cmd		:	TABLE		table_spec
 			}
 			|	RULE		ruleid_spec
 			{
-				$$ = cmd_alloc(CMD_DELETE, CMD_OBJ_RULE, &$2, &@$, NULL);
+				$$ = cmd_alloc(CMD_DELETE, CMD_OBJ_RULE, &$2->handle, &@$, $2);
 			}
 			|	SET		set_spec
 			{
@@ -1276,8 +1277,15 @@ rule_position		:	chain_spec
 
 ruleid_spec		:	chain_spec handle_spec
 			{
-				$$ = $1;
-				handle_merge(&$$, &$2);
+				$$ = rule_alloc(&@$, NULL);
+				$$->handle = $1;
+				handle_merge(&$$->handle, &$2);
+			}
+			|
+				chain_spec rule
+			{
+				$$ = $2;
+				handle_merge(&$$->handle, &$1);
 			}
 			;
 
diff --git a/src/rule.c b/src/rule.c
index 14e57f2..e9672a7 100644
--- a/src/rule.c
+++ b/src/rule.c
@@ -20,6 +20,7 @@
 #include <rule.h>
 #include <utils.h>
 #include <netlink.h>
+#include <erec.h>
 
 #include <libnftnl/common.h>
 #include <libnftnl/ruleset.h>
@@ -402,6 +403,32 @@ void rule_print(const struct rule *rule)
 		printf(" # handle %" PRIu64, rule->handle.handle.id);
 }
 
+static struct rule *rule_find_first(const struct rule *rule)
+{
+	struct nftnl_rule *nlr1, *nlr2;
+	struct rule *rule_idx;
+	struct table *table;
+	struct chain *chain;
+
+	table = table_lookup(&rule->handle);
+	if (!table)
+		return NULL;
+	chain = chain_lookup(table, &rule->handle);
+	if (!chain)
+		return NULL;
+
+	nlr1 = alloc_nftnl_rule(&rule->handle);
+	netlink_linearize_rule(NULL, nlr1, rule);
+
+	list_for_each_entry(rule_idx, &chain->rules, list) {
+		nlr2 = alloc_nftnl_rule(&rule_idx->handle);
+		netlink_linearize_rule(NULL, nlr2, rule_idx);
+		if (nftnl_rule_cmp(nlr1, nlr2))
+			return rule_idx;
+	}
+	return NULL;
+}
+
 struct rule *rule_lookup(const struct chain *chain, uint64_t handle)
 {
 	struct rule *rule;
@@ -1010,6 +1037,26 @@ static int do_delete_setelems(struct netlink_ctx *ctx, const struct handle *h,
 	return 0;
 }
 
+static int do_delete_rule(struct netlink_ctx *ctx, const struct cmd *cmd)
+{
+	struct rule *rule;
+
+	/* Delete by handle */
+	if (cmd->handle.handle.id)
+		return netlink_del_rule_batch(ctx, &cmd->handle, &cmd->location);
+
+	/* Delete by description */
+	rule = rule_find_first(cmd->rule);
+	if (!rule) {
+		struct error_record *e = erec_create(EREC_ERROR, &cmd->location,
+			"Could not delete rule to batch: Rule not found");
+		erec_queue(e, ctx->msgs);
+		return -1;
+	}
+	return netlink_del_rule_batch(ctx, &rule->handle,
+				      &rule->handle.position.location);
+}
+
 static int do_command_delete(struct netlink_ctx *ctx, struct cmd *cmd)
 {
 	switch (cmd->obj) {
@@ -1018,8 +1065,7 @@ static int do_command_delete(struct netlink_ctx *ctx, struct cmd *cmd)
 	case CMD_OBJ_CHAIN:
 		return netlink_delete_chain(ctx, &cmd->handle, &cmd->location);
 	case CMD_OBJ_RULE:
-		return netlink_del_rule_batch(ctx, &cmd->handle,
-					      &cmd->location);
+		return do_delete_rule(ctx, cmd);
 	case CMD_OBJ_SET:
 		return netlink_delete_set(ctx, &cmd->handle, &cmd->location);
 	case CMD_OBJ_SETELEM:
-- 
2.8.3

Re: [PATCH 2/4 V6 nft] Implement deleting rule by description

[Pablo Neira Ayuso]

On Sun, Aug 21, 2016 at 11:22:08PM +0200, Carlos Falgueras García wrote:
> This patch introduces deletion in a similar fashion as in iptables, thus,
> we can delete the first rule that matches our description, for example:
> 
> 	$ nft list -a ruleset
> 	table ip t {
> 		chain c {
> 			ip saddr 1.1.1.1 counter packets 0 bytes 0 # handle 1
> 			ip saddr 1.1.1.2 counter packets 0 bytes 0 # handle 2
> 			ip saddr 1.1.1.2 counter packets 0 bytes 0 # handle 3
> 			ip saddr 1.1.1.4 counter packets 0 bytes 0 # handle 4
> 		}
> 	}
> 	$ nft delete rule table chain ip saddr 1.1.1.2 counter
> 	$ nft list -a ruleset
> 	table ip t {
> 		chain c {
> 			ip saddr 1.1.1.1 counter packets 0 bytes 0 # handle 1
> 			ip saddr 1.1.1.2 counter packets 0 bytes 0 # handle 3
> 			ip saddr 1.1.1.4 counter packets 0 bytes 0 # handle 4
> 		}
> 	}
> 
> The parser rule 'ruleid_spec' is now of the type 'struct rule' in order to
> hold a rule description. When rule is identified with its handle a dummy
> 'struct rule' is allocated to hold the specified handle.

Applied, thanks Carlos.

[PATCH 3/4 V6 nft] test: shell: Add tests for deleting rule by description

[Carlos Falgueras García]

They checks if commands like "nft delete rule <table> <chain> <rule desc>"
works as is expected.

First one checks if command deletes only one of the matched rules.
Second one checks if command fails when rule did not found.

Signed-off-by: Carlos Falgueras García <carlosfg@riseup.net>
---
 .../testcases/rule_management/0010delete-by-desc_0 | 39 ++++++++++++++++++++++
 .../testcases/rule_management/0011delete-by-desc_1 | 20 +++++++++++
 2 files changed, 59 insertions(+)
 create mode 100755 tests/shell/testcases/rule_management/0010delete-by-desc_0
 create mode 100755 tests/shell/testcases/rule_management/0011delete-by-desc_1

diff --git a/tests/shell/testcases/rule_management/0010delete-by-desc_0 b/tests/shell/testcases/rule_management/0010delete-by-desc_0
new file mode 100755
index 0000000..6afdec1
--- /dev/null
+++ b/tests/shell/testcases/rule_management/0010delete-by-desc_0
@@ -0,0 +1,39 @@
+#!/bin/bash
+
+# positive tests for rule deletion by description:
+#	$ nft delete rule <table> <chain> <rule description>
+
+RULE2DEL="ip saddr 1.1.1.1 counter"
+
+set -e
+$NFT add table t
+$NFT add chain t c
+$NFT add rule t c ip saddr 1.1.1.1
+$NFT add rule t c $RULE2DEL
+$NFT add rule t c ip saddr 1.1.1.1 accept
+$NFT add rule t c $RULE2DEL
+
+$NFT delete rule t c $RULE2DEL
+if [ $? -ne 0 ]; then
+	echo "E: unable to delete rule \"$RULE2DEL\"" >&2
+	exit 1
+fi
+
+set +e; # Next commands can return 0
+REMAINS_RULE2DEL=$($NFT list -a ruleset | grep -c "$RULE2DEL")
+REMAINS_RULES=$(( $($NFT list -a ruleset | wc -l) - 4 ))
+set -e
+
+if   [ $REMAINS_RULE2DEL -eq 2 ]; then
+	echo "E: First rule \"$RULE2DEL\" should have been deleted" >&2
+	exit 1
+elif [ $REMAINS_RULE2DEL -eq 0 ]; then
+	echo "E: Second rule \"$RULE2DEL\" should not have been deleted" >&2
+	exit 1
+fi
+
+if [ $REMAINS_RULES -ne 3 ]; then
+	echo "E: Rest of rules should not have been deleted" >&2
+	$NFT list -a ruleset
+	exit 1
+fi
diff --git a/tests/shell/testcases/rule_management/0011delete-by-desc_1 b/tests/shell/testcases/rule_management/0011delete-by-desc_1
new file mode 100755
index 0000000..3ddb5ef
--- /dev/null
+++ b/tests/shell/testcases/rule_management/0011delete-by-desc_1
@@ -0,0 +1,20 @@
+#!/bin/bash
+
+# negative tests for rule deletion by description:
+#	$ nft delete rule <table> <chain> <rule description>
+
+set -e
+$NFT add table t
+$NFT add chain t c
+$NFT add rule t c ip saddr 1.1.1.1
+$NFT add rule t c ip saddr 1.1.1.1 accept
+
+set +e; # Next command must fail
+$NFT delete rule t c ip saddr 2.2.2.2
+RET=$?
+if [ $RET -ne 1 ]; then
+	echo "E: Try to delete a nonexistent rule should throw an error" >&2
+	exit $RET
+fi
+
+exit $RET
-- 
2.8.3

Re: [PATCH 3/4 V6 nft] test: shell: Add tests for deleting rule by description

[Pablo Neira Ayuso]

On Sun, Aug 21, 2016 at 11:22:09PM +0200, Carlos Falgueras García wrote:
> They checks if commands like "nft delete rule <table> <chain> <rule desc>"
> works as is expected.
> 
> First one checks if command deletes only one of the matched rules.
> Second one checks if command fails when rule did not found.

Also applied, thanks.

Re: [PATCH 3/4 V6 nft] test: shell: Add tests for deleting rule by description

[Pablo Neira Ayuso]

Hi Carlos,

One of this test fails... so please send me a follow up to fix it.

W: [FAILED]     ./testcases/rule_management/0010delete-by-desc_0

This chunk also looks a bit strange to me.

set +e; # Next commands can return 0
REMAINS_RULE2DEL=$($NFT list -a ruleset | grep -c "$RULE2DEL")
REMAINS_RULES=$(( $($NFT list -a ruleset | wc -l) - 4 ))
set -e

Please, send me follow up patchset to fix this.

Re: [PATCH 3/4 V6 nft] test: shell: Add tests for deleting rule by description

[Pablo Neira Ayuso]

On Mon, Aug 22, 2016 at 06:21:46PM +0200, Pablo Neira Ayuso wrote:
> Hi Carlos,
> 
> One of this test fails... so please send me a follow up to fix it.
> 
> W: [FAILED]     ./testcases/rule_management/0010delete-by-desc_0
> 
> This chunk also looks a bit strange to me.
> 
> set +e; # Next commands can return 0
> REMAINS_RULE2DEL=$($NFT list -a ruleset | grep -c "$RULE2DEL")
> REMAINS_RULES=$(( $($NFT list -a ruleset | wc -l) - 4 ))
> set -e
> 
> Please, send me follow up patchset to fix this.

Also hitting this.

rule.c: In function ‘rule_find_first’:
rule.c:427:3: warning: implicit declaration of function
‘nftnl_rule_cmp’ [-Wimplicit-function-declaration]
   if (nftnl_rule_cmp(nlr1, nlr2))
   ^

Sorry, I'm tossing this patchset, send a v7.

[PATCH 4/4 V6 nft] parser: Improve syntax errors

[Carlos Falgueras García]

Shows a more informative message when user commits a syntax error:

	$ nft add rule t c handle 3 ...
	<cmdline>:1:14-19: Error: Did you mean `position'?
	add rule t c handle 3 ...
	             ^^^^^^
	$ nft delete rule t c position 3 ...
	<cmdline>:1:17-24: Error: Did you mean `handle' or insert a rule description?
	delete rule t c position 3 ...
	                ^^^^^^^^

Adds function 'erec_del_last' that deletes last error from the error queue.
This is needed to do not show two error messages.

Signed-off-by: Carlos Falgueras García <carlosfg@riseup.net>
---
 include/erec.h     |  5 +++++
 src/parser_bison.y | 17 +++++++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/include/erec.h b/include/erec.h
index 36e0efa..95ed24a 100644
--- a/include/erec.h
+++ b/include/erec.h
@@ -58,6 +58,11 @@ static inline void erec_queue(struct error_record *erec,
 	list_add_tail(&erec->list, queue);
 }
 
+static inline void erec_del_last(struct list_head *queue)
+{
+	list_del(queue->prev);
+}
+
 extern void erec_print(FILE *f, const struct error_record *erec);
 extern void erec_print_list(FILE *f, struct list_head *list);
 
diff --git a/src/parser_bison.y b/src/parser_bison.y
index beea38b..9739cb1 100644
--- a/src/parser_bison.y
+++ b/src/parser_bison.y
@@ -1273,6 +1273,14 @@ rule_position		:	chain_spec
 				$$ = $1;
 				handle_merge(&$$, &$2);
 			}
+			|	chain_spec HANDLE error
+			{
+				erec_del_last(state->msgs);
+				erec_queue(error(&@2, "Did you mean `position'?"),
+					   state->msgs);
+				$$ = $1;
+				YYERROR;
+			}
 			;
 
 ruleid_spec		:	chain_spec handle_spec
@@ -1287,6 +1295,15 @@ ruleid_spec		:	chain_spec handle_spec
 				$$ = $2;
 				handle_merge(&$$->handle, &$1);
 			}
+			|	chain_spec POSITION error
+			{
+				erec_del_last(state->msgs);
+				erec_queue(error(&@2, "Did you mean `handle' or insert a rule description?"),
+					   state->msgs);
+				$$ = rule_alloc(&@$, NULL);
+				handle_merge(&$$->handle, &$1);
+				YYERROR;
+			}
 			;
 
 comment_spec		:	COMMENT		string
-- 
2.8.3

Re: [PATCH 1/4 V6 nft] Simplify parser rule_spec tree

[Pablo Neira Ayuso]

On Sun, Aug 21, 2016 at 11:22:07PM +0200, Carlos Falgueras García wrote:
> This patch separates the rule identification from the rule localization, so
> the logic moves from the evaluator to the parser. This allows to revert the
> patch "evaluate: improve rule managment checks"
> (4176c7d30c2ff1b3f52468fc9c08b8df83f979a8) and saves a lot of code.

Applied, thanks.