Server crash when using duplicate segmentby column

The segmentby column info array is populated by using the column
attribute number as an array index. This is done as part of validating
and creating segment by column info in function `compresscolinfo_init`.

Since the column is duplicated the attribute number for both the
segmentby column is same. When this attribute number is used as an
index, only one of the array element is populated correctly with the
detailed column info whareas the other element of the array ramins
NULL. This segmentby column info is updated in catalog as part of
processing compression options (ALTER TABLE ...).

When the chunk is being compressed this segmentby column information is
being retrieved from the catalog to create the scan key in order to
identify any existing index on the table that matches the segmentby
column. Out of the two keys one key gets updated correctly whereas the
second key contains NULL values. This results into a crash during index
scan to identify any existing index on the table.

The proposed change avoid this crash by raising an error if user has
specified duplicated columns as part of compress_segmentby option.

.unreleased/bugfix_6044

@@ -0,0 +1 @@
+Fixes: #6044 Server crash when using duplicate segmentby column

tsl/src/compression/create.c

@@ -245,6 +245,14 @@ compresscolinfo_init(CompressColInfo *cc, Oid srctbl_relid, List *segmentby_cols
 					 errhint("The timescaledb.compress_segmentby option must reference a valid "
 							 "column.")));
 		}
+
+		/* Check if a column is duplicated in segment by column list. */
+		if (segorder_colindex[col_attno - 1] != 0)
+			ereport(ERROR,
+					(errcode(ERRCODE_SYNTAX_ERROR),
+					 errmsg("duplicate column name \"%s\"", NameStr(col->colname)),
+					 errhint("The timescaledb.compress_segmentby option must reference distinct "
+							 "column.")));
 		segorder_colindex[col_attno - 1] = i++;
 	}
 	/* the column indexes are numbered as seg_attnolen + <orderby_index>

tsl/test/expected/compression_errors.out

@@ -195,6 +195,9 @@ HINT:  The option timescaledb.compress_segmentby must be a set of columns separa
 ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_orderby = 'a, p');
 ERROR:  invalid ordering column type point
 DETAIL:  Could not identify a less-than operator for the type.
+ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_segmentby = 'b, b');
+ERROR:  duplicate column name "b"
+HINT:  The timescaledb.compress_segmentby option must reference distinct column.
 --should succeed
 ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_orderby = 'a, b');
 --ddl on ht with compression

tsl/test/sql/compression_errors.sql

@@ -99,6 +99,7 @@ ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_segmentby = 'ran
 ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_segmentby = 'c LIMIT 1');
 ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_segmentby = 'c + b');
 ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_orderby = 'a, p');
+ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_segmentby = 'b, b');
 
 --should succeed
 ALTER TABLE foo set (timescaledb.compress, timescaledb.compress_orderby = 'a, b');