fix(database_postgresql): extract the correct pk names, the type is nameoid not textoid

Author: andinux

src/postgresql/database_postgresql.c

@@ -1671,12 +1671,6 @@ int database_update_schema_hash (cloudsync_context *data, uint64_t *hash) {
 
 // MARK: - PRIMARY KEY -
 
-int database_pk_rowid (cloudsync_context *data, const char *table_name, char ***names, int *count) {
-    // PostgreSQL doesn't have rowid concept like SQLite
-    // Use OID or primary key columns instead
-    return database_pk_names(data, table_name, names, count);
-}
-
 int database_pk_names (cloudsync_context *data, const char *table_name, char ***names, int *count) {
     if (!table_name || !names || !count) return DBRES_MISUSE;
 
@@ -1711,10 +1705,10 @@ int database_pk_names (cloudsync_context *data, const char *table_name, char ***
         bool isnull;
         Datum datum = SPI_getbinval(tuple, SPI_tuptable->tupdesc, 1, &isnull);
         if (!isnull) {
-            text *txt = DatumGetTextP(datum);
-            char *name = text_to_cstring(txt);
-            pk_names[i] = (name) ? cloudsync_string_dup(name) : NULL;
-            if (name) pfree(name);
+            // information_schema.column_name is of type 'name', not 'text'
+            Name namedata = DatumGetName(datum);
+            char *name = NameStr(*namedata);
+            pk_names[i] = cloudsync_string_dup(name);
         }
 
         // Cleanup on allocation failure