only re-parse data if it's a bare word

if we auto-convert XLAT_TMPL + TMPL_TYPE_DATA to XLAT_BOX when
parsing expressions, then this code is likely no longer necessary

diff --git a/src/lib/unlang/xlat_expr.c b/src/lib/unlang/xlat_expr.c
index 3a184c4fa54fe92fba0856cee3c2dffee6546946..20a27062a2aae275721a6597dcf28b3710b5574c 100644 (file)
--- a/src/lib/unlang/xlat_expr.c
+++ b/src/lib/unlang/xlat_expr.c
@@ -1427,7 +1427,7 @@ static xlat_action_t xlat_func_unary_complement(TALLOC_CTX *ctx, fr_dcursor_t *o
        return xlat_func_unary_op(ctx, out, xctx, request, in, T_COMPLEMENT);
 }
 
-/** Convert XLAT_BOX arguments to XLAT_TMPL
+/** Convert XLAT_BOX arguments to the correct data type.
  *
  *  xlat_tokenize() just makes all unknown arguments into XLAT_BOX, of data type FR_TYPE_STRING.  Whereas
  *  xlat_tokenize_expr() calls tmpl_afrom_substr(), which tries hard to create a particular data type.
@@ -1451,6 +1451,7 @@ static int xlat_function_args_to_tmpl(xlat_inst_ctx_t const *xctx)
                if (!node) continue;
                if (node->type != XLAT_BOX) continue;
                if (node->data.type != FR_TYPE_STRING) continue;
+               if (node->quote != T_BARE_WORD) continue;
 
                /*
                 *      Try to parse it.  If we can't, leave it for a run-time error.