Re: [PATCH v2 21/31] Use ui_file_as_string in gdb/compile/

[Simon Marchi <simon dot marchi at polymtl dot ca>]

The doc for get_new_file_names should probably be updated:

/* Compute the names of source and object files to use.  The names are
   allocated by malloc and should be freed by the caller.  */

>  static void
> -get_new_file_names (char **source_file, char **object_file)
> +get_new_file_names (std::string *source_file, std::string 
> *object_file)
>  {


Same for compile_to_object, both for the "freeing" part and the part 
that talks about the return value.

We should use $other_language instead of C++, to be able to do

  source_filep, object_filep = compile_to_object (...)

:)

Btw, would it be good to use references for those strings?

>     error condition, error () is called.  The caller is responsible for
>     freeing both strings.  */
>
> -static char *
> +static void
>  compile_to_object (struct command_line *cmd, const char *cmd_string,
>  		   enum compile_i_scope_types scope,
> -		   char **source_filep)
> +		   std::string *source_filep,
> +		   std::string *object_filep)
>  {

...

> @@ -609,30 +606,27 @@ void
>  eval_compile_command (struct command_line *cmd, const char 
> *cmd_string,
>  		      enum compile_i_scope_types scope, void *scope_data)
>  {
> -  char *object_file, *source_file;
> +  std::string source_file, object_file;
>
> -  object_file = compile_to_object (cmd, cmd_string, scope, 
> &source_file);
> -  if (object_file != NULL)
> +  compile_to_object (cmd, cmd_string, scope, &source_file, 
> &object_file);
> +  if (!object_file.empty ())

Can this actually happen?