Re: [PATCH] Fix thread-unsafe lazy initializations

[Ludovic Courtès]

Mark H Weaver <address@hidden> skribis:

> address@hidden (Ludovic Courtès) writes:
>
>> Mark H Weaver <address@hidden> skribis:
>>
>>> --- a/doc/ref/api-modules.texi
>>> +++ b/doc/ref/api-modules.texi
>>> @@ -942,14 +942,15 @@ the @var{name} is not bound in the module, signals an 
>>> error.  Returns a
>>>  variable, always.
>>>  
>>>  @example
>>> -SCM my_eval_string (SCM str)
>>> address@hidden
>>> -  static SCM eval_string_var = SCM_BOOL_F;
>>> +static SCM eval_string_var;
>>>  
>>> -  if (scm_is_false (eval_string_var))
>>> -    eval_string_var =
>>> -      scm_c_public_lookup ("ice-9 eval-string", "eval-string");
>>> +void my_init (void)
>>> address@hidden
>>> +  eval_string_var = scm_c_public_lookup ("ice-9 eval-string", 
>>> "eval-string");
>>> address@hidden
>>>  
>>> +SCM my_eval_string (SCM str)
>>> address@hidden
>>>    return scm_call_1 (scm_variable_ref (eval_string_var), str);
>>>  @}
>>>  @end example
>>
>> The doc should say something about ‘my_init’, and perhaps mention
>> locking?
>
> Can you suggest some text?  'my_init' seems obvious to me, and given
> that this is just an example for 'scm_c_*_lookup', it seems out of place
> to talk about general locking issues here, but feel free to propose some
> text to go here.

Agreed.  Perhaps just something like: “Note that the program should
ensure that ‘my_init’ is called only once, and in a thread-safe way.”

Ludo’.