@nogc

[via Digitalmars-d]

On Saturday, 12 July 2014 at 09:44:16 UTC, Piotr Szturman wrote:
> W dniu 2014-07-11 05:18, Kapps pisze:
>> On Friday, 11 July 2014 at 02:32:00 UTC, Manu via 
>> Digitalmars-d wrote:
>>> So, we allow assert() in nothrow functions, the argument is 
>>> that
>>> assert is non-recoverable, so it's distinct from user 
>>> exceptions.
>>>
>>> I have this in my 'nothrow @nogc' function:
>>>  assert(false, "Message " ~ details);
>>>
>>> It complains "Error: cannot use operator ~ in @nogc function"
>>>
>>> I think it should be allowed to invoke the GC for formatting 
>>> error
>>> messages inside of assert statements, just the same as 
>>> assert() is
>>> allowed inside of nothrow functions.
>>>
>>> Thoughts?
>>
>> More generally, I'm somewhat surprised that @nogc does not 
>> still allow
>> allocating Errors (including with assert), as who cares if 
>> your program
>> may slightly pause when it's about to crash in a theoretically
>> unrecoverable way.
>
> It does matter when entire program is marked with @nogc, so 
> there may be no GC code at all (GC code may not be linked).

Well, it wouldn't even link then...

But I think the more common route to go in such a case is to make 
a stubbed out GC that provides only allocation functionality 
(e.g. forwarding to malloc/free), but doesn't do any garbage 
collection. After all, some kind of memory allocation needs to 
present in almost all cases, it's just that it's done manually.