Re: How far should one go preventing potential memory leaks?

[Thomas Jericke <tjericke@...>]

On 05/10/2016 09:42 AM, Marc Balmer wrote:
> > Am 10.05.2016 um 09:13 schrieb Thomas Jericke <tjericke@indel.ch>:
> >
> > On 05/10/2016 09:10 AM, Thomas Jericke wrote:
> > > On 05/10/2016 08:41 AM, Marc Balmer wrote:
> > > > This prevents a memory leak if lua_pushstring() raises an error, but I will end
> > > > up with a number of rather small allocations that are only used for a short
> > > > period of time.
> > > If they buffers are really that small, I would just use the stack.
> > >
> > > char s[32];
> > > snprintf(s, sizeof(s), "bumblebee");
> > > lua_pushstring(L, s);
> > PS. I swear I typed this before I saw Sean's answer :-/
> Hehe ;)
>
> Well, in the real code the allocations are made for user supplied parameters, so I don't
> know neither the size of the parameters not their count (which can be up to 65535), so
> I really need to allocate the memory.
>
> However, many thanks to everyone who replied.
You can reuse the buffer and you can make a fall-back for large parameters.

char buff[1024];

for(unsigned long index = 0; index < 65535; index++) {
  size_t len = strlen(args[index]);
  if(len + 6 <= sizeof(buff)) {
    sprintf(buff, "arg1=%s", args[index]);
    lua_pushstring(L, buff);
  }
  else {
    char *s = lua_newuserdata(L, len + 6);
    sprintf(s, "bumblebee");
    lua_pushstring(L, s)
  }
}

Assuming you need the strlen anyway to know how large your buffer must 
be, it is only an additional "if".
I think most of the time the additional "if" statement will pay off. If 
you want to be sure you can simply profile your parameter length.
--
Thomas