[Ovmsdev] OvmsMetricString memory lifecycle

[Mark Webb-Johnson]

I think it pretty much ends the same.

Internally, there is a small fixed buffer in the string object that is used for small strings (I think 16 bytes). Anything larger ends up as a pointer to heap memory.

Regards, Mark.

> On 21 Aug 2018, at 4:06 AM, Tom Parker <tom at carrott.org> wrote:
> 
> On 21/08/18 05:19, Michael Balzer wrote:
>> in no case can passing & returning string objects by value result in a use after free or memory leak. Pass by value guarantees you'll be operating on a copy as
>> soon as necessary, and all compiler generated copies are guaranteed to be freed by the compiler.
> 
> Ah, pass by value isn't something I had considered.
> 
> However what happens in:
> 
> void OvmsMetricString::SetValue(std::string value)
>   {
>   if (m_value.compare(value)!=0)
>     {
>     m_value = value;
>     SetModified(true);
>     }
>   else
>     SetModified(false);
>   }
> 
> 
> value is passed in and then assigned to the member variable. If value is freed when SetValue() returns, is another copy made during the assignment? Will this potentially fragment the memory? Would m_value.assign(value); be better (assuming the backing store buffer isn't resized too often), or does m_value = value end up doing the same thing?
> 
> 
> I found my bug with the simcom status command -- I messed up the parameter somewhere between when I tested it and finished up for the day. First, it's not even the right parameter, and second, I'm printf %s'ing the object not the c_str().
> _______________________________________________
> OvmsDev mailing list
> OvmsDev at lists.openvehicles.com
> http://lists.openvehicles.com/mailman/listinfo/ovmsdev