Re: [xmlblaster-devel] c socket lib

[Marcel Ruff <mr at marcelruff.info>]

Martin Johnson wrote:
> Hi All,
>
>
> I've come accross a couple of minor probs with the c-socket lib under
> windows for the current release. Some of these may have old news by now.
>
> There are memory allocations made by the lib under the semantics that the
> client shall 'free' the memory.
>
> Unfortunatly this will not work on this platform, different heaps etc, lib
> could be compiled under vc7 and client under vc6 etc...
>
> There should be a library supported functions for memory allocation etc.

The C-library has a free() wrapper which support this:

Dll_Export void xmlBlasterFree(char *p)
{
   if (p != (char *)0) {
      free(p);
   }
}

> Also for the record, 'bool' though probably supported by 'C' for a long time
> is another gotcha under vc6 this would compile though when if(func()) was
> tested it would fail.
>
> with return bool it would go something like mov ah, 1 but when if() came it
> would go like cmp eax, 1
> and this was the cause of the error. So beware.

So this is a bug in VC6,
do all our demo and test C clients run fine under VC6
or do we have code which is hit by the VC6 bug?
Is there any easy workaround like #ifdef VC6 -> typedef bool long; ?

> Also there is a memory leak in  'freeMsgUnitArrInternal(MsgUnitArr
> *msgUnitArr)'
>
> free(msgUnitArr->msgUnitArr); needs to be uncommented, otherwise I see a
> leak.

In my msgUtil.c the code is:

Dll_Export void freeMsgUnitArrInternal(MsgUnitArr *msgUnitArr)
{
   size_t i;
   if (msgUnitArr == (MsgUnitArr *)0) return;
   for (i=0; i<msgUnitArr->len; i++) {
      freeMsgUnitData(&msgUnitArr->msgUnitArr[i]);
   }
   free(msgUnitArr->msgUnitArr);
   msgUnitArr->len = 0;
}

This looks fine for me.

Thanks,

Marcel
> Thanks Guys
>
>
> Martin


--
http://www.xmlBlaster.org