dnsperf and BIND memory consumption

[Danny Mayer]

Doug Barton wrote:
> Vinny Abello wrote:
>> Just for clarification, is there any downside to this autoconf fix 
>> vs. how it was previously working?
> 
> It was not working correctly previously, so no.
> 
>> Does autoconf still not understand AMD64 on FreeBSD
> 
> You're confusing "autoconf" and "the configure script provided in the
> sources for BIND >= 9.5.x." There is also a variable of "what the
> FreeBSD ports provide to configure by default" that I have deliberately
> omitted for the sake of clarity.
> 
>> and is your fix just defaulting to something lesser?
> 
> No.
> 
>> It appeared that the switches being run were for 32 bit after this 
>> fix, but I'm no expert on compilers by far.
> 
> You'd have to dig into the source and really understand what's happening
> now vs. what was happening before in order for me to answer this
> question, and by the time you had done that work I would not need to
> answer this question for you. :)

You would have a very hard time finding it just using code inspection
since the bug was found several levels under the code since there were a
number of Macros in place and the source of the problem was in the
creation but the fix had to be elsewhere. I only found this leak because
I had accidentally left on a debugging tool when test running on a
Windows box and the debugging tool was causing it to break at the
creation point. It took me quite a while to peel back the layers and
find out why it was really complaining. The fix was easy, but
understanding what needed to be fixed and why was not.

Danny

> The short version is that right now the port is doing what it should be.
> 
> 
> hth,
> 
> Doug
> 
>