pcp
[Top] [All Lists]

Re: [pcp] Too much QA testing for IPv6!? ;)

To: pcp@xxxxxxxxxxx
Subject: Re: [pcp] Too much QA testing for IPv6!? ;)
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Thu, 21 Mar 2013 16:45:49 +1100
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <1344253533.22202391.1363740717091.JavaMail.root@xxxxxxxxxx>
References: <1344253533.22202391.1363740717091.JavaMail.root@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130308 Thunderbird/17.0.4
On 20/03/13 11:51, Nathan Scott wrote:
> Hi Dave,
> 
> Just a note for when you're back from vacation - I'm seeing this
> (below) after addition of the IPv6 host/address wildcarding code,
> which indicates a huge increase in the time to run this test (062).
> AFAICT, it is spending lots of time in the IPv6 loops, and I think
> the reverse name lookups (and so many of them) are where the time
> ends up going.

Further to this ...

062 is failing on lots of platforms because the output from __pmAccDumpHosts() 
is not deterministic ... the host-spec field sometimes contains :: and 
sometimes contains :0: for the octet after fec0, e.g. fec0::6:13:2:6:13:2 vs 
fec0:0:6:13:2:6:13:2 ... I suspect the problem is somewhere in or below 
__pmSockAddrToString()

062 is failing on other platforms because this representative error

[Thu Mar 21 08:57:58] pcp(10098) Error: __pmGetAddrInfo(fec0::0:11:0:0:11:0), 
Unknown host

appears 320 times in the output.

And finally (the time to run issue), I think 062 should not be doing IPv6 at 
all (if that is possible) because this test was motivated by an obscure error 
in IRIX that dates back to the Crimean War ... I would much rather see some 
shorter and different variant of the test used for IPv6 if the name/address 
translations are going to be so slow for IPv6.

Cheers, Ken.

<Prev in Thread] Current Thread [Next in Thread>