pcp
[Top] [All Lists]

Re: [pcp] new pdubuf vs. qa/367

To: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Subject: Re: [pcp] new pdubuf vs. qa/367
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Tue, 10 Mar 2015 06:50:46 +1100
Cc: 'pcp developers' <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <20150309140459.GJ27936@xxxxxxxxxx>
References: <20150308161756.GH27936@xxxxxxxxxx> <01aa01d059e0$f599ad20$e0cd0760$@internode.on.net> <20150308205348.GI27936@xxxxxxxxxx> <54FD64C3.9080901@xxxxxxxxxxxxxxxx> <20150309140459.GJ27936@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.5.0
On 10/03/15 01:04, Frank Ch. Eigler wrote:
...
No big deal, but why are PDUs multiples of an int?  It may matter for
the archive format, but for network & other temporary purposes, does
the multiples-of-an-int property carry any benefit?

On some architectures there is benefit from the fields _within_ a PDU being word aligned. But there is no real reason for the PDU to be padded to a word boundary in the rare cases where the PDU does not end with an integer. However that's the way it is, and for compatibility I think we're stuck with it staying that way.

Most of the PDUs are not used at all for archive contexts, and in particular the PMNS ones at the center of this discussion are not relevant to archives.

...
(Just that the qa/367{,.out} files assert '~' padding in a bunch of
places.)

Ah, thanks.

...
Righto, thanks for the pointer.  Please see pcpfans.git

Good.

I've cherrypicked this commit into my tree and reviewed. It looks fine and is relevant for _any_ pdubuf implementation, so I'll push it up stream after some QA soak (it already passes -g libpcp on one platform).

Thanks, Frank.

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