pcp
[Top] [All Lists]

Re: [pcp] Retiring libpcp Errors

To: pcp@xxxxxxxxxxx
Subject: Re: [pcp] Retiring libpcp Errors
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed, 27 Jul 2016 06:36:24 +1000
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <5797C005.5070507@xxxxxxxxxx>
References: <5797C005.5070507@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.2.0
On 27/07/16 05:54, Dave Brolley wrote:
Hi,

I'm looking at a change which removes an unnecessary check (that the
archives in a multi-archive context all have the same time zone. I see
that libpcp error code/messages have been retired from time to time and
there is a message associated with this check which is now a candidate.
Do we offer a deprecation period before doing that?

This is a place where we don't (and probably should) have a policy that is articulated and enforced.

It impinges on the much deeper and more important issue of ABI compatibility that was unfortunately lost when we moved from IRIX (where the toolchain and run-time linker support for this were very strong) to Linux (where the support was very weak by comparison, although it has improved over the intervening decades).

In the absence of a policy, I'd suggest that the "grace" period varies from none to infinity depending on how visible the error code was to any apps outside PCP code base that were using the PMAPI. In the case you're referring to, I think this tends towards the "none" end of the scale.

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