pcp
[Top] [All Lists]

Re: getdate.y in libpcp

To: "Frank Ch. Eigler" <fche@xxxxxxxxxx>
Subject: Re: getdate.y in libpcp
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Thu, 19 Feb 2015 20:32:29 +1100
Cc: Stan Cox <scox@xxxxxxxxxx>, pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <y0mh9ujjimu.fsf@xxxxxxxx>
References: <54E438F3.4070700@xxxxxxxxxxxxxxxx> <y0mh9ujjimu.fsf@xxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.4.0
On 19/02/15 03:20, Frank Ch. Eigler wrote:
Ken McDonell <kenj@xxxxxxxxxxxxxxxx> writes:

[...]
Where did the getdate.y source come from?

The initial pcp commit for getdate.y identified gnulib as the source.
(It would ideally have given the exact version.)

Thanks for the background Frank.

Pending a decision on the questions I pose below, I'll add comments to the head of the file to capture this history.

...
... a complication with this is that gnulib is now GPL3-licensed.
Because PCP is LGPL2+/GPL2+, scox was forced to import a pre-GPL3
snapshot of the gnulib code.  That means that any upstream
improvements need to be sort of reinvented.

OK.

I can see 3 possible ways forward:

1. move PCP to GPL3 ... by itself this issue is not important enough to open that Pandora's box of worms!

2. use the FreeBSD version instead ... looks like a fair amount of work and the potential differences in implementation may mean that the end result will not mimic gnu date(1) and so qa/752 will still fail

3. leave sleeping dogs lie ... the problem is a corner case involving the semantics for -S (or -T) "<someday>" when today is <someday>, e.g. thursday as I type. I doubt anyone has used this, and even if they have I'd argue they should not be surprised by the semantics of <someday> being the "next <someday>" (in this one case), especially given the totally confusing semantics of "first <someday>" and "next <someday>" that are implemented in date(1) and copied in libpcp ... not to mention the irrelevance of all of this for non-English speakers!

I vote for 3.

If we agree, I'd split qa/752 into a passing part and a retired part that captures the failure scenario with a big explanation at in the both qa tests to explain what has happened and why.

Opinions?

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