pcp
[Top] [All Lists]

Re: pcp updates: python3 work

To: pcp@xxxxxxxxxxx
Subject: Re: pcp updates: python3 work
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon, 23 Feb 2015 18:08:37 -0500 (EST)
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <y0md250he5j.fsf@xxxxxxxx>
References: <368618270.13212567.1424688934926.JavaMail.zimbra@xxxxxxxxxx> <763423700.13212590.1424688941302.JavaMail.zimbra@xxxxxxxxxx> <y0md250he5j.fsf@xxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: N6ZoUjVuQb3VCR+h0IBEXDa1uR6A1Q==
Thread-topic: pcp updates: python3 work

----- Original Message -----
> [...]
> What are your thoughts about the longevity of this
> approach (conditionally rewriting the #! line of python scripts)?  If
> only the build (or package-construction) scripts do rehashbanging,

It seems to be the best option available & from talking to the python
Fedora packagers, it's their recommended approach.

> developers/qa may not see how to easily test that any given script
> source is and stays workable in both python languages.

If you look at the commits closely, you will notice that they make
changes to QA allowing PCP_PYTHON_PROG (which defaults to python3 if
it is available, via configure.ac) to override the hash-bang line -
see e.g. qa/722, qa/985, etc.  They also ensure python3 is used when
new python PMDAs are installed in pmcd.conf.

For those of us doing QA, this gives us excellent coverage in testing
PCP using whichever python we choose to prefer.  IOW, that is the case
for either upstream QA runs from Makepkgs or internal Red Hat QE runs
on fedpkg builds.

It also gives platform distributors the freedom to choose whichever
python version they want as their default (as in Arch Linux kind of
default), and PCP will be ready and well-tested either way.


So, in summary, yes I think this is a really good way forward and it
is certainly going to get us to an end result of stable, widely-used
support for both python2 and python3 runtimes with the PCP python-ey
bits.

cheers.

--
Nathan

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