Proposal: default to concurrent

[Nikolai Lifanov]

On 01/29/17 14:08, John W. O'Brien wrote:
> Hello FreeBSD Python,
> 
> One of the most common problems I encounter with python3
> interoperability is when the concurrent option is needed and can be
> trivially enabled. There is a growing list of bugs where this, on a
> individual port basis, has been fixed ([0], [1]), is in progress ([2],
> [3]), or is in my queue and perhaps others' to submit ([4]). On the
> other hand, there are lots of ports for which concurrent is a no-op, and
> lots more that don't support python3 at all meaning that concurrent has
> little to no chance to cause harm.
> 
> What I propose is to enable the concurrent behavior by default and to
> provide a feature to disable it when necessary.
> 
> I welcome supportive and dissenting comments as well as cautionary
> remarks about the likely pitfalls of pursuing this.
> 
> <snark>
> While it would be tempting to call the new feature something
> unflattering like "i_hate_python3" or "archaic_upstream" or
> "regressive", the most natural name is probably just "noconcurrent".
> </snark>
> 
> Regards,
> John
> 
> [0] textproc/py-xlrd: Allow concurrent installs
>     https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=209689
> [1] [PATCH] textproc/py-xlsxwriter: Enable concurrent installs
>     https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=215331
> [2] textproc/py-pygments: add concurrent to USE_PYTHON
>     https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=214295
> [3] math/py-bottleneck: bump to 1.20, install concurrently
>     https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=215263
> [4] devel/py-boto, devel/py-tables, devel/pep8, devel/flake8
> 

I think the biggest controversy is that binary names are suffixed, which
may not be what upstream projects like and/or document.
It may be confusing to users as well.

If we can get rid of the suffix for PYTHON_CONCURRENT_INSTALL for
whichever version is PYTHON_DEFAULT, it should be a net positive to
treat it the same way we did USE_GCC in the past (enable
PYTHON_NO_CONCURRENT or something and remove it on per-port basis once
fixed).

- Nikolai Lifanov

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 992 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/freebsd-python/attachments/20170129/3fa458dc/attachment.sig>