Re: Doubling performance of pkgsrc/mk

[Jonathan Perkin <jperkin%mnx.io@localhost>]

* On 2022-10-24 at 19:04 BST, John Klos wrote:

> This makes a HUGE difference on slower hardware. For instance, here's 
> benchmark/flops (I wasn't going to wait for Python to compile) on a 33 
> MHz m68030 running NetBSD-9.99.100:
>
> => Creating binary package /usr/pkgsrc/packages/All/flops-2.0.tgz
>     1625.10 real       672.92 user       749.24 sys
>
> versus
>
> => Creating binary package /usr/pkgsrc/packages/All/flops-2.0.tgz
>      958.06 real       389.40 user       415.06 sys

Great!

> Are there edge cases we should test, or are you running bulk builds 
> with this tree to shake things out?

I'm running bulk builds, and the latest one just finished:

  https://us-central.manta.mnx.io/pkgsrc/public/reports/upstream-bulktest/20221023.1802/meta/report.html

I can't find any failures in there that are due to my changes, so I 
think I've now covered all of the issues that I was aware of, however 
I'd still love extra testing of this, just because it's such a large 
number of changes to critical parts of the infrastructure, and I can't 
rule out edge-cases that change behaviour and now no longer work.

I'll also let it sit in my release branches for a bit, with the aim of 
getting it in with plenty of time for additional widespread testing 
prior to the 2022Q4 freeze.

Cheers,

--
Jonathan Perkin   -   mnx.io   -   pkgsrc.smartos.org
Open Source Complete Cloud   www.tritondatacenter.com