Re: Testing mksh as bootstrap shell

[Jonathan Perkin <jperkin%joyent.com@localhost>]

* On 2020-06-24 at 13:48 BST, Jonathan Perkin wrote:

> So, on Benny's recommendation, I took a look at shells/mksh instead,
> and it actually looks pretty good.  Not much larger than dash, looks
> very portable, and so far has only needed one small patch to fix an
> issue with bootstrapping using the legacy Solaris /bin/sh.
> 
> I have a branch here:
> 
>   https://github.com/joyent/pkgsrc/tree/joyent/dev/bootstrap-mksh

I have two completed bulk builds with this enabled.

 Darwin:
  https://us-east.manta.joyent.com/pkgsrc/public/reports/Darwin/10.15/trunk/x86_64/20200624.1234/meta/report.html

 SmartOS:
  https://us-east.manta.joyent.com/pkgsrc/public/reports/upstream-bulktest/20200624.2229/meta/report.html

Both look good, and I've also now included the "ulimit -m" patch on
macOS as discussed in that branch which will fix the main fallout from
libcups.

Upstream have also included both of my patches so the next release
shouldn't require any manual pkgsrc patching:

  https://twitter.com/Knoblauchkeks/status/1276165545285562375

In light of successful tests and a responsive upstream, I'd like to go
ahead and import mksh as an alternative bootstrap shell, and enable it
at least initially on Darwin to work around the SIP issues.

Over time we can look at converting other platforms away from the
crufty pdksh as and when they are tested.

If you have any objections let me know, as I said earlier I want us to
make a one-time decision on this so as not to bloat the tree, and we
definitely don't want to end up with 3+ bootstrap shells.

Thanks,

-- 
Jonathan Perkin  -  Joyent, Inc.  -  www.joyent.com