Re: Spawn support in pkgsrc tools - Google Summer of Code 2016

[Andreas Theodosiou <andreas%atheod.net@localhost>]

On Sun, Mar 20, 2016 at 01:42:32AM +0100, Kamil Rytarowski wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
>
> > Greetings,
> >
> > I am writing to let you know that I have submitted a draft of the
> > proposal via the GSOC web interface. It is also available at
> > http://www.atheod.net/GSOC2016/atheod_project_proposal.pdf.
> >
> > As there is still some time until the deadline for subimission of
> > the final version of the proposal, I am open to suggestions and
> > feeback.
> >
>
> Looks good.
>
> Few notes from my side:
> - - when you start discussing fork(2) and vfork(2) please show the
> difference between them and their purposes, use-cases, applicability,
> - - please note that while vfork(2) is kind of optimization, this
> interface had been marked as obsolescent by POSIX and currently is
> absent in the standard:
> http://pubs.opengroup.org/onlinepubs/009695399/functions/vfork.html
> - - historically vfork(2) has been reimplemented in NetBSD,
> - - please note that fork(2) is expensive operation on some systems
> (notably Darwin),
> - - vfork(2) is currently out from at least Oracle Solaris,
> - - please list what operating systems do support posix_spawn(2),
> - - pdksh(1) there is no longer external active upstream, it's
> maintained in the NetBSD sources,

Thank you, these are reasonable and well thought suggestions.

> - - please try to check memory overhead complexity, not just time
> complexity,

Is there preferred way of doing this among the NetBSD project? Would a
heap analyser such as valgrind massif do?

> - - building release (with build.sh) is a good initial test, for the
> pgksrc world I'm suggesting to try out final tests with pbulk (if
> there is no shortage on quick hardware),
> - - personally I would focus on NetBSD base shells (with their pkgsrc
> equivalents): sh(1), ksh(1), csh(1) -- leaving bash(1), mksh(1) and
> others as optional,
> - - please ensure that there are ATF tests in the NetBSD base validating
> modified implementation of a functionality, if they are absent please
> add them.

This is a very reasonable suggestion. A quick look at base shows
that there are ATF tests for sh(1) and make(1) but not for ksh(1) and
csh(1).
>
> > Best regards,
> >
> > Andreas
> >
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v2
>
> iQIcBAEBCAAGBQJW7fH3AAoJEEuzCOmwLnZsFJ4P/3tneO9C/t2o1ag0eLZ23757
> VQQbLaQB26wgxpDMutVvzUwB5ui3W+GFgUJYeJJe9IdRtGHyRaoqSOKoDJ1Xj20A
> vsag2FHI3RCJrvPVZ4TWbiU5JV/LAhLDAIjI4+815xpUS/S92H4ErpiBzKe/SZl6
> rVkNWp0vmKx6TovD1kDwgM8X7O/+Qcefi+khYDxmQGO7qWwEZ9NgblH39rhgKVvp
> 2VCUNT35bB2itt/vkS/HKEs5nvlb0zphgOzqNn3vh2IHv1GTmxuFa5FQ1Coap0iG
> bqqmyKKndtKnpvet+KUfrTPb9CXHZBzff4CRMHnbOlE7KpNjOWq0lZyed29wJJdP
> SKUJI8NzSZg9LSnR30Kw6rGRvbuVLPtdfdkfCHpzBijrZoi/UGJ75dmrgxEXNX6f
> TT3S7IRuR6CbPfTIvpCwyjEP2zC5dBdRS3Sdx+DArKA6pvkpXCtf3p12wAH9IIz1
> 427G5fmicwJgXdg+TBmvb2Rcgt+vWxzdrrFV9huFzMxuuX5bNxJFa+CqekTE1QWJ
> Od8T6q+9U+pShrfao406Az8F+TOI3/WmSigQL9lyCDsHN/OWGBjRIY9dz2rVxtB+
> Hlz8bDC9yQlpnbKShb7kOrX9iS0LIIGGoCoLvybxdqhCIz3wEphfH0FpzFrVABAY
> kKgup5cNo8tZHeSujbrN
> =w5MN
> -----END PGP SIGNATURE-----

Best,

Andreas