Re: import wip/unit

["Sergey A. Osokin" <osa%freebsd.org@localhost>]

On Sun, Feb 07, 2021 at 02:36:29PM +0100, Roland Illig wrote:
> On 07.02.2021 09:49, nia wrote:
> > On Fri, Feb 05, 2021 at 08:55:07PM +0100, Roland Illig wrote:
> >> In all the patches, I'm missing a remark about the upstream status of
> >> the patches.  The pkgsrc policy is to have as few patches as possible,
> >> therefore each patch should be submitted upstream and the issue URL or
> >> other identifying information should be noted in the comment section of
> >> the patch.
> >>
> >> This is something that is not yet implemented in pkglint since I didn't
> >> find time to discuss whether this should be an enforced rule or how the
> >> proper format of the upstream status should look like, so that pkglint
> >> can interpret it reliable.  It is documented in the pkgsrc guide though:
> >>
> >> https://www.netbsd.org/docs/pkgsrc/pkgsrc.html#components.patches.caveats
> >
> > This is nice but doesn't really reflect the reality of pkgsrc, there
> > are many packages with a huge chunk of patches with unclear status.
> >
> > I think going through the ordeal of upstreaming patches is a worthwhile
> > clause, but there's often valid reasons for avoiding it (upstream projects
> > not caring or being actively hostile to minority platforms like NetBSD
> > and illumos for example). I personally got burnt out from upstreaming
> > in some cases due to stress.
> >
> > So, it shouldn't be required for new packages.
> 
> I'm totally fine with a remark in the patch that "upstream doesn't
> accept patches" or "upstream is dead" or anything like this.  I just
> don't like to have patches around whose status is unclear.
> 
> pkgsrc has around 20000 patches that are missing this kind of
> documentation, and for several of these it isn't even clear what the
> goal of the patch is.  That's what I want to avoid, especially for new
> packages where we can still get this information.  It's much more
> difficult to guess the goal of a patch for a package from 2001 that is
> "maintained" by pkgsrc-users%NetBSD.org@localhost and the original author of that
> patch is not reachable anymore.
> 
> For several times I thought about removing the check for undocumented
> patches from pkglint since it produces the majority of errors, but that
> would only make the situation worse than it is right now.  It still
> needs to be marked as error (not as warning) since undocumented patches
> make future maintenance far too laborious.

Hi Roland,

thanks for the update.

Let me add my 2 cents to this discussion.

To be honest, I've already reported about the patches we have for NGINX
Unit package to the vendor, cause I'm working for the vendor.  And I
got a response about some of those:

https://wip.pkgsrc.org/cgi-bin/gitweb.cgi?p=pkgsrc-wip.git;a=blob;f=unit/patches/patch-src_nxt__php__sapi.c
looks strange not only for you, but for the Unit development team as well.
From other side, otis@ added that patch to the package due to a build
error on NetBSD-current.

https://wip.pkgsrc.org/cgi-bin/gitweb.cgi?p=pkgsrc-wip.git;a=blob;f=unit/patches/patch-src_nxt__kqueue__engine.c
is related to NetBSD-current only.

Some other patches are going to be accepted in case they don't brake
anything on other platforms.

Hope this helps to import NGINX Unit to the main package tree.

-- 
Sergey Osokin