From nobody Sat Aug 09 20:35:34 2025 X-Original-To: freebsd-current@mlmmj.nyi.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mlmmj.nyi.freebsd.org (Postfix) with ESMTP id 4bzt2W4TMsz64Kw4 for ; Sat, 09 Aug 2025 20:35:47 +0000 (UTC) (envelope-from vermaden@interia.pl) Received: from smtpo73.interia.pl (smtpo73.interia.pl [217.74.67.73]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-SHA256 (128/128 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 4bzt2W147Lz3H8j; Sat, 09 Aug 2025 20:35:46 +0000 (UTC) (envelope-from vermaden@interia.pl) Authentication-Results: mx1.freebsd.org; none Date: Sat, 09 Aug 2025 22:35:34 +0200 From: vermaden Subject: Re: PKGBASE Removes FreeBSD Base System Feature To: Warner Losh , David G Lawrence Cc: Tomoaki AOKI , Michal Meloun , FreeBSD Current X-Mailer: interia.pl/pf09 In-Reply-To: References: <20250809062925.GN26557@mailx.dglawrence.com> <929543B2-633E-44B5-B6F6-F292CCEADAB3@freebsd.org> <20250809065247.GO26557@mailx.dglawrence.com> <96820ff6-bdb0-4d25-ad78-502e30b7e479@FreeBSD.org> <20250809185418.7d272536dd5862d0bdfd39c2@dec.sakura.ne.jp> <20250809101145.GS26557@mailx.dglawrence.com> X-Originating-IP: 45.148.42.1 Message-Id: List-Id: Discussions about the use of FreeBSD-current List-Archive: https://lists.freebsd.org/archives/freebsd-current List-Help: List-Post: List-Subscribe: List-Unsubscribe: Sender: owner-freebsd-current@FreeBSD.org MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="=-Uqe0AMNuwGRd2AX0vwOJ" DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=interia.pl; s=biztos; t=1754771739; bh=kPNp78HL1pg/y3IlvvRlnjkBPMDM01MmqPHlEhNk91Q=; h=Date:From:Subject:To:Message-Id:MIME-Version:Content-Type; b=nCm16LBhuQkL+mjcux3XlIgrK/lQerQI5LyGq6mo0ox3FogJaMd7e+DSYfl5G7auQ hRCZYDEuc7ZMJE0DdNJDBjiXdjDwCRHNlLFWr2gi7l6BTWUmTRRCeERDkdlDwoARi/ BbTrUTkg2hiaKFBf85veJGh13JRbyDyrKiQrgAP0= X-Rspamd-Queue-Id: 4bzt2W147Lz3H8j X-Spamd-Bar: ---- X-Rspamd-Pre-Result: action=no action; module=replies; Message is reply to one we originated X-Spamd-Result: default: False [-4.00 / 15.00]; REPLY(-4.00)[]; ASN(0.00)[asn:16138, ipnet:217.74.64.0/22, country:PL] --=-Uqe0AMNuwGRd2AX0vwOJ Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable > Fun fact: bectl would completely fix the pkg delete issue.Does bectl(8= ) works with UFS filesystem?I know that this is rhetorical question.If prop= erly installed FreeBSD can even have UFS Boot Environments:- https://vermad= en.wordpress.com/2021/04/02/ufs-boot-environments/But that is just workarou= nd for the problem - not a solution.Regards,vermadenTemat: Re: PKGBASE Remo= ves FreeBSD Base System FeatureData: 2025-08-09 13:17Nadawca: "Warner Losh"= <imp@bsdimp.com>Adresat: "David G Lawrence" <dg@dglawrence.com>= ;; DW: "Tomoaki AOKI" <junchoon@dec.sakura.ne.jp>; "Michal Meloun" &l= t;mmel@freebsd.org>; "FreeBSD Current" <freebsd-current@freebsd.org&g= t;; On Sat, Aug 9, 2025, 4:13=E2=80=AFAM David G Lawrence <dg@dglawrence= .com> wrote:> On Sat, 9 Aug 2025 09:08:53 +0200 > Michal Meloun &l= t;mmel@FreeBSD.org> wrote: > > > On 8/9/2025 8:52 AM, David G = Lawrence wrote: > > >> On 9 Aug 2025, at 07:29, David Greenman-= Lawrence <dg@dglawrence.com> wrote: > > >>> > > = >>> FWIW, I do have an opinion on this: I think that "pkg delete -= af" is > > >>> a useful thing that should not destroy your b= ase system. We should find > > >>> a way to make that work a= s it always has > > >> > > >> Today, it will destro= y any kernel modules installed from packages, including those necessary for= the display to work. It will destroy the ability for WiFi to work if you??= ?re using wifibox. There are a lot of other things in ports that are essent= ial to the system for some systems. > > > > > >  &n= bsp;  We can fix that, too. ...but destroying kernel modules doesn't m= ean your > > > system doesn't work - it just means X11 won't work = when you reboot - but then > > > it wouldn't anyway because you de= leted X11. But, here's the thing: X11 didn't > > > work when you f= irst installed the base system, either. And perhaps your > > > Wif= i didn't work out of the box, either. So you have some work to do to > &= gt; > get back basic functionality - but you knew that when you did the = > > > "pkg delete -af" in the first place. > > > > &g= t; > -DG > > > > I cannot but agree wholeheartedly.. The ac= tual situation with > > pkg delete and pkgbase is (for me) simply abs= urd. > > > > kmod in the ports is a different problem ??? it s= hows the inability of > > FBSD developers to implement these things = on their own,  so this > > solution  has some problems.&nb= sp; And don't kill me, I fully understand that > > it's not possible= , but that doesn't change the previous fact. > > > > Michal &= gt; > Sometimes yes, but sometimes no. > > On early but widely t= estable developement phase for drivers, especially > SD card drivers, ne= twork (including but not limited with WiFi) drivers > and disk controlle= rs, base is not a good place even for FreeBSD-native > drivers. > &g= t; This is because turnaround time for implememt (fix) / test / commit >= on base is usually take much longer days (or even months!) than in > po= rts. So recently, AFAIK, some drivers are first developed as > kmod port= s, and once stabilized, merged into main branch of src. > > What com= es in my mind is rtsx driver for Realtek SD card reader driver. Tomoaki, =    I see what you're saying and I agree completely. But, I think = this is pointing squarely at problem in the development paradigm for src co= mmitters. It should not take weeks or months to fix/test/commit/repeat in s= rc. It didn't used to be that way, so if it is now, then something has brok= en in the paradigm, and _that_ needs to be fixed.Fun fact: bectl would comp= letely fix the pkg delete issue. But I digress: rm -rf In the wrong spot al= so will kill base. It's a strange hill to die on. It also ignores common us= e cases, like wifibox, that make a system critically dependent on ports tha= t in simplwr times didn't happen...But some perspective on rtsx.  the = rtsx driver is an obscure edge case 1000 times less popular than the sdhci = driver. And even that is 1000x less popular than nvme. Given limited time a= nd lack of ability to buy the rtsx hardware easily, it's hard to justify us= ing my time for that driver when testing patches for other drivers is easie= r and benefits more people. That's why I passed over some of the changes th= ere, especially since there were big issues with that driver initially that= would have taken a lot of time to articulate. That is how I have prioritiz= ed my time on the thousands of fixes i have done for people, many the same = day. Using it as a posterchild for src being slow overstates the problems t= ypical patches have gwtting in.I have been trying to solve the actual, unde= rlying problems behind it: getting the pipeline flowing better through redu= ced friction for submissions (some good, but many lousy and it takes time t= o sort and you never know if a lot of feedback will produce a better outcom= e for any given problematic patch). Getting a deeper bench onboard and grow= ing aspects of our culture are also key areas needing help. I've had a hard= time getting others to help, assume ownership, follow through on promises,= etc. If you want to help fix things, it's helping me fix this problem. Fix= ing that increases the scarse developer resources and helps make it easier = to fix more issues. But 4 years in, it is a problem resistant to easy solut= ions.Warnet-DG  *  Dr. David G. Lawrence * * DG Labs   &nbs= p; Pave the road of life with opportunities.=20= --=-Uqe0AMNuwGRd2AX0vwOJ Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
> Fun fact: bectl would completely fix the pkg delete i= ssue.

Does bectl(8) works with UFS filesystem?

I know that th= is is rhetorical question.

If properly installed FreeBSD can even ha= ve UFS Boot Environments:
- https://vermaden.wordpre= ss.com/2021/04/02/ufs-boot-environments/

But that is just workar= ound for the problem - not a solution.

Regards,
vermaden

<= br>

Temat: Re: PKGBASE Removes FreeBSD= Base System Feature
Data: 2025-08-09 13:17
Nadawca: "Warner Losh" &l= t;imp@bsdimp.com>
Adresat: "David G Lawrence" <dg@dglawrence.com&g= t;;
DW: "Tomoaki AOKI" <junchoon@dec.sakura.ne.jp>; "Michal Melou= n" <mmel@freebsd.org>; "FreeBSD Current" <freebsd-current@freebsd.= org>;





On Sat, Aug 9, 2025, 4:13=E2=80= =AFAM David G Lawrence <dg@dglawrence.com> wrote:
<= blockquote style=3D"margin: 0 0 0 .8ex; border-left: 1px #ccc solid; paddin= g-left: 1ex;" class=3D"gmail_quote" data-mce-style=3D"margin: 0 0 0 .8ex; b= order-left: 1px #ccc solid; padding-left: 1ex;">> On Sat, 9 Aug 2025 09:= 08:53 +0200
> Michal Meloun <mmel@FreeBSD.org> wrote:
>=
> > On 8/9/2025 8:52 AM, David G Lawrence wrote:
> > = >> On 9 Aug 2025, at 07:29, David Greenman-Lawrence <dg@dglawrence.com> wrote:
> > >>>
= > > >>> FWIW, I do have an opinion on this: I think that "pk= g delete -af" is
> > >>> a useful thing that should not = destroy your base system. We should find
> > >>> a way t= o make that work as it always has
> > >>
> > >= > Today, it will destroy any kernel modules installed from packages, inc= luding those necessary for the display to work. It will destroy the ability= for WiFi to work if you???re using wifibox. There are a lot of other thing= s in ports that are essential to the system for some systems.
> >= >
> > >=C2=A0 =C2=A0 =C2=A0We can fix that, too. ...but d= estroying kernel modules doesn't mean your
> > > system doesn'= t work - it just means X11 won't work when you reboot - but then
> &= gt; > it wouldn't anyway because you deleted X11. But, here's the thing:= X11 didn't
> > > work when you first installed the base syste= m, either. And perhaps your
> > > Wifi didn't work out of the = box, either. So you have some work to do to
> > > get back bas= ic functionality - but you knew that when you did the
> > > "p= kg delete -af" in the first place.
> > >
> > > -= DG
> >
> > I cannot but agree wholeheartedly.. The act= ual situation with
> > pkg delete and pkgbase is (for me) simply = absurd.
> >
> > kmod in the ports is a different probl= em ??? it shows the inability of
> > FBSD developers to implemen= t these things on their own,=C2=A0 so this
> > solution=C2=A0 ha= s some problems.=C2=A0 And don't kill me, I fully understand that
>= > it's not possible , but that doesn't change the previous fact.
&g= t; >
> > Michal
>
> Sometimes yes, but sometim= es no.
>
> On early but widely testable developement phase f= or drivers, especially
> SD card drivers, network (including but not= limited with WiFi) drivers
> and disk controllers, base is not a go= od place even for FreeBSD-native
> drivers.
>
> This = is because turnaround time for implememt (fix) / test / commit
> on = base is usually take much longer days (or even months!) than in
> po= rts. So recently, AFAIK, some drivers are first developed as
> kmod = ports, and once stabilized, merged into main branch of src.
>
&= gt; What comes in my mind is rtsx driver for Realtek SD card reader driver.=

Tomoaki,

=C2=A0 =C2=A0I see what you're saying and I agr= ee completely. But, I think this is
pointing squarely at problem in the= development paradigm for src committers.
It should not take weeks or m= onths to fix/test/commit/repeat in src. It
didn't used to be that way, = so if it is now, then something has broken
in the paradigm, and _that_ = needs to be fixed.

=
Fun fact: bectl would completely fix the pkg delete issue= . But I digress: rm -rf In the wrong spot also will kill base. It's a stran= ge hill to die on. It also ignores common use cases, like wifibox, that mak= e a system critically dependent on ports that in simplwr times didn't happe= n...

But some perspectiv= e on rtsx.=C2=A0 the rtsx driver is an obscure edge case 1000 times less po= pular than the sdhci driver. And even that is 1000x less popular than nvme.= Given limited time and lack of ability to buy the rtsx hardware easily, it= 's hard to justify using my time for that driver when testing patches for o= ther drivers is easier and benefits more people. That's why I passed over s= ome of the changes there, especially since there were big issues with that = driver initially that would have taken a lot of time to articulate. That is= how I have prioritized my time on the thousands of fixes i have done for p= eople, many the same day. Using it as a posterchild for src being slow over= states the problems typical patches have gwtting in.

I have been trying to solve the actual, underl= ying problems behind it: getting the pipeline flowing better through reduce= d friction for submissions (some good, but many lousy and it takes time to = sort and you never know if a lot of feedback will produce a better outcome = for any given problematic patch). Getting a deeper bench onboard and growin= g aspects of our culture are also key areas needing help. I've had a hard t= ime getting others to help, assume ownership, follow through on promises, e= tc. If you want to help fix things, it's helping me fix this problem. Fixin= g that increases the scarse developer resources and helps make it easier to= fix more issues. But 4 years in, it is a problem resistant to easy solutio= ns.

Warnet

-DG

= =C2=A0*=C2=A0 Dr. David G. Lawrence
* * DG Labs
=C2=A0 =C2=A0 Pave = the road of life with opportunities.



--=-Uqe0AMNuwGRd2AX0vwOJ--