NetBSD Problem Report #59998

From www@netbsd.org  Fri Feb 13 03:22:12 2026
Return-Path: <www@netbsd.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
	 key-exchange X25519 server-signature RSA-PSS (2048 bits)
	 client-signature RSA-PSS (2048 bits))
	(Client CN "mail.netbsd.org", Issuer "R13" (verified OK))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 007201A923C
	for <gnats-bugs@gnats.NetBSD.org>; Fri, 13 Feb 2026 03:22:11 +0000 (UTC)
Message-Id: <20260213032210.F0E291A923D@mollari.NetBSD.org>
Date: Fri, 13 Feb 2026 03:22:10 +0000 (UTC)
From: campbell+netbsd@mumble.net
Reply-To: campbell+netbsd@mumble.net
To: gnats-bugs@NetBSD.org
Subject: index packages by plists
X-Send-Pr-Version: www-1.0
X-From4GNATS: "campbell+netbsd@mumble.net via gnats" <gnats-admin@NetBSD.org>

>Number:         59998
>Category:       pkg
>Synopsis:       index packages by plists
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    pkg-manager
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Fri Feb 13 03:25:00 +0000 2026
>Last-Modified:  Fri Feb 13 07:10:01 +0000 2026
>Originator:     Taylor R Campbell
>Release:        
>Organization:
The IndexBSD Packaginc.
>Environment:
>Description:

	pkgin can search packages by name.  It would be nice if we
	could also search packages by what files they install.

>How-To-Repeat:

	Ask: What package do I install to get, say, /usr/pkg/bin/pdfjoin?

>Fix:

	Quick hack to create a locate(1) database of package files from
	my notes the other year:

# generate newline-separated list of $PATHNAME/$PKG
# (hope packages don't have any pathnames with newlines in them...)
# about 500 MB uncompressed for 2022Q4, 30 MB gzipped
# XXX handle long package names wider than 80 columns for progress output
(n=0; for x in *.tgz; do n=$((n + 1)); done; i=0; for x in *.tgz; do printf '\r%80s\r%s/%s %s' '' $i $n $x >&2; pkg_info -qL $x | sed -e 's,$,/'"$x"','; i=$((i + 1)); done | gzip -nc > /tmp/pkgindex.2022Q4.gz; printf '\r%80s\r\n' '' >&2)

# generate bigrams
BIGRAM=$(gunzip -c < /tmp/pkgindex.2022Q4.gz | /usr/libexec/locate.bigram)

# generate locate(1) database
# about 30 MB uncompressed for 2022Q4, 23 MB gzipped
/usr/libexec/locate.code "$BIGRAM" | gzip -nc >/tmp/pkglocate.2022Q4.gz

	Part of this pipeline, of course, is embarrassingly
	parallelizable.

	TBD:
	1. Integrate into pbulk or other infrastructure.
	2. Decide how to publish this.
	3. Ship tool (or pkgin patches) for automatically downloading,
	   caching, and querying the index.

>Audit-Trail:
From: Taylor R Campbell <riastradh@NetBSD.org>
To: gnats-bugs@NetBSD.org, netbsd-bugs@NetBSD.org
Cc: 
Subject: Re: pkg/59998: index packages by plists
Date: Fri, 13 Feb 2026 03:27:08 +0000

 Credit: This locate(1) method was inspired by a talk by Rodrigo Osorio
 at FOSDEM a couple years ago about FreeBSD's pkg-provides.  (I haven't
 looked into how it works beyond that talk, nor into others like
 Debian's apt-file search.)

From: Jonathan Perkin <jperkin@pkgsrc.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: pkg/59998: index packages by plists
Date: Fri, 13 Feb 2026 07:08:06 +0000

 >	Quick hack to create a locate(1) database of package files from
 >	my notes the other year:

 FWIW my repositories have shipped pkg_files.* archives for nearly 10 
 years now, generated with this:

    https://github.com/NetBSD/pkgsrc/commit/0b8a13aa9282c91030817692a2c38bea97599917

 I have not yet gotten around to hooking them up to pkgin though, and I'm 
 not sure I'm still keen on this approach, but just FYI.

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

Home	PR Database Search