netbsd-8 'pkg_info' segfaults during package builds?

["John D. Baker" <jdbaker%mylinuxisp.com@localhost>]

On the machine I'm using to update during the (barely noticed) pkgsrc
freeze, I'm seeing segfaults from 'pkg_info' on nearly every package.
This is seen while running 'pkg_rolling-replace -u'.

A typical occurrence is like:

RR> Selecting xfce4-genmon-plugin (sysutils/xfce4-genmon-plugin) as next package to replace
RR> Checking if xfce4-genmon-plugin has new depends...
RR> Replacing xfce4-genmon-plugin-4.0.0
===> Cleaning for xfce4-genmon-plugin-4.0.0nb1
=> Bootstrap dependency digest>=20010302: found digest-20160304
===> Skipping vulnerability checks.
WARNING: No /var/db/pkg/pkg-vulnerabilities file found.
WARNING: To fix run: `/usr/sbin/pkg_admin -K /var/db/pkg fetch-pkg-vulnerabilities'.
=> Checksum SHA1 OK for xfce4-genmon-plugin-4.0.0.tar.bz2
=> Checksum RMD160 OK for xfce4-genmon-plugin-4.0.0.tar.bz2
=> Checksum SHA512 OK for xfce4-genmon-plugin-4.0.0.tar.bz2
===> Installing dependencies for xfce4-genmon-plugin-4.0.0nb1
=> Tool dependency glib2-tools-[0-9]*: found glib2-tools-2.54.2
=> Tool dependency libtool-base>=2.4.2nb9: found libtool-base-2.4.6
=> Tool dependency intltool>=0.40.0: found intltool-0.51.0nb2
=> Tool dependency gmake>=3.81: found gmake-4.1nb3
=> Tool dependency perl>=5.0: found perl-5.26.1nb1
=> Tool dependency pkgconf-[0-9]*: found pkgconf-1.3.5
=> Build dependency x11-links>=1.06: found x11-links-1.06
=> Build dependency cwrappers>=20150314: found cwrappers-20171108
=> Full dependency gtk3+>=3.22.25nb2: found gtk3+-3.22.26
=> Full dependency xfce4-panel>=4.13.0nb1: found xfce4-panel-4.13.0nb1
[sits here a long time, then:]
[1]   Segmentation fault (core dumped) /usr/sbin/pkg_in...
===> Overriding tools for xfce4-genmon-plugin-4.0.0nb1
===> Extracting for xfce4-genmon-plugin-4.0.0nb1
===> Patching for xfce4-genmon-plugin-4.0.0nb1
=> Applying pkgsrc patches for xfce4-genmon-plugin-4.0.0nb1
===> Creating toolchain wrappers for xfce4-genmon-plugin-4.0.0nb1
[...]

The build continues and completes, so the 'pkg_info' segfault is
non-fatal to the build, but perhaps the information it was gathering
is incomplete.

'gdb' says:

(gdb) bt
#0  0x00007ae18336bdcd in __sfvwrite () from /usr/lib/libc.so.12
#1  0x00007ae18334627c in ?? () from /usr/lib/libc.so.12
#2  0x00007ae18334b88f in __vfprintf_unlocked_l () from /usr/lib/libc.so.12
#3  0x00007ae18330fd1a in vasprintf_l () from /usr/lib/libc.so.12
#4  0x00007ae18330fe02 in vasprintf () from /usr/lib/libc.so.12
#5  0x0000000026e11b3c in xasprintf ()
#6  0x0000000026e10930 in pkgdb_pkg_file ()
#7  0x0000000026e05c2c in read_meta_data_from_pkgdb ()
#8  0x0000000026e05f3e in build_full_reqby ()
#9  0x0000000026e05f62 in build_full_reqby ()
#10 0x0000000026e05f62 in build_full_reqby ()
[...]
#8999 0x0000000026e05f62 in build_full_reqby ()
#9000 0x0000000026e05f62 in build_full_reqby ()
#9001 0x0000000026e05f62 in build_full_reqby ()
[... "It's over 9000!"]
[ad nauseum]

So, it appears something is throwing it into an infinite loop and likely
terminates only when it has run out of RAM (5GB on this machine).

Only seen it on this machine so far:

NetBSD brenowin.reltuk.net 8.0_BETA NetBSD 8.0_BETA (KEPLER) #15: Fri Dec 22 15:38:11 CST 2017  sysop%brenowin.reltuk.net@localhost:/d0/build/netbsd-8/obj/amd64/sys/arch/amd64/compile/KEPLER amd64

as it's the crash-test dummy for problem packages in the run-up to to
pkgsrc-2017Q4.

There's no room on the root/usr filesystem for debug sets so I'll have
to see about redirecting that with a symlink to see if I can get more
information.

-- 
|/"\ John D. Baker, KN5UKS               NetBSD     Darwin/MacOS X
|\ / jdbaker[snail]mylinuxisp[flyspeck]com    OpenBSD            FreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645