Re: bin/50121: makemandb / man crashes on certain manual pages in netbsd-6

[David Holland <dholland-bugs%netbsd.org@localhost>]

The following reply was made to PR bin/50121; it has been noted by GNATS.

From: David Holland <dholland-bugs%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: bin/50121: makemandb / man crashes on certain manual pages in
 netbsd-6
Date: Thu, 6 Aug 2015 16:04:57 +0000

 On Mon, Aug 03, 2015 at 03:30:00PM +0000, he%NetBSD.org@localhost wrote:
  > 	The makemandb program crashes on certain man pages.  In my
  > 	case, the zabbix_agentd page from the pkgsrc-wip zabbix-agent
  > 	package makes this happen (under the -v option):
  > 
  > Parsing: /usr/pkg/man/man8/zabbix_agentd.8
  > assertion "MAN_BLOCK == p->parent->type" failed: file "/usr/src/external/bsd/mdocml/lib/libmandoc/../../dist/man.c", line 212, function "man_node_append"
  > Abort (core dumped)
  > 
  > 	The same happens if you do "man" on that manual page.
  > 	However, if you do "nroff -man" on the manpage, it formats
  > 	~ok.  The manpage doesn't use the new man page macros.
  > 
  > 	This has the slightly nasty side effect that "man -k" on this
  > 	machine is useless, as any lookup results in:
  > 
  > apropos: Unable to query schema version: disk I/O error
  > 
  > 	(I thought that error code / message was reserved for actual
  > 	hard IO failures, which this isn't an instance of.)
 
 That's a sqlite error message, which is probably a sqlite bug; it
 should probably be producing SQLITE_CORRUPT instead of
 SQLITE_IOERROR.
 
 But that's a side issue; crashing shouldn't corrupt the database.
 That's part of the point of using database software. This suggests
 that makemandb and/or apropos isn't using sqlite correctly.
 
  > 	There's quite a few updates to man/mkmandb in newer NetBSD
  > 	code, but it appears to drag along also an update of mdocml
  > 	and sqlite...
  > 
  > 	However, with a wholesale upgrade to HEAD of all those 4
  > 	components, makemandb no longer crashes, and "man
  > 	zabbix_agentd" works as expected.
  > 	Is this too invasive to contemplate pulling up to netbsd-6?
  > 	Or does there exist a "point fix" for this problem?
 
 Dunno. Updating sqlite is probably a fine idea, but by itself it's
 unlikely to fix the problem.
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost