NetBSD Problem Report #60219

From wiz@netbsd.org  Thu Apr 30 19:30:57 2026
Return-Path: <wiz@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) server-digest SHA256
	 client-signature RSA-PSS (2048 bits) client-digest SHA256)
	(Client CN "mail.netbsd.org", Issuer "R12" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id E90D01A9239
	for <gnats-bugs@gnats.NetBSD.org>; Thu, 30 Apr 2026 19:30:56 +0000 (UTC)
Message-Id: <20260430193056.7376084D27@mail.netbsd.org>
Date: Thu, 30 Apr 2026 19:30:56 +0000 (UTC)
From: wiz@NetBSD.org
Reply-To: wiz@NetBSD.org
To: gnats-bugs@NetBSD.org
Subject: sysconf overpromises TZNAME_MAX
X-Send-Pr-Version: 3.95
X-From4GNATS: "wiz@NetBSD.org via gnats" <gnats-admin@NetBSD.org>

>Number:         60219
>Category:       lib
>Synopsis:       sysconf overpromises TZNAME_MAX
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kre
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Apr 30 19:35:00 +0000 2026
>Closed-Date:    Thu May 07 18:25:29 +0000 2026
>Last-Modified:  Thu May 07 18:25:29 +0000 2026
>Originator:     Thomas Klausner
>Release:        NetBSD 11.99.5
>Organization:

>Environment:
11.99.5/20260408 and 11.0_RC3
Architecture: x86_64
Machine: amd64
>Description:

When running the gnulib tests for parse-datetime, there is a test failure.
The problem is that the code wants to test extreme cases, so it generates
a time zone string of maximum allowed length and expects that libc will parse
it, but instead it is refused as too long.

Bruno Haible investigated and found that even though sysconf(3) returns
511 for sysconf (_SC_TZNAME_MAX), which is also what I see on my system:

# getconf TZNAME_MAX
511

libc seems to have a limit of 254 instead.

I guess, but haven't verified, that this is from src/lib/libc/time/localtime.c

#ifndef TZNAME_MAXIMUM
# define TZNAME_MAXIMUM 254
#endif

I'm not sure where the 511 for _SC_TZNAME_MAX comes from; however, while
looking for this I found in lib/libc/gen/sysctl.c:

                    _INT("tzname_max", USER_TZNAME_MAX, NAME_MAX,
                      "The minimum maximum number of types supported for the "
                      "name of a timezone"),

where I think 'types' should be replaced by 'characters'.


>How-To-Repeat:
$ git clone --depth=1 https://git.savannah.gnu.org/git/gnulib.git
$ cd gnulib
$ ./gnulib-tool --create-testdir --dir t parse-datetime
$ cd t
$ ./configure && make && make check

>Fix:
Adapt the real limit to the reported limited, or the other way round.

>Release-Note:

>Audit-Trail:

Responsible-Changed-From-To: lib-bug-people->kre
Responsible-Changed-By: kre@NetBSD.org
Responsible-Changed-When: Fri, 01 May 2026 04:36:04 +0000
Responsible-Changed-Why:
I am looking into this PR


From: Robert Elz <kre@munnari.OZ.AU>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: lib/60219: sysconf overpromises TZNAME_MAX
Date: Fri, 01 May 2026 12:19:08 +0700

     Date:        Thu, 30 Apr 2026 19:35:00 +0000 (UTC)
     From:        "wiz@NetBSD.org via gnats" <gnats-admin@NetBSD.org>
     Message-ID:  <20260430193500.7870C1A9241@mollari.NetBSD.org>


   | Bruno Haible investigated and found that even though sysconf(3) returns
   | 511 for sysconf (_SC_TZNAME_MAX), which is also what I see on my system:
   |
   | # getconf TZNAME_MAX
   | 511
   |
   | libc seems to have a limit of 254 instead.

 Yes, that is clearly broken, sysconf(_SC_TZNAME_MAX) is supposed to
 return the minimum value that TZNAME_MAX can be - the POSIX default for that
 is 6 (which is perhaps a bit short, but these are the abbreviations
 ("EST" etc) and don't need to be all that long).    The current tzcode
 max length is 254 (way more than 6) - chosen so the length of the abbrev
 will fit in one byte.

 Returning NAME_MAX (which has nothing whatever to do with the issue)
 which is what we were doing is nonsense, it is more than libc (and hence
 zic) allows, when it is supposed to be less (or equal).   It is almost
 as if the proverbial someone believed that TZNAME_MAX has something to
 do with the file name (America/New_York) - for which NAME_MAX might be
 appropriate.

 There are two possible solutions, one is to treat sysconf(_SC_TZNAME_MAX)
 as it is defined, and make it the min value we'd ever support (say 31 or
 something like that) even though we actually support lots more than that
 (254), or we set it to be the value we actually support, rather than being
 the minimum allowed value.

 In some ways I prefer the former, it means the value will never need to
 change in the future (unless something very weird happens), whereas setting
 it to the actual supported value means every new tzcode release might alter
 what it gets set to (tzcode changed it from 255 to 254 not all that long
 ago, so the (unsigned)"-1" byte value could be used for other purposes).

 For now I am going to do the latter however, as that seems more consistent
 with how we typically set these "minimum allowed" sysconf values - making
 them represent what we currently allow, which is clearly both the minimum
 max value, and the max max value (for now).

   | I'm not sure where the 511 for _SC_TZNAME_MAX comes from;

 libc/gen/sysctl.c - the user.whatever sysctl() values are all embedded
 in there.   (sysconf() uses sysctl() to get the value - no actual system
 call is required for it).

   | where I think 'types' should be replaced by 'characters'.

 That typo exists in several places, I found several, but no idea if
 I got all that exist.

 kre


State-Changed-From-To: open->analyzed
State-Changed-By: kre@NetBSD.org
State-Changed-When: Fri, 01 May 2026 05:43:52 +0000
State-Changed-Why:
Issue understood, fix coming (undergoing testing)


From: "Robert Elz" <kre@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/60219 CVS commit: src
Date: Fri, 1 May 2026 07:19:46 +0000

 Module Name:	src
 Committed By:	kre
 Date:		Fri May  1 07:19:45 UTC 2026

 Modified Files:
 	src/include: time.h
 	src/lib/libc/gen: sysconf.3 sysctl.c
 	src/lib/libc/time: localtime.c
 	src/share/man/man7: sysctl.7

 Log Message:
 PR lib/60219 -- Fix sysconf(_SC_TZNAME_MAX)

 That value is supposed to be the minimum value allowed for
 the maximum length of a timezone abbreviation.   It cannot
 be something larger than is allowed for that (and NAME_MAX
 has nothing to do with it)

 It defines the max lengths allowed for the words in

 	TZ=Frankenstein-7Monster-6[transition rules]

 in old style POSIX TZ variable settings - the POSIX required
 minimum value is 6 (so "Frankenstein" would not fit in a minimalist
 POSIX implementation).

 For now, sync the values between libc/sysctl() and libc/localtime()
 via a new _TZNAME_MAXIMUM definition in <time.h> and use that for
 both purposes (as a possible default anyway for localtime() - there
 the default value used will always be at least 254, and building with
 -DTZNAME_MAXIMUM=N can override that value with anything larger than
 _TZNAME_MAXIMUM if desired).

 While here fix as many instances of the "max number of types" (which
 should be "max number of bytes") string in related contexts that I could
 find - there might perhaps be more.

 XXX pullup -11 (at least)


 To generate a diff of this commit:
 cvs rdiff -u -r1.56 -r1.57 src/include/time.h
 cvs rdiff -u -r1.57 -r1.58 src/lib/libc/gen/sysconf.3
 cvs rdiff -u -r1.39 -r1.40 src/lib/libc/gen/sysctl.c
 cvs rdiff -u -r1.153 -r1.154 src/lib/libc/time/localtime.c
 cvs rdiff -u -r1.171 -r1.172 src/share/man/man7/sysctl.7

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

State-Changed-From-To: analyzed->needs-pullups
State-Changed-By: kre@NetBSD.org
State-Changed-When: Fri, 01 May 2026 07:23:05 +0000
State-Changed-Why:
Needs a pullup to -11, this one probably isn't worth fixing in older releases.


From: Thomas Klausner <wiz@netbsd.org>
To: gnats-bugs@netbsd.org
Cc: kre@netbsd.org
Subject: Re: lib/60219: sysconf overpromises TZNAME_MAX
Date: Fri, 1 May 2026 12:57:20 +0200

 Thanks, Robert, for fixing this.
 I can confirm that the test now completes successfully.
  Thomas

State-Changed-From-To: needs-pullups->pending-pullups
State-Changed-By: kre@NetBSD.org
State-Changed-When: Fri, 01 May 2026 12:06:59 +0000
State-Changed-Why:
[pullup-11 #267]


From: "Martin Husemann" <martin@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/60219 CVS commit: [netbsd-11] src
Date: Thu, 7 May 2026 15:43:06 +0000

 Module Name:	src
 Committed By:	martin
 Date:		Thu May  7 15:43:06 UTC 2026

 Modified Files:
 	src/include [netbsd-11]: time.h
 	src/lib/libc/gen [netbsd-11]: sysconf.3 sysctl.c
 	src/lib/libc/time [netbsd-11]: localtime.c
 	src/share/man/man7 [netbsd-11]: sysctl.7

 Log Message:
 Pull up following revision(s) (requested by kre in ticket #267):

 	lib/libc/gen/sysctl.c: revision 1.40
 	lib/libc/gen/sysconf.3: revision 1.58
 	share/man/man7/sysctl.7: revision 1.172
 	lib/libc/time/localtime.c: revision 1.154
 	include/time.h: revision 1.57

 PR lib/60219 -- Fix sysconf(_SC_TZNAME_MAX)

 That value is supposed to be the minimum value allowed for
 the maximum length of a timezone abbreviation.   It cannot
 be something larger than is allowed for that (and NAME_MAX
 has nothing to do with it)

 It defines the max lengths allowed for the words in
 	TZ=Frankenstein-7Monster-6[transition rules]
 in old style POSIX TZ variable settings - the POSIX required
 minimum value is 6 (so "Frankenstein" would not fit in a minimalist
 POSIX implementation).

 For now, sync the values between libc/sysctl() and libc/localtime()
 via a new _TZNAME_MAXIMUM definition in <time.h> and use that for
 both purposes (as a possible default anyway for localtime() - there
 the default value used will always be at least 254, and building with
 - -DTZNAME_MAXIMUM=N can override that value with anything larger than
 _TZNAME_MAXIMUM if desired).

 While here fix as many instances of the "max number of types" (which
 should be "max number of bytes") string in related contexts that I could
 find - there might perhaps be more.


 To generate a diff of this commit:
 cvs rdiff -u -r1.56 -r1.56.2.1 src/include/time.h
 cvs rdiff -u -r1.57 -r1.57.4.1 src/lib/libc/gen/sysconf.3
 cvs rdiff -u -r1.38.10.1 -r1.38.10.2 src/lib/libc/gen/sysctl.c
 cvs rdiff -u -r1.147 -r1.147.2.1 src/lib/libc/time/localtime.c
 cvs rdiff -u -r1.167.4.1 -r1.167.4.2 src/share/man/man7/sysctl.7

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

State-Changed-From-To: pending-pullups->closed
State-Changed-By: kre@NetBSD.org
State-Changed-When: Thu, 07 May 2026 18:25:29 +0000
State-Changed-Why:
Pullups completed


>Unformatted:

Home	PR Database Search