GNU bug report logs -
#17774
AIX and lbracket ([) program - will not install on AIX using installp
Previous Next
Reported by: Michael Felt <mamfelt <at> gmail.com>
Date: Fri, 13 Jun 2014 14:21:02 UTC
Severity: normal
Tags: notabug
Done: Assaf Gordon <assafgordon <at> gmail.com>
Bug is archived. No further changes may be made.
To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 17774 in the body.
You can then email your comments to 17774 AT debbugs.gnu.org in the normal way.
Toggle the display of automated, internal messages from the tracker.
Report forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Fri, 13 Jun 2014 14:21:02 GMT)
Full text and
rfc822 format available.
Acknowledgement sent
to
Michael Felt <mamfelt <at> gmail.com>
:
New bug report received and forwarded. Copy sent to
bug-coreutils <at> gnu.org
.
(Fri, 13 Jun 2014 14:21:03 GMT)
Full text and
rfc822 format available.
Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
I know now why I am still using version 8-15
1. I never tried 8.16
2. since version 8.17 (or 16 which I never tried) lbracket gets linked as
the program [
Now, except that I would have a terrible time entering that as a command,
as is it a special character for ksh, and I would think bash the additional
problem with
[ as a commandname, or even as a character in a command name - AIX does not
permit files in an installp package are refused when they include certain
special characters (: and the square brackets I am sure of, there are
probably more).
So, while it is a valid filename - I, personally, see it as an extremely
non-portable command name.
If someone, after installation, would want to create a hardlink to that
name - great. But to have a name like that, I must be too old fashioned -
where is the win? typing
/usr/bin/\[ gets me what?
regards,
Michael
p.s. hints on how to massage Makefile is fine. Not trying to save the world
here :)
[Message part 2 (text/html, inline)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Fri, 13 Jun 2014 15:10:02 GMT)
Full text and
rfc822 format available.
Message #8 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
Michael Felt wrote:
> But to have a name like that, I must be too old fashioned -
> where is the win?
It's so that execlp ("FOO") acts like the shell command FOO, or, more
precisely, so that the attached C program works like '[ -d / ]' at the
shell level. POSIX requires that all standard utilities (except for a
very short list) must work the same way from a C program as from the
shell. See the last sentence of:
http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap01.html#tag_17_06
'[' is not on the list of exceptions, so coreutils arranges for it to be
an executable, as POSIX requires.
> AIX does not
> permit files in an installp package are refused when they include certain
> special characters
It may be simpler to just omit '[' from your installp package (I assume
that's some downstream thing). I doubt whether anybody but POSIX nerds
will care. AIX itself doesn't seem to be POSIX-conforming here, as the
attached C program fails on AIX. (If *you* are a POSIX nerd please feel
free to file a bug report with IBM....)
[t.c (text/x-csrc, attachment)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Sat, 14 Jun 2014 15:55:02 GMT)
Full text and
rfc822 format available.
Message #11 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
Yes, I suppose I could just delete it. However, my comment is "also" that
at 8.15 it packaged fine. Starting with 8.17 (or maybe 8.16, I can download
and try).
FYI - I can execute the program [ (as ./? --version as the shell does not
like [ entered directly, and I am not counting the backslashes correctly).
So, maybe your sample program would execute. I will try it in a day or so,
just checking mail for now.
On Fri, Jun 13, 2014 at 5:09 PM, Paul Eggert <eggert <at> cs.ucla.edu> wrote:
> Michael Felt wrote:
>
>> But to have a name like that, I must be too old fashioned -
>> where is the win?
>>
>
> It's so that execlp ("FOO") acts like the shell command FOO, or, more
> precisely, so that the attached C program works like '[ -d / ]' at the
> shell level. POSIX requires that all standard utilities (except for a very
> short list) must work the same way from a C program as from the shell. See
> the last sentence of:
>
> http://pubs.opengroup.org/onlinepubs/9699919799/
> utilities/V3_chap01.html#tag_17_06
>
> '[' is not on the list of exceptions, so coreutils arranges for it to be
> an executable, as POSIX requires.
>
>
> AIX does not
>> permit files in an installp package are refused when they include certain
>> special characters
>>
>
> It may be simpler to just omit '[' from your installp package (I assume
> that's some downstream thing). I doubt whether anybody but POSIX nerds
> will care. AIX itself doesn't seem to be POSIX-conforming here, as the
> attached C program fails on AIX. (If *you* are a POSIX nerd please feel
> free to file a bug report with IBM....)
>
[Message part 2 (text/html, inline)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Sat, 14 Jun 2014 16:09:02 GMT)
Full text and
rfc822 format available.
Message #14 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
I believe IBMis usually quite commited to being inline with published and
accepted standards. However, this is a standard with 2008 in it's name, and
AIX 5.3 is from 2004, and the TL/SP level I am compiling on, for backwards
compatibility is dated 2007 - so hard to complain that it is not up to a
2008 standard - although I expect the program you sent will compile and
execute.
It is the installp installer that has issues because it uses the []
characters for special purposes.
Maybe I can find a way/modification to the mkinstallp program so that
installp will accept the input.
And yes, deleting/not including it is probably the simplest solution.
On Fri, Jun 13, 2014 at 5:09 PM, Paul Eggert <eggert <at> cs.ucla.edu> wrote:
> Michael Felt wrote:
>
>> But to have a name like that, I must be too old fashioned -
>> where is the win?
>>
>
> It's so that execlp ("FOO") acts like the shell command FOO, or, more
> precisely, so that the attached C program works like '[ -d / ]' at the
> shell level. POSIX requires that all standard utilities (except for a very
> short list) must work the same way from a C program as from the shell. See
> the last sentence of:
>
> http://pubs.opengroup.org/onlinepubs/9699919799/
> utilities/V3_chap01.html#tag_17_06
>
> '[' is not on the list of exceptions, so coreutils arranges for it to be
> an executable, as POSIX requires.
>
>
> AIX does not
>> permit files in an installp package are refused when they include certain
>> special characters
>>
>
> It may be simpler to just omit '[' from your installp package (I assume
> that's some downstream thing). I doubt whether anybody but POSIX nerds
> will care. AIX itself doesn't seem to be POSIX-conforming here, as the
> attached C program fails on AIX. (If *you* are a POSIX nerd please feel
> free to file a bug report with IBM....)
>
[Message part 2 (text/html, inline)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Sat, 14 Jun 2014 17:29:01 GMT)
Full text and
rfc822 format available.
Message #17 received at 17774 <at> debbugs.gnu.org (full text, mbox):
Michael Felt wrote:
> this is a standard with 2008 in it's name, and
> AIX 5.3 is from 2004
AIX 5.3 is no longer supported, so I wouldn't bother filing a bug for
it. I was thinking about AIX 7.1, which has the same problem; if you
want to be a POSIX nerd you can file a bug report against it.
That part of POSIX has been standardized since POSIX.2 (IEEE Std
1003.2-1992), and the wording hasn't changed since then if I recall
correctly, so AIX has had this conformance bug for decades and nobody
has cared....
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Sun, 15 Jun 2014 10:53:01 GMT)
Full text and
rfc822 format available.
Message #20 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
then if nobody has cared up til now I doubt I will have much chance of
making a difference on that.
thx for the correction. I will mention it, but will not be holding my
breath.
Michael
On Jun 14, 2014 7:28 PM, "Paul Eggert" <eggert <at> cs.ucla.edu> wrote:
> Michael Felt wrote:
>
>> this is a standard with 2008 in it's name, and
>> AIX 5.3 is from 2004
>>
>
> AIX 5.3 is no longer supported, so I wouldn't bother filing a bug for it.
> I was thinking about AIX 7.1, which has the same problem; if you want to
> be a POSIX nerd you can file a bug report against it.
>
> That part of POSIX has been standardized since POSIX.2 (IEEE Std
> 1003.2-1992), and the wording hasn't changed since then if I recall
> correctly, so AIX has had this conformance bug for decades and nobody has
> cared....
>
[Message part 2 (text/html, inline)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Tue, 17 Jun 2014 20:51:01 GMT)
Full text and
rfc822 format available.
Message #23 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
FYI: the program runs fine, and even from the command line (the extra ] at
the end must satisify the ksh syntax checking).
Note: the 8.15 one is "suppossed" to fail, because I packaged that on AIX
6.1 - and then it does not work on AIX 5.3.
root <at> x093:[/data/prj/gnu/coreutils]find . -name ? -ls
85925819 4 drwxrwxr-x 7 michael 1954 4096 Dec 16 2013 .
85988590 126 -rwxr-xr-x 1 root system 128624 Jun 13 13:52
./coreutils-8.20/src/[
85986625 130 -rwxrwxr-- 1 michael 1954 132118 Nov 3 2012
./coreutils-8.17/src/[
85932793 111 -rwxrwxr-- 1 michael 1954 113168 Nov 3 2012
./coreutils-8.15/src/[
143394625 4 drwxr-xr-x 2 root system 4096 Jun 13 14:08
./coreutils-8.15/build/aix/x
120261651 127 -rwxr-xr-x 1 root system 129823 Jun 13 13:15
./coreutils-8.22/src/[
root <at> x093:[/data/prj/gnu/coreutils]./coreutils-8.15/src/[ -d / ]
Segmentation fault(coredump)
root <at> x093:[/data/prj/gnu/coreutils]./coreutils-8.17/src/[ -d / ]
root <at> x093:[/data/prj/gnu/coreutils]echo $?
0
root <at> x093:[/data/prj/gnu/coreutils]./coreutils-8.20/src/[ -d / ]
root <at> x093:[/data/prj/gnu/coreutils]echo $?
0
On an AIX 6.1 system they all work.
root <at> x094:[/data/prj/gnu/coreutils]find . -name ?
.
./coreutils-8.20/src/[
./coreutils-8.17/src/[
./coreutils-8.15/src/[
./coreutils-8.15/build/aix/x
./coreutils-8.22/src/[
root <at> x094:[/data/prj/gnu/coreutils]./coreutils-8.22/src/[ -d / ]
root <at> x094:[/data/prj/gnu/coreutils]./coreutils-8.22/src/[ -d / ]; echo $?
0
root <at> x094:[/data/prj/gnu/coreutils]./coreutils-8.20/src/[ -d / ]; echo $?
0
root <at> x094:[/data/prj/gnu/coreutils]./coreutils-8.17/src/[ -d / ]; echo $?
0
root <at> x094:[/data/prj/gnu/coreutils]./coreutils-8.15/src/[ -d / ]; echo $?
0
So, I am thinking, when I did 8.15 way back when, I must have customized it
to not include "lbracket" in the package - as, obviously, it was built.
p.s. - I did not try your t.c program, but I expect it to work fine.
On Sat, Jun 14, 2014 at 6:08 PM, Michael Felt <mamfelt <at> gmail.com> wrote:
> I believe IBMis usually quite commited to being inline with published and
> accepted standards. However, this is a standard with 2008 in it's name, and
> AIX 5.3 is from 2004, and the TL/SP level I am compiling on, for backwards
> compatibility is dated 2007 - so hard to complain that it is not up to a
> 2008 standard - although I expect the program you sent will compile and
> execute.
>
> It is the installp installer that has issues because it uses the []
> characters for special purposes.
>
> Maybe I can find a way/modification to the mkinstallp program so that
> installp will accept the input.
>
> And yes, deleting/not including it is probably the simplest solution.
>
>
> On Fri, Jun 13, 2014 at 5:09 PM, Paul Eggert <eggert <at> cs.ucla.edu> wrote:
>
>> Michael Felt wrote:
>>
>>> But to have a name like that, I must be too old fashioned -
>>> where is the win?
>>>
>>
>> It's so that execlp ("FOO") acts like the shell command FOO, or, more
>> precisely, so that the attached C program works like '[ -d / ]' at the
>> shell level. POSIX requires that all standard utilities (except for a very
>> short list) must work the same way from a C program as from the shell. See
>> the last sentence of:
>>
>> http://pubs.opengroup.org/onlinepubs/9699919799/
>> utilities/V3_chap01.html#tag_17_06
>>
>> '[' is not on the list of exceptions, so coreutils arranges for it to be
>> an executable, as POSIX requires.
>>
>>
>> AIX does not
>>> permit files in an installp package are refused when they include certain
>>> special characters
>>>
>>
>> It may be simpler to just omit '[' from your installp package (I assume
>> that's some downstream thing). I doubt whether anybody but POSIX nerds
>> will care. AIX itself doesn't seem to be POSIX-conforming here, as the
>> attached C program fails on AIX. (If *you* are a POSIX nerd please feel
>> free to file a bug report with IBM....)
>>
>
>
[Message part 2 (text/html, inline)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Tue, 17 Jun 2014 21:27:02 GMT)
Full text and
rfc822 format available.
Message #26 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
On 06/17/2014 02:50 PM, Michael Felt wrote:
> FYI: the program runs fine, and even from the command line (the extra ] at
> the end must satisify the ksh syntax checking).
Rather, the 'test' binary and the 'lbracket' binary differ in one
crucial aspect: 'lbracket' requires its last argument in argv[] to be
"]". It is not ksh, but lbracket itself that requires the matching ],
and it is this fact that allows us to obey POSIX while still allowing '[
--help' to produce help text, even though 'test --help' cannot do so
(that is, '[' has an escape hatch that 'test' does not).
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
[signature.asc (application/pgp-signature, attachment)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Wed, 18 Jun 2014 06:24:01 GMT)
Full text and
rfc822 format available.
Message #29 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
Well, you guys are the experts. I was trying to be "smart" - thinking that
lbracket 'required' the closing right bracket to keep the shell syntax
checkers happy. Maybe I am expecting too much from my shells need to check
syntax.
FYI - It seems to be working as expected, rather designed - so I shall make
some specialized 'buildaix' components to get around the filenameing
problem.
Will keep you posted.
On Tue, Jun 17, 2014 at 11:16 PM, Eric Blake <eblake <at> redhat.com> wrote:
> On 06/17/2014 02:50 PM, Michael Felt wrote:
> > FYI: the program runs fine, and even from the command line (the extra ]
> at
> > the end must satisify the ksh syntax checking).
>
> Rather, the 'test' binary and the 'lbracket' binary differ in one
> crucial aspect: 'lbracket' requires its last argument in argv[] to be
> "]". It is not ksh, but lbracket itself that requires the matching ],
> and it is this fact that allows us to obey POSIX while still allowing '[
> --help' to produce help text, even though 'test --help' cannot do so
> (that is, '[' has an escape hatch that 'test' does not).
>
> --
> Eric Blake eblake redhat com +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
>
>
[Message part 2 (text/html, inline)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Wed, 18 Jun 2014 18:31:01 GMT)
Full text and
rfc822 format available.
Message #32 received at 17774 <at> debbugs.gnu.org (full text, mbox):
[Message part 1 (text/plain, inline)]
revisted my 8.15 package in detail. Seems that two years ago I had already
done something special for coreutils - so my apology - not a bug, just
something in the way. I will find a way to make it transparent for installp.
My question: once done, would you be interested in what I have done to
include either in an AIX INSTALL or README?
On Wed, Jun 18, 2014 at 8:23 AM, Michael Felt <mamfelt <at> gmail.com> wrote:
> Well, you guys are the experts. I was trying to be "smart" - thinking that
> lbracket 'required' the closing right bracket to keep the shell syntax
> checkers happy. Maybe I am expecting too much from my shells need to check
> syntax.
>
> FYI - It seems to be working as expected, rather designed - so I shall
> make some specialized 'buildaix' components to get around the filenameing
> problem.
>
> Will keep you posted.
>
>
> On Tue, Jun 17, 2014 at 11:16 PM, Eric Blake <eblake <at> redhat.com> wrote:
>
>> On 06/17/2014 02:50 PM, Michael Felt wrote:
>> > FYI: the program runs fine, and even from the command line (the extra ]
>> at
>> > the end must satisify the ksh syntax checking).
>>
>> Rather, the 'test' binary and the 'lbracket' binary differ in one
>> crucial aspect: 'lbracket' requires its last argument in argv[] to be
>> "]". It is not ksh, but lbracket itself that requires the matching ],
>> and it is this fact that allows us to obey POSIX while still allowing '[
>> --help' to produce help text, even though 'test --help' cannot do so
>> (that is, '[' has an escape hatch that 'test' does not).
>>
>> --
>> Eric Blake eblake redhat com +1-919-301-3266
>> Libvirt virtualization library http://libvirt.org
>>
>>
>
[Message part 2 (text/html, inline)]
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Sat, 20 Oct 2018 03:36:02 GMT)
Full text and
rfc822 format available.
Message #35 received at 17774 <at> debbugs.gnu.org (full text, mbox):
tags 17774 notabug
close 17774
stop
(triaging old bugs)
On 18/06/14 12:30 PM, Michael Felt wrote:
> revisted my 8.15 package in detail. Seems that two years ago I had already
> done something special for coreutils - so my apology - not a bug, just
> something in the way. I will find a way to make it transparent for installp.
>
Confirmed not to be a bug - so closing.
-assaf
Added tag(s) notabug.
Request was from
Assaf Gordon <assafgordon <at> gmail.com>
to
control <at> debbugs.gnu.org
.
(Sat, 20 Oct 2018 03:36:02 GMT)
Full text and
rfc822 format available.
bug closed, send any further explanations to
17774 <at> debbugs.gnu.org and Michael Felt <mamfelt <at> gmail.com>
Request was from
Assaf Gordon <assafgordon <at> gmail.com>
to
control <at> debbugs.gnu.org
.
(Sat, 20 Oct 2018 03:36:02 GMT)
Full text and
rfc822 format available.
bug archived.
Request was from
Debbugs Internal Request <help-debbugs <at> gnu.org>
to
internal_control <at> debbugs.gnu.org
.
(Sat, 17 Nov 2018 12:24:07 GMT)
Full text and
rfc822 format available.
bug unarchived.
Request was from
Paul Eggert <eggert <at> cs.ucla.edu>
to
control <at> debbugs.gnu.org
.
(Fri, 31 Dec 2021 20:41:01 GMT)
Full text and
rfc822 format available.
Information forwarded
to
bug-coreutils <at> gnu.org
:
bug#17774
; Package
coreutils
.
(Fri, 31 Dec 2021 20:55:02 GMT)
Full text and
rfc822 format available.
Message #46 received at 17774 <at> debbugs.gnu.org (full text, mbox):
On 6/14/14 10:28, Paul Eggert wrote:
> That part of POSIX has been standardized since POSIX.2 (IEEE Std
> 1003.2-1992), and the wording hasn't changed since then if I recall
> correctly, so AIX has had this conformance bug for decades and nobody
> has cared....
I just ran into a related problem when building bleeding-edge coreutils
on AIX 7.1, and so I thought I'd document it here. On AIX 7.1, 'make
check' fails with:
FAIL: tests/misc/help-version.sh
because 'make' never built a '[' program. And AIX 'make' doesn't build a
'[' program because it mishandles '[' in 'make' macros, which means that
with Makefiles like this:
EXEEXT =
bin_PROGRAMS = ... src/[$(EXEEXT) ...
PROGRAMS = $(bin_PROGRAMS) ...
all-am: ... $(PROGRAMS) ...
'make all-am' does not "see" the 'src/[' and so doesn't build it.
Because of this bug, 'make install' obviously will not work correctly,
as there's no '[' command to install.
A simple workaround is to use GNU Make, which doesn't have this bug.
This bug in AIX 'make' is so obscure that I'm not going to bother
documenting it in the Autoconf manual under its portability guidelines
for 'make'. Anyway, nowadays "just use GNU Make" is a good recipe for
just about every package other than GNU Make itself.
bug archived.
Request was from
Debbugs Internal Request <help-debbugs <at> gnu.org>
to
internal_control <at> debbugs.gnu.org
.
(Sat, 29 Jan 2022 12:24:06 GMT)
Full text and
rfc822 format available.
This bug report was last modified 3 years and 199 days ago.
Previous Next
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.