Mailing List Archive

op/groups......FAILED on test 1
perl5.001n
SunOS 4.1.3

Is this anything to worry about?

--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
> perl5.001n
> SunOS 4.1.3

> Is this anything to worry about?

i'm temporarily unconnected so can't look at source, but i'll
bet that either you have an oddity between /etc/passwd and /etc/group,
or else you gave the wrong answer to the size of the groups returned
(git_t vs int).

when i ran ./configure by itself on that system, it worked ok.

--5tom
Re: op/groups......FAILED on test 1 [ In reply to ]
On Tue, 7 Nov 1995, Jim Anderson wrote:

> >> perl5.001n SunOS 4.1.3
>
> >> Is this anything to worry about?

It shouldn't happen. SunOS 4.1.3 is one of the places I tested
5.001n before releasing it.

> Tom> groups returned (git_t vs int).

The sunos_4 hints file assumes you're *not* using SysV stuff (which is
documented to take gid_t * as the groupstype). I haven't tested
whether this actually makes a difference or not.

> I ran Configure with -des. I didn't have this problem with 1m,
> although I installed it in it (almost) in exactly the same way. I said
> 'almost' 'cause I remember using ony -d as the option. Might this
> account for the difference?

No. -des vs. -d won't change this at all.
However, re-using your old config.sh might matter (depending on
what you did with the first -d).

Just to be sure what you've done, I'd recommend
make -k distclean
rm -f config.sh
sh Configure -des

Andy Dougherty doughera@lafcol.lafayette.edu
Re: op/groups......FAILED on test 1 [ In reply to ]
>>>>> "Tom" == Tom Christiansen <tchrist@alumni.cs.colorado.edu> writes:

>> perl5.001n SunOS 4.1.3

>> Is this anything to worry about?

Tom> i'm temporarily unconnected so can't look at source, but i'll bet
Tom> that either you have an oddity between /etc/passwd and
Tom> /etc/group, or else you gave the wrong answer to the size of the
Tom> groups returned (git_t vs int).

Tom> when i ran ./configure by itself on that system, it worked ok.

Tom> --5tom

I ran Configure with -des. I didn't have this problem with 1m,
although I installed it in it (almost) in exactly the same way. I said
'almost' 'cause I remember using ony -d as the option. Might this
account for the difference?

jim

--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
>
> On Tue, 7 Nov 1995, Jim Anderson wrote:
> > >> perl5.001n SunOS 4.1.3
> >
> > >> Is this anything to worry about?
>
> It shouldn't happen. SunOS 4.1.3 is one of the places I tested
> 5.001n before releasing it.
>
> > Tom> groups returned (git_t vs int).
>
> The sunos_4 hints file assumes you're *not* using SysV stuff (which is
> documented to take gid_t * as the groupstype). I haven't tested
> whether this actually makes a difference or not.

I just ran it specifying only -d, and got the same results. But I did
notice that gid_t was selected.

> > I ran Configure with -des. I didn't have this problem with 1m,
> > although I installed it in it (almost) in exactly the same way. I said
> > 'almost' 'cause I remember using ony -d as the option. Might this
> > account for the difference?
>
> No. -des vs. -d won't change this at all.
> However, re-using your old config.sh might matter (depending on
> what you did with the first -d).
>
> Just to be sure what you've done, I'd recommend
> make -k distclean
> rm -f config.sh
> sh Configure -des

I'll try this just to be sure, but I did _not_ reuse my config.sh from 1m.


--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
On Nov 7, 11:17am, Tom Christiansen wrote:
} Subject: Re: op/groups......FAILED on test 1
} > perl5.001n
} > SunOS 4.1.3
}
} > Is this anything to worry about?
}
} i'm temporarily unconnected so can't look at source, but i'll
} bet that either you have an oddity between /etc/passwd and /etc/group,
} or else you gave the wrong answer to the size of the groups returned
} (git_t vs int).

You might see this probelm if you're using gcc. The fixincludes program
gets the type of the grouplist passed to getgroups() wrong. It turns
out that the /usr/lib version of getgroups() wants an int array, and
the /usr/5lib version wants a gid_t array. The <unistd.h> created by
fixincludes uses (unsigned short *) in the prototype.

--- Truck
Re: op/groups......FAILED on test 1 [ In reply to ]
On Tue, 7 Nov 1995, Don Lewis wrote:

> On Nov 7, 11:17am, Tom Christiansen wrote:
> } Subject: Re: op/groups......FAILED on test 1
> } > perl5.001n
> } > SunOS 4.1.3
> }
> } > Is this anything to worry about?

The problem was indeed gid_t * vs. int * for groupstype. Configure reads
the <unistd.h> file and guesses gid_t, but the sunos_4_1.sh hint file
overrides that and tries to force the "correct" answer by guessing whether
you are using a SysV-ish universe or BSD-ish universe, based on the error
output of cat -b.

The SysV-ish part sets groupstype=gid_t. Apparently we need a better way
to check if the user will link against /usr/5lib vs. /usr/lib.

Volunteers?

Andy Dougherty doughera@lafcol.lafayette.edu
Re: op/groups......FAILED on test 1 [ In reply to ]
> On Nov 7, 11:17am, Tom Christiansen wrote:
> } Subject: Re: op/groups......FAILED on test 1
> } > perl5.001n
> } > SunOS 4.1.3
> }
> } > Is this anything to worry about?
> }
> } i'm temporarily unconnected so can't look at source, but i'll
> } bet that either you have an oddity between /etc/passwd and /etc/group,
> } or else you gave the wrong answer to the size of the groups returned
> } (git_t vs int).
>
> You might see this probelm if you're using gcc. The fixincludes program
> gets the type of the grouplist passed to getgroups() wrong. It turns
> out that the /usr/lib version of getgroups() wants an int array, and
> the /usr/5lib version wants a gid_t array. The <unistd.h> created by
> fixincludes uses (unsigned short *) in the prototype.

I'm not running gcc. Turns out the problem was that the hints file was
incorrect.

Thanks for responding.

jim

--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
Guess I should have read your reply more closely the first time :-) I
think you're right on the money. Thanks again.

jim

On Nov 7, 11:17am, Tom Christiansen wrote:
} Subject: Re: op/groups......FAILED on test 1
} > perl5.001n
} > SunOS 4.1.3
}
} > Is this anything to worry about?
}
} i'm temporarily unconnected so can't look at source, but i'll
} bet that either you have an oddity between /etc/passwd and /etc/group,
} or else you gave the wrong answer to the size of the groups returned
} (git_t vs int).

You might see this probelm if you're using gcc. The fixincludes program
gets the type of the grouplist passed to getgroups() wrong. It turns
out that the /usr/lib version of getgroups() wants an int array, and
the /usr/5lib version wants a gid_t array. The <unistd.h> created by
fixincludes uses (unsigned short *) in the prototype.

--- Truck



--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
In <9511071810.AA06755@gusvnm1.lehman.com>
On Tue, 7 Nov 95 13:10:12 EST
Jim Anderson <jander@lehman.com> writes:
>perl5.001n
>SunOS 4.1.3
>
>Is this anything to worry about?

Passed here.
Re: op/groups......FAILED on test 1 [ In reply to ]
In <Pine.3.89.9511071729.B13923-0100000@lafcol>
On Tue, 7 Nov 1995 17:36:38 -0500 (EST)
Andy Dougherty <doughera@lafcol.lafayette.edu> writes:
>
>The SysV-ish part sets groupstype=gid_t. Apparently we need a better way
>to check if the user will link against /usr/5lib vs. /usr/lib.

IMHO using /usr/5lib on SunOS4 is nearly as risky as /usr/ucb on Solaris2.
Re: op/groups......FAILED on test 1 [ In reply to ]
>>>>> "Nick" == Nick Ing-Simmons <nik@tiuk.ti.com> writes:

>> perl5.001n SunOS 4.1.3
>>
>> Is this anything to worry about?


Thanks for responding. I got this yesterday:

============================================================
Return-Path: <doughera@lafcol.lafayette.edu>
Date: Tue, 7 Nov 1995 17:36:38 -0500 (EST)
From: Andy Dougherty <doughera@lafcol.lafayette.edu>
Subject: Re: op/groups......FAILED on test 1
To: Don Lewis <gdonl@gv.ssi1.com>
Cc: tchrist@cs.colorado.edu, Jim Anderson <jander@lehman.com>,
perl5-porters@africa.nicoh.com
In-Reply-To: <199511072217.OAA06006@salsa.gv.ssi1.com>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Tue, 7 Nov 1995, Don Lewis wrote:

> On Nov 7, 11:17am, Tom Christiansen wrote:
> } Subject: Re: op/groups......FAILED on test 1
> } > perl5.001n
> } > SunOS 4.1.3
> }
> } > Is this anything to worry about?

The problem was indeed gid_t * vs. int * for groupstype. Configure reads
the <unistd.h> file and guesses gid_t, but the sunos_4_1.sh hint file
overrides that and tries to force the "correct" answer by guessing whether
you are using a SysV-ish universe or BSD-ish universe, based on the error
output of cat -b.

The SysV-ish part sets groupstype=gid_t. Apparently we need a better way
to check if the user will link against /usr/5lib vs. /usr/lib.

Volunteers?

Andy Dougherty doughera@lafcol.lafayette.edu



--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
> From: Nick Ing-Simmons <nik@tiuk.ti.com>
>
> In <Pine.3.89.9511071729.B13923-0100000@lafcol>
> On Tue, 7 Nov 1995 17:36:38 -0500 (EST)
> Andy Dougherty <doughera@lafcol.lafayette.edu> writes:
> >
> >The SysV-ish part sets groupstype=gid_t. Apparently we need a better way
> >to check if the user will link against /usr/5lib vs. /usr/lib.
>
> IMHO using /usr/5lib on SunOS4 is nearly as risky as /usr/ucb on Solaris2.

I've used it, and the SunOS4 SysV environment in general, for years with no
significant problems.

It certainly doesn't deserve to be compared to /usr/ucb on Solaris2.

This problem with perl in the SysV environment is only a problem if you
use GCC since it doesn't understand that it should automatically use
/usr/5lib if /usr/5bin comes before /usr/bin on the path. Building perl
with cc in the SunOS4 SysV environment works fine.

If you scratch the surface you'll notice that Configure has no way to
search /usr/5include _and_then_ /usr/include when looking for header files.
In practice this does not seem to be a significant problem.

It would be nice to resolve the group types issues for GCC in the SunOS4
SysV environment but it's not a high priority for me.

Tim.
Re: op/groups......FAILED on test 1 [ In reply to ]
On Thu, 9 Nov 1995, Tim Bunce wrote:

> If you scratch the surface you'll notice that Configure has no way to
> search /usr/5include _and_then_ /usr/include when looking for header files.

There's always a way with Configure :-).
> In practice this does not seem to be a significant problem.

True. Especially since it's picking up the gid_t from /usr/include/unistd.h
anyway.

In fact, using the SysV environment might help fix those _tzname[]
problems in POSIX.c.

> It would be nice to resolve the group types issues for GCC in the SunOS4
> SysV environment but it's not a high priority for me.

Nor me. However, if someone else does it, I won't object.

Andy Dougherty doughera@lafcol.lafayette.edu
Re: op/groups......FAILED on test 1 [ In reply to ]
> From: jander@lehman.com (Jim Anderson)
>
> >>>>> "Tim" == Tim Bunce <Tim.Bunce@ig.co.uk> writes:
>
> >> From: Nick Ing-Simmons <nik@tiuk.ti.com>
> >>
> >> In <Pine.3.89.9511071729.B13923-0100000@lafcol> On Tue, 7 Nov 1995
> >> 17:36:38 -0500 (EST) Andy Dougherty <doughera@lafcol.lafayette.edu>
> >> writes: > >The SysV-ish part sets groupstype=gid_t. Apparently we
> >> need a better way >to check if the user will link against /usr/5lib
> >> vs. /usr/lib.
> >>
> >> IMHO using /usr/5lib on SunOS4 is nearly as risky as /usr/ucb on
> >> Solaris2.
>
> Tim> I've used it, and the SunOS4 SysV environment in general, for
> Tim> years with no significant problems.
>
> Tim> It certainly doesn't deserve to be compared to /usr/ucb on
> Tim> Solaris2.
>
> Tim> This problem with perl in the SysV environment is only a problem
> Tim> if you use GCC since it doesn't understand that it should
> Tim> automatically use /usr/5lib if /usr/5bin comes before /usr/bin on
> Tim> the path. Building perl with cc in the SunOS4 SysV environment
> Tim> works fine.
>
> Tim> If you scratch the surface you'll notice that Configure has no
> Tim> way to search /usr/5include _and_then_ /usr/include when looking
> Tim> for header files. In practice this does not seem to be a
> Tim> significant problem.
>
> Tim> It would be nice to resolve the group types issues for GCC in the
> Tim> SunOS4 SysV environment but it's not a high priority for me.
>
> I was the one who reported this problem, and I was using the Sun
> compiler, _not_ gcc.

I've just tried and I don't get any problems using the default SunOS 4.1.3
cc. It works fine.

make clobber && Configure -des && make && make test => All tests successful.

$ ldd perl
-ldl.1 => /usr/lib/libdl.so.1.0
-lc.2 => /usr/5lib/libc.so.2.8

Tim.
Re: op/groups......FAILED on test 1 [ In reply to ]
>>>>> "Tim" == Tim Bunce <Tim.Bunce@ig.co.uk> writes:

>> From: Nick Ing-Simmons <nik@tiuk.ti.com>
>>
>> In <Pine.3.89.9511071729.B13923-0100000@lafcol> On Tue, 7 Nov 1995
>> 17:36:38 -0500 (EST) Andy Dougherty <doughera@lafcol.lafayette.edu>
>> writes: > >The SysV-ish part sets groupstype=gid_t. Apparently we
>> need a better way >to check if the user will link against /usr/5lib
>> vs. /usr/lib.
>>
>> IMHO using /usr/5lib on SunOS4 is nearly as risky as /usr/ucb on
>> Solaris2.

Tim> I've used it, and the SunOS4 SysV environment in general, for
Tim> years with no significant problems.

Tim> It certainly doesn't deserve to be compared to /usr/ucb on
Tim> Solaris2.

Tim> This problem with perl in the SysV environment is only a problem
Tim> if you use GCC since it doesn't understand that it should
Tim> automatically use /usr/5lib if /usr/5bin comes before /usr/bin on
Tim> the path. Building perl with cc in the SunOS4 SysV environment
Tim> works fine.

Tim> If you scratch the surface you'll notice that Configure has no
Tim> way to search /usr/5include _and_then_ /usr/include when looking
Tim> for header files. In practice this does not seem to be a
Tim> significant problem.

Tim> It would be nice to resolve the group types issues for GCC in the
Tim> SunOS4 SysV environment but it's not a high priority for me.

I was the one who reported this problem, and I was using the Sun
compiler, _not_ gcc.

jim

--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
On Nov 9, 1:08pm, Tim Bunce wrote:
} Subject: Re: op/groups......FAILED on test 1
} This problem with perl in the SysV environment is only a problem if you
} use GCC since it doesn't understand that it should automatically use
} /usr/5lib if /usr/5bin comes before /usr/bin on the path.

It's worse than that. You would need to run fixinclude on both the
both /usr/include and /usr/5include and gcc would have to use the
correct set of fixed include files. Also, /usr/5include doesn't
have unistd.h, and it's contents should differ between the two
environments, since the type of the group list returned by getgroups()
differs between the two.

--- Truck
Re: op/groups......FAILED on test 1 [ In reply to ]
> > From: jander@lehman.com (Jim Anderson)
> >
> > >>>>> "Tim" == Tim Bunce <Tim.Bunce@ig.co.uk> writes:
> >
> > >> From: Nick Ing-Simmons <nik@tiuk.ti.com>
> > >>
> > >> In <Pine.3.89.9511071729.B13923-0100000@lafcol> On Tue, 7 Nov 1995
> > >> 17:36:38 -0500 (EST) Andy Dougherty <doughera@lafcol.lafayette.edu>
> > >> writes: > >The SysV-ish part sets groupstype=gid_t. Apparently we
> > >> need a better way >to check if the user will link against /usr/5lib
> > >> vs. /usr/lib.
> > >>
> > >> IMHO using /usr/5lib on SunOS4 is nearly as risky as /usr/ucb on
> > >> Solaris2.
> >
> > Tim> I've used it, and the SunOS4 SysV environment in general, for
> > Tim> years with no significant problems.
> >
> > Tim> It certainly doesn't deserve to be compared to /usr/ucb on
> > Tim> Solaris2.
> >
> > Tim> This problem with perl in the SysV environment is only a problem
> > Tim> if you use GCC since it doesn't understand that it should
> > Tim> automatically use /usr/5lib if /usr/5bin comes before /usr/bin on
> > Tim> the path. Building perl with cc in the SunOS4 SysV environment
> > Tim> works fine.
> >
> > Tim> If you scratch the surface you'll notice that Configure has no
> > Tim> way to search /usr/5include _and_then_ /usr/include when looking
> > Tim> for header files. In practice this does not seem to be a
> > Tim> significant problem.
> >
> > Tim> It would be nice to resolve the group types issues for GCC in the
> > Tim> SunOS4 SysV environment but it's not a high priority for me.
> >
> > I was the one who reported this problem, and I was using the Sun
> > compiler, _not_ gcc.
>
> I've just tried and I don't get any problems using the default SunOS 4.1.3
> cc. It works fine.
>
> make clobber && Configure -des && make && make test => All tests successful.
>
> $ ldd perl
> -ldl.1 => /usr/lib/libdl.so.1.0
> -lc.2 => /usr/5lib/libc.so.2.8

ldd /home/technoids/bin/perl5
-ldl.1 => /usr/lib/libdl.so.1.0
-lc.1 => /usr/lib/libc.so.1.8

Might this account for it?

--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com
Re: op/groups......FAILED on test 1 [ In reply to ]
> > From: jander@lehman.com (Jim Anderson)
> >
> > >>>>> "Tim" == Tim Bunce <Tim.Bunce@ig.co.uk> writes:
> >
> > >> From: Nick Ing-Simmons <nik@tiuk.ti.com>
> > >>
> > >> In <Pine.3.89.9511071729.B13923-0100000@lafcol> On Tue, 7 Nov 1995
> > >> 17:36:38 -0500 (EST) Andy Dougherty <doughera@lafcol.lafayette.edu>
> > >> writes: > >The SysV-ish part sets groupstype=gid_t. Apparently we
> > >> need a better way >to check if the user will link against /usr/5lib
> > >> vs. /usr/lib.
> > >>
> > >> IMHO using /usr/5lib on SunOS4 is nearly as risky as /usr/ucb on
> > >> Solaris2.
> >
> > Tim> I've used it, and the SunOS4 SysV environment in general, for
> > Tim> years with no significant problems.
> >
> > Tim> It certainly doesn't deserve to be compared to /usr/ucb on
> > Tim> Solaris2.
> >
> > Tim> This problem with perl in the SysV environment is only a problem
> > Tim> if you use GCC since it doesn't understand that it should
> > Tim> automatically use /usr/5lib if /usr/5bin comes before /usr/bin on
> > Tim> the path. Building perl with cc in the SunOS4 SysV environment
> > Tim> works fine.
> >
> > Tim> If you scratch the surface you'll notice that Configure has no
> > Tim> way to search /usr/5include _and_then_ /usr/include when looking
> > Tim> for header files. In practice this does not seem to be a
> > Tim> significant problem.
> >
> > Tim> It would be nice to resolve the group types issues for GCC in the
> > Tim> SunOS4 SysV environment but it's not a high priority for me.
> >
> > I was the one who reported this problem, and I was using the Sun
> > compiler, _not_ gcc.
>
> I've just tried and I don't get any problems using the default SunOS 4.1.3
> cc. It works fine.
>
> make clobber && Configure -des && make && make test => All tests successful.
>
> $ ldd perl
> -ldl.1 => /usr/lib/libdl.so.1.0
> -lc.2 => /usr/5lib/libc.so.2.8

ldd /home/technoids/bin/perl5
-ldl.1 => /usr/lib/libdl.so.1.0
-lc.1 => /usr/lib/libc.so.1.8

Might this account for it?

--
Jim Anderson Phone: (201)524-4076
Lehman Brothers, Inc. Fax: (201)524-5153
101 Hudson Street, 34th Floor E-mail: jander@lehman.com
Jersey City, NJ 07302 jander@jander.com