headers
Lisa Seelye | 1 Nov 2003 03:47
Picon
Favicon

Re: locking user accounts doesn't really lock them.

On Fri, 2003-10-31 at 17:18, Kurt Lieber wrote:
> As was just pointed out to me on irc, expiring the account, as opposed to
> locking the password, will do what I want.  So:
> 
> usermod -e 0000-00-00 <user>

Will this leave the /etc/passwd entry intact?
--

-- 
Regards,
-Lisa
<Vix ulla tam iniqua pax, quin bello vel aequissimo sit potior>
Permalink | Reply | 0 comments |
headers
Norberto Bensa | 1 Nov 2003 07:43
Picon

Re: glibc-2.3.2-r[68] libpthread -> segmentation fault

> On Fri, 2003-10-31 at 03:22, Norberto Bensa wrote:
> > I've tried swaping the microprocessor: same message. Seems compiled wrong
> > by the box. I'll dedicate a Pentium MMX box to update glibc and see how
> > it goes.

Martin,

I've tried something better. My firewall runs on a Pentium MMX box. I've 
tarred the filesystem and moved it to my main box, a Pentium III [1]

Once on the P3 box, I chrooted and then I've 'emerge -u glibc' This is the 
result:

	# ls
	Segmentation fault

So the problem are not K6. The problem, IMHO, is gcc. CFLAGS on the firewall 
(P-MMX) are these:

	CFLAGS="-march=pentium -Os -ffast-math -fomit-frame-pointer -pipe"

"-march=pentium" is what I use on the K6 box where I first experienced this 
problem. I'll try -march=pentium{2,3} tomorrow and keep you informed if you 
want.

Regards,
Norberto

[1] the Pentium III box runs glibc-2.3.2-r8.
(Continue reading)

Permalink | Reply | 11 comments |
headers
Robin H. Johnson | 1 Nov 2003 07:50
Picon
Favicon
Gravatar

Re: glibc-2.3.2-r[68] libpthread -> segmentation fault

On Sat, Nov 01, 2003 at 03:43:45AM -0300, Norberto Bensa wrote:
> 	CFLAGS="-march=pentium -Os -ffast-math -fomit-frame-pointer -pipe"
do NOT use -ffast-math, it's asking for trouble (look it up in 'man gcc'
for the reason).

--

-- 
Robin Hugh Johnson
E-Mail     : robbat2 <at> orbis-terrarum.net
Home Page  : http://www.orbis-terrarum.net/?l=people.robbat2
ICQ#       : 30269588 or 41961639
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85
Permalink | Reply | 2 comments |
headers
Martin Lesser | 1 Nov 2003 09:52
X-Face
Picon

Re: vpopmail's emerge directory structure

"Robin H. Johnson" <robbat2 <at> gentoo.org> writes:

> The file doesn't belong to vpopmail exclusively. It really belongs to
> qmail, and vpopmail wants to add and remove items from it for it's
> misguiding implementation of relaying.

What do you mean with misguiding? vpopmail - like others - only tries to
record the REMOTEIP for SMTP after POP purposes.

> Qmail looks at /etc/tcp.smtp via tcpserver, which only allows a single
> file to be specified, so there is also a tcp.smtp is in
> /var/vpopmail/etc, then qmail NEVER looks at it, as it really needs
> /etc/tcp.smtp.

Just for clarifying: tcpserver (and not qmail) looks into a cdb-file
which you define as option for tcpserver with -x /path/to/file.cdb. In
case of qmail-smtpd tcpserver sets additional environment vars if
REMOTEIP is found in the cdb-file - e.g. RELAYCLIENT, on which
qmail-smtpd decides whether the remote-client may send mail for
non-local domains. You may run qmail-smtpd without any cdb-file if you
don't want to relay any mails.

tcpserver (package ucspi-tcp) at least is "only" a reliable and stable
replacement for [x]inetd with extended possibilities so one may run
qmail-smtpd also under control of [x]inetd but this is really not the
recommended way.

So if one uses vpopmail the run-file for qmail-smtpd could be changed in
a way that tcpserver looks in another cdb-file for which vpopmail has
write access. The vanilla vpopmail suggests this IIRC.
(Continue reading)

Permalink | Reply | 3 comments |
headers
Eldad Zack | 1 Nov 2003 12:50

Re: locking user accounts doesn't really lock them.


On Fri, 31 Oct 2003, Kurt Lieber wrote:

> On Fri, Oct 31, 2003 at 01:55:13PM -0800 or thereabouts, Kevyn Shortell wrote:
> > It's often overlooked but a much easier method for locking a user out is
> > simply to change their default shell to /bin/false or something like it.
> > SSH keys or not, they won't be getting access to the box anytime soon
> > without a default shell.
> 
> A valid point, but iirc, this still allows the user to do things which
> don't require an interactive shell.  (scp, for instance)  

I don't think that is the case - actually, I've managed to break scp by 
changing bashrc output.

scp does require the user to have a valid shell.

--
gentoo-dev <at> gentoo.org mailing list
Permalink | Reply | 0 comments |
headers
Mamoru KOMACHI | 1 Nov 2003 13:06
Picon
Favicon

Re: better handling of multibyte characters (nls/cjk/unicode)

Hi,

At Fri, 31 Oct 2003 16:05:30 +0100,
Spider wrote:

> > 1. Create global USE flag named "unicode" which enables
> >    Unicode(including UTF-7/UTF-8/UTF-16 and so on) support.
> > 2. Don't make it default in any profiles. (We will consider it later
> >    when those applications settle down.)

> Sounds great.

I've just added unicode USE flag :)

> proposed addition to baselayout:
> 
> if [ `use unicode` ] ;
> then
> cat >> ${D}/etc/rc.conf << EOF
> # This setting enables a default UTF-8 locale for your system. 
> # please look inside /usr/share/locale for more examples
> LANG="en_US.UTF-8"
> 
> EOF
> fi
> 
> This is rough, but something like it perhaps?

If http://bugs.gentoo.org/show_bug.cgi?id=9988 "setting LANG variable"
is implemented we may think about it, but I don't think the system
(Continue reading)

Permalink | Reply | 1 comment |
headers
Robin H. Johnson | 1 Nov 2003 13:15
Picon
Favicon
Gravatar

Re: vpopmail's emerge directory structure

On Sat, Nov 01, 2003 at 09:52:41AM +0100, Martin Lesser wrote:
> > The file doesn't belong to vpopmail exclusively. It really belongs to
> > qmail, and vpopmail wants to add and remove items from it for it's
> > misguiding implementation of relaying.
> What do you mean with misguiding? vpopmail - like others - only tries to
> record the REMOTEIP for SMTP after POP purposes.
I personally believe that /etc/tcp.smtp should not be writable by
anybody other than the root user himself, setting it up.

> > Qmail looks at /etc/tcp.smtp via tcpserver, which only allows a single
> > file to be specified, so there is also a tcp.smtp is in
> > /var/vpopmail/etc, then qmail NEVER looks at it, as it really needs
> > /etc/tcp.smtp.
> Just for clarifying: tcpserver (and not qmail) looks into a cdb-file
> which you define as option for tcpserver with -x /path/to/file.cdb. 
That is exactly what I said: 'Qmail looks at /etc/tcp.smtp via
tcpserver'.

> So if one uses vpopmail the run-file for qmail-smtpd could be changed in
> a way that tcpserver looks in another cdb-file for which vpopmail has
> write access. The vanilla vpopmail suggests this IIRC.
The problem is that tcpserver only takes the last '-x' parameter it is
passed, so you cannot give it multiple cdbfiles.

I'd like to enforce a clean seperation between the tcp.smtp that is set
by the administrator and the tcp.smtp that vpopmail wants to create.

No application should ever re-write configuration files as it goes,
there is too much potential for disaster, and ideally should have no
permissions to write to the files even.
(Continue reading)

Permalink | Reply | 2 comments |
headers
Spider | 1 Nov 2003 14:44
Picon
Favicon

Re: better handling of multibyte characters (nls/cjk/unicode)

begin  quote
On Sat, 01 Nov 2003 21:06:56 +0900
Mamoru KOMACHI <usata <at> gentoo.org> wrote:

> 
> I've just added unicode USE flag :)

Good.

> If http://bugs.gentoo.org/show_bug.cgi?id=9988 "setting LANG variable"
> is implemented we may think about it, but I don't think the system
> should set LANG variable (at least) until then.
> 

Well, ok.  Except that if you use utf-8 enabled locales and the like,
you really want to have LANG set to a utf-8 locale, rather than a
"normal" one. 
//Spider

--

-- 
begin  .signature
This is a .signature virus! Please copy me into your .signature!
See Microsoft KB Article Q265230 for more information.
end
Permalink | Reply | 0 comments |
headers
James Harlow | 1 Nov 2003 15:40
Picon

Re: glibc-2.3.2-r[68] libpthread -> segmentation fault

> On Sat, Nov 01, 2003 at 03:43:45AM -0300, Norberto Bensa wrote:
> > 	CFLAGS="-march=pentium -Os -ffast-math -fomit-frame-pointer -pipe"

On Fri, Oct 31, 2003 at 10:50:14PM -0800, Robin H. Johnson wrote:
> do NOT use -ffast-math, it's asking for trouble (look it up in 'man gcc'
> for the reason).

Leaving aside the question of sillyness of using -ffast-math in the
system CFLAGS, I'd be surprised if `ls` made use of any floating-point
math at all, let alone requiring a strict implementation of the IEEE
standard.

--

-- 
When a true genius appears in the world, you may know him by this sign, that the dunces are all in confederacy
against him. - Jonathan Swift

--
gentoo-dev <at> gentoo.org mailing list
Permalink | Reply | 0 comments |
headers
Martin Schlemmer | 1 Nov 2003 16:02
Picon
Favicon

Re: glibc-2.3.2-r[68] libpthread -> segmentation fault

On Sat, 2003-11-01 at 08:43, Norberto Bensa wrote:
> > On Fri, 2003-10-31 at 03:22, Norberto Bensa wrote:
> > > I've tried swaping the microprocessor: same message. Seems compiled wrong
> > > by the box. I'll dedicate a Pentium MMX box to update glibc and see how
> > > it goes.
> 
> Martin,
> 
> I've tried something better. My firewall runs on a Pentium MMX box. I've 
> tarred the filesystem and moved it to my main box, a Pentium III [1]
> 
> Once on the P3 box, I chrooted and then I've 'emerge -u glibc' This is the 
> result:
> 
> 	# ls
> 	Segmentation fault
> 
> So the problem are not K6. The problem, IMHO, is gcc. CFLAGS on the firewall 
> (P-MMX) are these:
> 
> 	CFLAGS="-march=pentium -Os -ffast-math -fomit-frame-pointer -pipe"
> 
> "-march=pentium" is what I use on the K6 box where I first experienced this 
> problem. I'll try -march=pentium{2,3} tomorrow and keep you informed if you 
> want.
> 

So basically so far on a k6 and a p3, with -march=pentium it borks ?

Just to refresh (sorry if I asked before), what glibc, gcc and binutils
(Continue reading)

Permalink | Reply | 7 comments |

Gmane