Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-17 Thread Ewald Wasscher

David Douthitt wrote:

>>I'd vote for 2.2. It may be bigger, but 2.1 will be unmaintained rather
>>soon I'm afraid. So when we choose for glibc 2.1 we might end up with
>>the same mess as we have for glibc 2.0 now in a year or so. Unless one
>>of us  is capable of backporting security fixes 2.2 is the way to go I
>>think.
>>
>
>I tend to agree, though I've already got 2.1.3 going.  Upgrading
>libraries is a hassle - biggest of which is compiling the libraries
>from scratch - but I imagine I'll be doing it soon enough.
>
Well, that should be doable I think. But compiling it takes ages.

>>That makes me wonder if anyone has seriously considered uClibc. It
>>probably has some limitations. One is of course that binaries compiled
>>with glibc won't run on a uClibc system.
>>
>
>That's one of the main reasons I went to glibc 2.1.3.  No need to have
>to scrounge for old binaries; just load a current (now semi-current)
>distribution and use that.
>
If you want to copy binaries this is obviously an advantage. If you want 
to compile your own stuff: with the recent gcc and binutils wrappers 
that come with uClibc compiling is a breeze. And installing a uClibc 
compile environtment is as easy, though some installation instructions 
wouldn't hurt.

Ewald Wasscher


___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-17 Thread David Douthitt

Ewald Wasscher wrote:
> 
> David Douthitt wrote:
> 
> >Pim van Riezen wrote:

> >>if I want to produce binaries I'll have to use three different
> >>environments if I want to cater for all glibc variations. Now that
> >>RH7/glibc2.2 is gaining acceptance that'll be four:
> >>
> >>  libc5

> Is anyone still using this?

Just today someone on the busybox list said that there was a problem
compiling busybox from CVS with libc5.  Some people do seem to be
still using it.

> I may be wrong, but I think that anything using ipv6 won't compile with
> glibc 2.0.

Sounds about right, but I don't know for sure either.

> I'd vote for 2.2. It may be bigger, but 2.1 will be unmaintained rather
> soon I'm afraid. So when we choose for glibc 2.1 we might end up with
> the same mess as we have for glibc 2.0 now in a year or so. Unless one
> of us  is capable of backporting security fixes 2.2 is the way to go I
> think.

I tend to agree, though I've already got 2.1.3 going.  Upgrading
libraries is a hassle - biggest of which is compiling the libraries
from scratch - but I imagine I'll be doing it soon enough.

> That makes me wonder if anyone has seriously considered uClibc. It
> probably has some limitations. One is of course that binaries compiled
> with glibc won't run on a uClibc system.

That's one of the main reasons I went to glibc 2.1.3.  No need to have
to scrounge for old binaries; just load a current (now semi-current)
distribution and use that.

___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-17 Thread George Metz

On Thu, 17 May 2001, David Douthitt wrote:

> > Why I never went anywhere with mine was mostly because I sent out several
> > e-mails to this list, and the lack of a response was almost deafening in
> > it's silence. If I recall, not even you commented David. I assumed that
> > people had weighed the concept and decided it wasn't worth the effort to
> > switch, so I moved onward.
>
> You said you get several questions a week :-)

Whoops! E-mails, 4am again, you understand. =)

I meant that to be in reference to my glibc 2.1 update efforts. My Kernel
2.4 efforts have been suddenly halted by a mysteriously sudden lack of an
autoconf.h file on my Linux system. I discovered that when I tried to
compile a 2.4.4 kernel, and I haven't had the time/opportunity to fix it
since.

2 days until vacation, then I can write the Abuse Reporting HOWTO, an
adventure for GenCon that I got tricked into volunteering for, and fixing
my system to do some hardcore devel work. =)

> I hadn't realized you stopped development or whatever; I've been
> watching the Linux 2.4 development with interest.  I had originally

Again, glibc 2.1. Kernel 2.4.4 is on hold until I can pull autoconf.h out
of my tail and stick it on my compy again.

> thought that 2.4 would take another 100k or something the way glibc
> 2.1 did; when your versions of Linux 2.4 showed up under 500k that was
> very interesting.

They're not MUCH under 500k, and most of that is due to UPX, but the
pulling of the PCI Device Name Database seems to be key; apparently it's a
LOT larger than the 20K advertised in the help file.

--
George Metz
Commercial Routing Engineer
[EMAIL PROTECTED]

"We know what deterrence was with 'mutually assured destruction' during
the Cold War. But what is deterrence in information warfare?" -- Brigadier
General Douglas Richardson, USAF, Commander - Space Warfare Center


___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-17 Thread Ewald Wasscher

David Douthitt wrote:

>Pim van Riezen wrote:
>
>>On Wed, 16 May 2001, David Douthitt wrote:
>>
>>>I must say I've been surprised at all the excitement over Linux
>>>2.4.  I've noticed that all of you kernel wizards are scrambling to
>>>get Linux 2.4 installed on LRP, while glibc 2.1 gets ignored.
>>>
>>For me, it's basically the fact that I'm silently still pissed off with
>>the mess I, as a developer, get when dealing with different glibc
>>versions. No matter what glibc I use on my development system, at the end
>>if I want to produce binaries I'll have to use three different
>>environments if I want to cater for all glibc variations. Now that
>>RH7/glibc2.2 is gaining acceptance that'll be four:
>>
>>  libc5
>>
Is anyone still using this?

>>
>>  glibc2.0
>>  glibc2.1
>>  glibc2.2
>>
I may be wrong, but I think that anything using ipv6 won't compile with 
glibc 2.0. I'm not an expert on this matter but I thought that with 
symbol versioning applications compiled with glibc 2.0 should run on 
glibc > 2.0 systems.

>
>Sounds like a good reason to shift from using glibc 2.0 to using glibc
>2.1 or 2.2.
>
I'd vote for 2.2. It may be bigger, but 2.1 will be unmaintained rather 
soon I'm afraid. So when we choose for glibc 2.1 we might end up with 
the same mess as we have for glibc 2.0 now in a year or so. Unless one 
of us  is capable of backporting security fixes 2.2 is the way to go I 
think.

>I, too, have seen teh MESS that comes from trying to
>compile things for glibc 2.0.  In particular, there are several
>applications which don't seem like they'll compile under glibc 2.0:
>
>* ax25-tools
>* zebra
>* brctl
>
>Of course, they all compile WITHOUT ERRORS OR WARNINGS under glibc
>2.1.
>
>>If you add to that the fact that a typical embedded application shouldn't
>>be using much more than the stdio, string and socket functions and you see
>>that I'm reluctant to change over to a newer glibc, which will probably
>>take more space without offering much in exchange.
>>
That makes me wonder if anyone has seriously considered uClibc. It 
probably has some limitations. One is of course that binaries compiled 
with glibc won't run on a uClibc system. I've been playing with uClibc 
lately and there seems to be considerable progress in it's development. 
With the last snapshot I tried also the libdl seemed to work, which 
makes it usable for an LRP like distro.

Also in this article: 
http://embedded.linuxjournal.com/magazine/issue00/4335/ Bruce Perens 
mentions a way of stripping down libraries. May be interesting for us.

>
>How about not having to compile for the old glibc versions?  Sounds
>like a good reason to me.  You gave lots of good reasons for why LRP
>(and variants) should move to glibc 2.1 or 2.2, instead of arguing
>against it.
>

Ewald Wasscher


___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-17 Thread David Douthitt

George Metz wrote:

> Why I never went anywhere with mine was mostly because I sent out several
> e-mails to this list, and the lack of a response was almost deafening in
> it's silence. If I recall, not even you commented David. I assumed that
> people had weighed the concept and decided it wasn't worth the effort to
> switch, so I moved onward.

You said you get several questions a week :-)

I hadn't realized you stopped development or whatever; I've been
watching the Linux 2.4 development with interest.  I had originally
thought that 2.4 would take another 100k or something the way glibc
2.1 did; when your versions of Linux 2.4 showed up under 500k that was
very interesting.

Perhaps the lack of interest is because:

1) users don't understand how to change kernels
2) no one has an overriding need to change to 2.4
3) no one has been given a need to change to 2.4 (i.e., no marketing)
4) people still think there isn't enough room on disk

Solutions are, in order:

1) Convert the developers over so that new versions of LEAF and of
firewall software use Linux 2.4
2) -- no fix ---
3) Advertize Linux 2.4, pitching the advantages on the list
4) Post sizes too

___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-17 Thread Pim van Riezen

On Wed, 16 May 2001, David Douthitt wrote:

> Sounds like a good reason to shift from using glibc 2.0 to using glibc
> 2.1 or 2.2.  I, too, have seen teh MESS that comes from trying to
> compile things for glibc 2.0.  In particular, there are several
> applications which don't seem like they'll compile under glibc 2.0:
>
> * ax25-tools
> * zebra
> * brctl

I haven't run into any compile troubles so far, specifically brctl works
fine for me.

> Of course, they all compile WITHOUT ERRORS OR WARNINGS under glibc
> 2.1.
>
> > If you add to that the fact that a typical embedded application shouldn't
> > be using much more than the stdio, string and socket functions and you see
> > that I'm reluctant to change over to a newer glibc, which will probably
> > take more space without offering much in exchange.
>
> How about not having to compile for the old glibc versions?  Sounds
> like a good reason to me.  You gave lots of good reasons for why LRP
> (and variants) should move to glibc 2.1 or 2.2, instead of arguing
> against it.

My point was that I have to compile for umpteen versions anyway and
glibc2.0 is smaller. Keep in mind that what I do with cish leads to much
less of a "generic linux platform" where it would make sense to include
arbitrary software packages than most of the LRP projects right now.

Pi

-- 
Head Development - Vuurwerk Internet (http://www.vuurwerk.nl/)
Brainbench MVP Unix Programming, twisted artist and Free Software idiot.

I need a mental stoma.


___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-16 Thread George Metz

On Wed, 16 May 2001, David Douthitt wrote:

> I must say I've been surprised at all the excitement over Linux
> 2.4.  I've noticed that all of you kernel wizards are scrambling to
> get Linux 2.4 installed on LRP, while glibc 2.1 gets ignored.

Not entirely. I've got a newlibs.tgz sitting on my hard drive right now
that contains glibc 2.1.3 libs in their correct directory. This was done
before 2.4.0 had hit test12.

> To me, Linux 2.4 offers only this:
>
> * Stateful firewalling

It also has support builtin or available for some pretty useful stuff -
the Alcatel SpeedTouch USB DSL modem being one I'm aware of - as well as
improved, updated, and all around more device drivers. For instance, with
my workstation's Ethernet card, a 3c905C NIC, I can use 3c59x.o under
2.4.1 or later. Under 2.2.xx, I have to find an obscure copy of the 3c90x
in order to actually use it. (I've provided this module to at least two
people on the LRP list.)

Other than that though, you're quite correct; all it adds is stateful
firewalling.

> We don't see question after question on the LRP list asking about
> stateful firewalling, nor are people banging down the doors looking
> for 2.4.

Maybe not YOUR door, but I get a few queries each week asking a question
about my 2.4 images.

> I hope no one thinks I'm criticizing - I just find that all the people
> I respect are doing this and I don't know why - so I thought I'd ask.

Why I never went anywhere with mine was mostly because I sent out several
e-mails to this list, and the lack of a response was almost deafening in
it's silence. If I recall, not even you commented David. I assumed that
people had weighed the concept and decided it wasn't worth the effort to
switch, so I moved onward.

--
George Metz
Commercial Routing Engineer
[EMAIL PROTECTED]

"We know what deterrence was with 'mutually assured destruction' during
the Cold War. But what is deterrence in information warfare?" -- Brigadier
General Douglas Richardson, USAF, Commander - Space Warfare Center


___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-16 Thread David Douthitt

Pim van Riezen wrote:
> 
> On Wed, 16 May 2001, David Douthitt wrote:
> 
> > I must say I've been surprised at all the excitement over Linux
> > 2.4.  I've noticed that all of you kernel wizards are scrambling to
> > get Linux 2.4 installed on LRP, while glibc 2.1 gets ignored.

> For me, it's basically the fact that I'm silently still pissed off with
> the mess I, as a developer, get when dealing with different glibc
> versions. No matter what glibc I use on my development system, at the end
> if I want to produce binaries I'll have to use three different
> environments if I want to cater for all glibc variations. Now that
> RH7/glibc2.2 is gaining acceptance that'll be four:
> 
>   libc5
>   glibc2.0
>   glibc2.1
>   glibc2.2

Sounds like a good reason to shift from using glibc 2.0 to using glibc
2.1 or 2.2.  I, too, have seen teh MESS that comes from trying to
compile things for glibc 2.0.  In particular, there are several
applications which don't seem like they'll compile under glibc 2.0:

* ax25-tools
* zebra
* brctl

Of course, they all compile WITHOUT ERRORS OR WARNINGS under glibc
2.1.

> If you add to that the fact that a typical embedded application shouldn't
> be using much more than the stdio, string and socket functions and you see
> that I'm reluctant to change over to a newer glibc, which will probably
> take more space without offering much in exchange.

How about not having to compile for the old glibc versions?  Sounds
like a good reason to me.  You gave lots of good reasons for why LRP
(and variants) should move to glibc 2.1 or 2.2, instead of arguing
against it.

___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel



Re: [Leaf-devel] Linux 2.4 versus glibc 2.1/2.2

2001-05-16 Thread Pim van Riezen

On Wed, 16 May 2001, David Douthitt wrote:

> I must say I've been surprised at all the excitement over Linux
> 2.4.  I've noticed that all of you kernel wizards are scrambling to
> get Linux 2.4 installed on LRP, while glibc 2.1 gets ignored.
>
> To me, Linux 2.4 offers only this:
>
> * Stateful firewalling
>
> We don't see question after question on the LRP list asking about
> stateful firewalling, nor are people banging down the doors looking
> for 2.4.
>
> Howver, we DO see constant questions about "Why do I get a SegFault in
> this app from my full distribution?" and we have a FAQ for it.  The
> answer, of course, is "You don't have glibc 2.1 installed."
>
> I hope no one thinks I'm criticizing - I just find that all the people
> I respect are doing this and I don't know why - so I thought I'd ask.

For me, it's basically the fact that I'm silently still pissed off with
the mess I, as a developer, get when dealing with different glibc
versions. No matter what glibc I use on my development system, at the end
if I want to produce binaries I'll have to use three different
environments if I want to cater for all glibc variations. Now that
RH7/glibc2.2 is gaining acceptance that'll be four:

  libc5
  glibc2.0
  glibc2.1
  glibc2.2

If you add to that the fact that a typical embedded application shouldn't
be using much more than the stdio, string and socket functions and you see
that I'm reluctant to change over to a newer glibc, which will probably
take more space without offering much in exchange.

Cheers,
Pi

-- 
Head Development - Vuurwerk Internet (http://www.vuurwerk.nl/)
Brainbench MVP Unix Programming, twisted artist and Free Software idiot.

I need a mental stoma.


___
Leaf-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/leaf-devel