On Fri, 2009-10-30 at 10:01 -0400, Duncan Smith wrote:
> The company I work for is using gentoo on all its machines. We just
> got a license to a commercial tool which does not support gentoo. The
> closest thing it supports is RHEL v4.
>
> Running any command provided by the tool results in an explosive
> memory leak (virtual memory hits 400G in 1 second, and continues to
> climb).
>
> I suspect the problem is that RHEL v4 uses =sys-libs/glibc-2.3.4,
> whereas we have =sys-libs/glibc-2.9_p20081201-r2 installed.
>
> I have three questions:
> 1. Am I posting to the right list?

You are just just as likely to get support from Gentoo about software we
have no access to as your distributer is to support Gentoo.

> 2. Any idea what's going on? Could it be something other than glibc
> causing the problem?

It could be one of a hundred million things. Without access to the
program it's really hard to tell.

> 3. If it is glibc, is there some way to install glibc slotted? Could
> I install an old version of glibc to some other lib folder (like
> /opt/lib64), and then use LD_LIBRARY_PATH somehow to get the tool to
> look there first? How?

You can't have multiple versions of glibc. And you can't downgrade
glibc. Attempting to do so may result in having more than just that
program misbehaving

My suggestion, for your sanity and support: if you insist on Gentoo then
at least run RHEL4 (or CentOS or whatever) inside a virtual machine and
run your app from there.

On Freitag 30 Oktober 2009, Albert Hopkins wrote:

> > 3. If it is glibc, is there some way to install glibc slotted? Could
> > I install an old version of glibc to some other lib folder (like
> > /opt/lib64), and then use LD_LIBRARY_PATH somehow to get the tool to
> > look there first? How?
>
> You can't have multiple versions of glibc. And you can't downgrade
> glibc. Attempting to do so may result in having more than just that
> program misbehaving

you can have multiple glibc's. Just not via portage.
But yes, it is a mess. A mess that is most likely to explode violently.

>
> My suggestion, for your sanity and support: if you insist on Gentoo then
> at least run RHEL4 (or CentOS or whatever) inside a virtual machine and
> run your app from there.
>

yeah, I would go down that route too.

On Freitag 30 Oktober 2009, Duncan Smith wrote:
> Thank you both for your quick response.
>
> I'll probably end up taking the virtual machine approach. I may also
> try some sort of chroot solution... I'll see how much of a hassle
> vmware is.

chroot can work nicely, but you have to create a gentoo with a very old glibc.
The problems might not be worth it.

Virtualbox on the other hand is pretty much hassle free in my experience.
Can't talk about vmware - haven't used that in years

On Friday 30 October 2009 23:52:10 Kyle Bader wrote:
> Avoiding 1, 2, and 3 but thought I'd propose a 4 other than a virtual
> machine. Ask the vendor if they can provide a statically compiled
> version, that way you don't have to worry about libc. I dunno how
> flexible the vendor is but its worth asking


If it's a somewhat critical machine for business, just drop a new stand-alone
box running RHEL4. Critical machines usually generate|save more cash than the
cost of the box they run on








> On 10/30/09, Duncan Smith <duncanphilipnorman@gmail.com> wrote:
> > The company I work for is using gentoo on all its machines. We just
> > got a license to a commercial tool which does not support gentoo. The
> > closest thing it supports is RHEL v4.
> >
> > Running any command provided by the tool results in an explosive
> > memory leak (virtual memory hits 400G in 1 second, and continues to
> > climb).
> >
> > I suspect the problem is that RHEL v4 uses =sys-libs/glibc-2.3.4,
> > whereas we have =sys-libs/glibc-2.9_p20081201-r2 installed.
> >
> > I have three questions:
> > 1. Am I posting to the right list?
> > 2. Any idea what's going on? Could it be something other than glibc
> > causing the problem?
> > 3. If it is glibc, is there some way to install glibc slotted? Could
> > I install an old version of glibc to some other lib folder (like
> > /opt/lib64), and then use LD_LIBRARY_PATH somehow to get the tool to
> > look there first? How?
> >
> > Thanks for any help or ideas.
> >
> > Duncan
> >
> > P.S. In case it's useful, here is the output of ldd:
> > linux-vdso.so.1 => (0x00007fff9e3ff000)
> > libncurses.so.5 => /lib/libncurses.so.5 (0x00007f49c871b000)
> > libresolv.so.2 => /lib/libresolv.so.2 (0x00007f49c8503000)
> > libm.so.6 => /lib/libm.so.6 (0x00007f49c827e000)
> > libdl.so.2 => /lib/libdl.so.2 (0x00007f49c807a000)
> > libc.so.6 => /lib/libc.so.6 (0x00007f49c7d07000)
> > /lib64/ld-linux-x86-64.so.2 (0x00007f49c897a000)
>

--
alan dot mckinnon at gmail dot com

On Fri, 2009-10-30 at 14:52 -0700, Kyle Bader wrote:
> I dunno how
> flexible the vendor is but its worth asking

They only support RHEL4. RHEL4 was released nearly 5 years ago and uses
the 2.6.9 kernel. I think that shows how flexible they are.

I was in a similar position some years ago - grab a copy of the needed
libs from somewhere and use "ldpreload" to load them into memory before
running the application. Google will help.

In some cases, you can symlink the needed lib names to existing later
libs and run ldconfig before trying to run the app. This does work
sometimes, but success varies ...

BillK





On Fri, 2009-10-30 at 19:27 -0400, Albert Hopkins wrote:
> On Fri, 2009-10-30 at 14:52 -0700, Kyle Bader wrote:
> > I dunno how
> > flexible the vendor is but its worth asking
>
> They only support RHEL4. RHEL4 was released nearly 5 years ago and uses
> the 2.6.9 kernel. I think that shows how flexible they are.
>
>
>
--
William Kenworthy <billk@iinet.net.au>
Home in Perth!