11:06 a.m.
Hi,
I've recently had to link a fair amount of my work statically so that
it'll run on a cluster of RHEL machines. Unfortunately, I am just a
user of these machines, and so I don't have the power to get them to run
Fedora or even to get the admins to install RHEL packages in a timely
manner. Building statically also helps me to eliminate as many of the
inevitable fractional differences between cluster nodes as possible, to
achieve reproducible results from simulation runs.
However, only a few packages in Fedora provide -static variants. This
has meant that I've had to locally build these, which is obviously not
desirable from a maintenance perspective.
So, would be acceptable to register requests for -static package
variants as tickets on bugzilla? Or is there a better way to try to
encourage people to generate these packages?
Cheers,
Rob
11:11 a.m.
On Wed, Sep 15, 2010 at 05:06:20PM +0100, Robert Spanton wrote:
I've recently had to link a fair amount of my work statically so
that
it'll run on a cluster of RHEL machines. Unfortunately, I am just a
user of these machines, and so I don't have the power to get them to run
Fedora or even to get the admins to install RHEL packages in a timely
manner. Building statically also helps me to eliminate as many of the
inevitable fractional differences between cluster nodes as possible, to
achieve reproducible results from simulation runs.
However, only a few packages in Fedora provide -static variants. This
has meant that I've had to locally build these, which is obviously not
desirable from a maintenance perspective.
So, would be acceptable to register requests for -static package
variants as tickets on bugzilla? Or is there a better way to try to
encourage people to generate these packages?
Better yet is not to link statically.
http://www.akkadia.org/drepper/no_static_linking.html
Jakub
11:19 a.m.
On Wed, Sep 15, 2010 at 06:11:21PM +0200, Jakub Jelinek wrote:
On Wed, Sep 15, 2010 at 05:06:20PM +0100, Robert Spanton wrote:
> I've recently had to link a fair amount of my work statically so that
> it'll run on a cluster of RHEL machines. Unfortunately, I am just a
> user of these machines, and so I don't have the power to get them to run
> Fedora or even to get the admins to install RHEL packages in a timely
> manner. Building statically also helps me to eliminate as many of the
> inevitable fractional differences between cluster nodes as possible, to
> achieve reproducible results from simulation runs.
>
> However, only a few packages in Fedora provide -static variants. This
> has meant that I've had to locally build these, which is obviously not
> desirable from a maintenance perspective.
>
> So, would be acceptable to register requests for -static package
> variants as tickets on bugzilla? Or is there a better way to try to
> encourage people to generate these packages?
Better yet is not to link statically.
http://www.akkadia.org/drepper/no_static_linking.html
There are times when static linking is a useful. Robert clearly
describes one in his original post. I also use static linking as a
way to distribute binaries that will run on a range of Linux
distros[*].
Uli's claim that static linking's benefits "never been the case and
never will be the case" is complete drivel IMHO.
Rich.
[*] In fact, I wrote a nice little script to help:
http://git.annexia.org/?p=libguestfs.git;a=blob;f=relink-static.sh;hb=HEAD
--
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
virt-top is 'top' for virtual machines. Tiny program with many
powerful monitoring features, net stats, disk stats, logging, etc.
http://et.redhat.com/~rjones/virt-top
11:27 a.m.
On Thu, 2010-09-16 at 17:19 +0100, Richard W.M. Jones wrote:
There are times when static linking is a useful. Robert clearly
describes one in his original post.
Only because we do not (yet) have a good per-user package manager to
make installing the required dynamic libraries, or assembling a bundle
containing them, comparably easy.
--
Matt
11:32 a.m.
Richard W.M. Jones píše v Čt 16. 09. 2010 v 17:19 +0100:
There are times when static linking is a useful. Robert clearly
describes one in his original post.
Static libraries solve the "unprivileged
user" problem only in a limited
set of cases.
As soon as any of the libraries needs additional files (e.g. glade UI
files in /usr/share/mylibrary/*), static linking doesn't help.
The only general solution for these cases is to recompile the software
with --prefix inside the user's home directory.
Mirek
3:36 p.m.
On Thu, Sep 16, 2010 at 06:32:20PM +0200, Miloslav Trmač wrote:
Richard W.M. Jones píše v Čt 16. 09. 2010 v 17:19 +0100:
> There are times when static linking is a useful. Robert clearly
> describes one in his original post.
Static libraries solve the "unprivileged user" problem only in a limited
set of cases.
As soon as any of the libraries needs additional files (e.g. glade UI
files in /usr/share/mylibrary/*), static linking doesn't help.
The only general solution for these cases is to recompile the software
with --prefix inside the user's home directory.
It's not just that. Debian, for example, has most of the libraries
that my software needs, but they use different sonames (eg.
libpcre.so.0 vs libpcre.so.3). The other libraries like hivex don't
use any external files.
Rich.
--
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming blog: http://rwmj.wordpress.com
Fedora now supports 80 OCaml packages (the OPEN alternative to F#)
http://cocan.org/getting_started_with_ocaml_on_red_hat_and_fedora
11:21 a.m.
On 09/15/2010 05:06 PM, Robert Spanton wrote:
Hi,
I've recently had to link a fair amount of my work statically so that
it'll run on a cluster of RHEL machines. Unfortunately, I am just a
user of these machines, and so I don't have the power to get them to run
Fedora or even to get the admins to install RHEL packages in a timely
manner. Building statically also helps me to eliminate as many of the
inevitable fractional differences between cluster nodes as possible, to
achieve reproducible results from simulation runs.
However, only a few packages in Fedora provide -static variants. This
has meant that I've had to locally build these, which is obviously not
desirable from a maintenance perspective.
So, would be acceptable to register requests for -static package
variants as tickets on bugzilla? Or is there a better way to try to
encourage people to generate these packages?
You might find the Fedora packaging guidelines for packaging static
libraries helps explain the rationale a bit more:
http://tinyurl.com/kz4rp [fedoraproject.org]
There have also been several discussions of this on the fedora lists
over the years, e.g.:
http://www.spinics.net/lists/fedora-devel/msg48361.html
Regards,
Bryn.
11:39 a.m.
"Bryn M. Reeves" <bmr(a)redhat.com> writes:
On 09/15/2010 05:06 PM, Robert Spanton wrote:
> So, would be acceptable to register requests for -static package
> variants as tickets on bugzilla? Or is there a better way to try to
> encourage people to generate these packages?
You might find the Fedora packaging guidelines for packaging static
libraries helps explain the rationale a bit more:
The important point here is that there is a distro policy against static
libraries (with only very narrow exceptions). Trying to get individual
package maintainers to go against the policy isn't a helpful way to
proceed.
regards, tom lane
6:14 a.m.
On Wed, 15 Sep 2010 12:39:09 -0400, Tom wrote:
"Bryn M. Reeves" writes:
> On 09/15/2010 05:06 PM, Robert Spanton wrote:
>> So, would be acceptable to register requests for -static package
>> variants as tickets on bugzilla? Or is there a better way to try to
>> encourage people to generate these packages?
> You might find the Fedora packaging guidelines for packaging static
> libraries helps explain the rationale a bit more:
The important point here is that there is a distro policy against static
libraries (with only very narrow exceptions).
It is against linking statically, not against building/packaging libs:
https://fedoraproject.org/wiki/Packaging:Guidelines#Packaging_Static_Libr...
| In general, packagers are strongly encouraged not to ship static libs
| unless a compelling reason exists.
In package reviews, package submitters typically don't give any
"compelling reason" for shipping the static libs. That's why many
-static packages are missing.
Trying to get individual package maintainers to go against the
policy
isn't a helpful way to proceed.
If shared libs are available and static linking requires FESCo approval in
that case, why ship the static libs? That's the question to begin with.
If it's asked regularly, the Fedora Packaging Committee might see a need
to answer it and add a comment to the guidelines.
11:50 a.m.
On Wed, 2010-09-15 at 17:06 +0100, Robert Spanton wrote:
I've recently had to link a fair amount of my work statically so
that
it'll run on a cluster of RHEL machines. Unfortunately, I am just a
user of these machines, and so I don't have the power to get them to run
Fedora or even to get the admins to install RHEL packages in a timely
manner. Building statically also helps me to eliminate as many of the
inevitable fractional differences between cluster nodes as possible, to
achieve reproducible results from simulation runs.
However, only a few packages in Fedora provide -static variants. This
has meant that I've had to locally build these, which is obviously not
desirable from a maintenance perspective.
So, would be acceptable to register requests for -static package
variants as tickets on bugzilla? Or is there a better way to try to
encourage people to generate these packages?
I think the answer is "no": Fedora wishes to discourage static linking.
You can continue to build your own -static packages, or you can install
the libraries you need manually in your home directory (which probabably
isn't any easier, but gives you the other benefits of dynamic linking),
or you could set up a QEMU or User Mode Linux VM and install whatever
you like.
Ideally, we would have a packaging tool that works for per-user software
collections as well as the system. That's something I plan to work on
in the next few years.
--
Matt
1:51 p.m.
On Wed, Sep 15, 2010 at 05:06:20PM +0100, Robert Spanton wrote:
So, would be acceptable to register requests for -static package
variants as tickets on bugzilla? Or is there a better way to try to
encourage people to generate these packages?
Providing static libraries for user needs like yours is acceptable in
Fedora and some packages ship static libraries, especially when the
maintainer faces the issue himself (heterogenous build/run environments,
reproducible builds, lasting functional binaries). However, the
majority of Fedora developpers are opposed to providing static libraries,
even when users ask for them, and the guidelines themselves strongly
discourage shipping static libraries. So, you can try to ask maintainers
to ship static libraries, but don't expect to have many of them
understand your needs.
--
Pat
4:53 a.m.
On Wed, 2010-09-15, I wrote:
However, only a few packages in Fedora provide -static variants.
This
has meant that I've had to locally build these, which is obviously not
desirable from a maintenance perspective.
So, would be acceptable to register requests for -static package
variants as tickets on bugzilla? Or is there a better way to try to
encourage people to generate these packages?
Ok, the it seems that some people got the idea about why I want to be
able to statically link things. Others seem to have completely
misunderstood [1].
So we have to hack around the situation to achieve what we want that
some other people don't want. There is an existing tool called
statifier [2], but this doesn't work with VDSO randomisation, so I can't
use that. Here's a script that takes an ELF file and makes a directory
containing the libraries that it needs to live:
https://www.studentrobotics.org/cgit/rspanton/dynpk.git/tree/
Yup, it has to take the local system's dynamic linker too (RHEL's
dynamic linker made moaning noises about some difference in the way the
ELF file were labelled). My script also includes an "audit" library
that will force the application to exit if it attempts to load a shared
library from outside the packaged directory. This approach is working
for me -- and I *think* it's working for dlopen, but this is untested at
this point (I'll fix it when it breaks...).
Notice that the absence of -static packages hasn't actually stopped
users from achieving the same results... all it's done is add a hurdle
of inconvenience.
Cheers,
Rob
[1] Yea, like I'm bothered about the security issues of my simulation!
Other things I also don't care about in this situation:
* Memory usage: These machines have >32GB. Size of libraries
<50MB. Number of instances of this process per node <= 4.
* Binary size: Unbelievably insignificant.
* Library updates: I want my simulations to all be comparable...
[2] http://statifier.sourceforge.net/
4983
days inactive
4989
days old
11 comments
9 participants
participants (9)
-
Bryn M. Reeves -
Jakub Jelinek -
Matt McCutchen -
Michael Schwendt -
Miloslav Trmač -
Patrice Dumas -
Richard W.M. Jones -
Rob Spanton -
Tom Lane