I'm attempting to build a binary package that can be run on multiple Linux distributions. It's currently built on Ubuntu 10.04, but it fails on Ubuntu 8.04 with the following error:
./test: /usr/lib/libstdc++.so.6: version `GLIBCXX_3.4.11' not found (required by ./test)
./test: /usr/lib/libstdc.so.6: version `GLIBC_2.11' not found (required by ./test)
What's the preferred way to solve this problem? Is there a way to install an old glibc on a new box and build against it, or do I have to build on an old distribution? And if I build against an old glibc, will it work on a new glibc?
Or, alternatively, are there just some handy compiler flags or packages I could install to solve the problem?
The best solution I've found is to install a virtual machine running Debian stable, and build on that. Debian stable is old enough that any packages built with it will run on any other Debian-based distribution like Ubuntu. You may have to work around non-critical bugs that have been fixed in later versions of various software but not backported to Debian stable.
If you really want to make sure it runs on every recent distribution, you might also consider statically linking against a libC you select. However you may then still run into problems if you use features that are only provided by newer kernels (newer system calls e.g.).
Related
So, I recently noticed that depending on which Linux machine I was running, valgrind would throw a solid 90k errors, for the exact same file/code that threw no errors using valgrind on a different machine. So, I looked into it, and it looks like while one is running valgrind-3.7.0, the other has valgrind-3.15.0. sudo apt update/upgrade didn't show anything, and reinstalling didn't make a difference. (For reference, I am using Raspbian on one and Ubuntu on the other)
Now the question: Why are packages installed with different versions on different distributions? Would it be advisable to manually install the newer version on a distro that defaults to an older version? If so, what is the easiest way to do so?
Yes i have encountered such an issue.Some packages require older versions.The errors depend on the dependencies and your system configurations.
I guess to install an older version of what you want to install, do some research and try changing /etc/sources.list.
Redhat 5 has the required 2.6.18 kernel but not the latest glibc, g++ and certainly no clang. The binary distribution doesn't run, complaining about glibc version.
Has anyone made an attempt to back port to old Linux? I could imagine cobbling together a frankensystem with an old kernel but new compilers and try to compile against old glibc or statically compile in some of new glibc but it seems like a fraught course.
Just wondering if anyone has tried and can offer guidance?
I'm thinking there must be a distro setup to do such builds. Anaconda, for example, includes lots of new technologies but works fine on Redhat 5. I wonder how they build it?
Update: Once you get some newer compilers working on RHEL5 it's still not possible to build Rust because it depends on a working binary of itself to bootstrap. See: https://github.com/rust-lang/rust/issues/9545#issuecomment-54865031
Anaconda packages are built against CentOS 5 (which is equivalent to Redhat 5). The important thing is to compile against an old enough glibc, as it is strictly forward compatible (as you saw). The version of glibc is typically tied to the version of the distro, so your best bet is to make a VM with CentOS 5 and build on that.
The version of the compiler used is not as important, and in fact in some cases it is necessary to use a newer compiler than the old one that comes with the system to get things to work. I have gcc 4.8 built for CentOS 5 which you can get with conda (conda install -c asmeurer gcc).
I hit this very issue this weekend, because Skylight supports Linux 2.6.18, including CentOS 5.10, and we use Rust in our agent.
Alex Crichton of the Rust core team was kind enough to get this working again by using CentOS 5.10 boxes for the build, which uses glibc 2.5 and Linux 2.6.18.
The mysql++ package that comes from tangentsoft, is version 3.1.0 and I am having trouble building on my linux system. To get it to configure I had to specify the with the flag
./configure --with-mysql-lib=/usr/lib/i386-linux-gnu/
Otherwise there was a problem with it finding the mysqlclient library. Now unfortunately it wont make, the error I get is ./lib/refcounted.h:258:2: error: ‘size_t’ does not name a type, after I type make and before I type make install.
Has anyone had the same problem with compiling it, and did you find a solution?
Any help it solving this problem is appreciated.
There are a few #include fixes in the svn version of MySQL++ that haven't yet been included in a release version. They're necessary on modern OSes like yours, which include recent GCCs and libstdc++'s.
Don't be nervous about using the svn version in production. It's quite stable at the moment. My reasons for not releasing it as-is aren't related to stability.
A C program compiled here runs fine on our Ubuntu servers. But when a somebody else tries to run it on their particular Linux server they get the following errors:
./myprog-install: /lib/tls/libc.so.6: version `GLIBC_2.4' not found (required by ./myprog-install)
./myprog-install: /lib/tls/libc.so.6: version `GLIBC_2.7' not found (required by ./myprog-install)
Do I need to upgrade our glibc libraries and recompile? Or are they missing something on their server?
If I run apt-cache show libc6 my Ubuntu tells me the version is:
Package: libc6
Priority: required
Section: libs
Installed-Size: 9368
Maintainer: Ubuntu Core developers <ubuntu-devel-discuss#lists.ubuntu.com>
Original-Maintainer: GNU Libc Maintainers <debian-glibc#lists.debian.org>
Architecture: i386
Source: eglibc
Version: 2.11.1-0ubuntu7.10
If I look at http://packages.ubuntu.com/hardy/libc6 the current version appears to be 2.7-10ubuntu8.1.
I'm confused by the numbers. On the one hand 2.11-1-0 is a higher number than 2.7-11. On the other hand 7.10 is a lower number than 8.1.
Is it just a matter of me upgrading the C library package and recompiling do you think? Or is the other person's server missing some needed library there for compatibility?
You have built on glibc-2.11 system. You are trying to run on a system with glibc-2.3 or older. That's not going to work.
Is it just a matter of me upgrading the C library package
No: upgrading your glibc will only make things worse.
You may want to try solutions listed here.
Is this something we can reasonably request the other party to upgrade their system to support, rather than downgrade our compiler?
Usually the client will strongly resist requests to upgrade their system: it's working fine for them as is, and any upgrade can break other existing applications.
If you are planning to distribute binaries on Linux (as opposed to building them on the target system), then you need to learn how to make binaries that will run everywhere, or you need to state your requirements (minimum kernel and libc versions, etc.) and turn clients who can't meet these requirements away.
Update:
Why did they get two errors. Why didn't they just get one for GLIBC_2.11.1 which is apparently what I built with?
Symbol versioning doesn't work that way.
When a new symbol is introduced, it is marked with the current libc version, e.g. readdir64##GLIBC_2.2, posix_spawn##GLIBC_2.15, etc.
When you link a program that uses both of the above symbols, and try to run it on e.g. glibc-2.1 system, you would get two errors.
But if you link a program that doesn't use any of the above symbols, e.g.
int main() { return 0; }
then your program will just run without any errors.
Update 2:
they don't have to add both GLIBC_2.4 and GLIBC2.7 to their Linux system, do they?
No, they don't. The GLIBC_2.11 will have all the previous symbols in it. In fact, they couldn't install both glibc-2.4 and 2.7 even if they wanted to: it is quite difficult to have multiple versions installed at the same time, and impossible to have multiple versions installed in default location.
You've built it against a version of glibc that is too new. Build it against an older version of glibc, preferably the one that they are using.
you need to build on a system that uses same versions of libraries as you do. This is where docker and VM's are very convenient. There is probably a pre-made docker image for whatever version the customer has.
Can I install an older version of gcc/g++ (4.1.3) on the latest Ubuntu (which comes with 4.4.3) and use it to compile a .so which should run on CentOS? The binary compiled with the Ubuntu version of gcc fails to load on CentOS because of missing imports (GLIB_2_11, ...). I need C++ (including exceptions), so I can't just statically link against glibc, which I already tried.
Can I install the older gcc without removing the newer one? How do I go about the libs required by the older gcc?
I'm currently developing code in CentOS, but it's such a pain to use. I really want to move to an Ubuntu desktop.
g++-4.1 is available for Ubuntu; just run apt-get install g++-4.1 then run g++-4.1 instead of g++. However, simply using an older compiler may not fix all of your library issues.
Like Joachim Sauer said, your best bet is to do your development on Ubuntu then do the final compilation on CentOS.
Even though you're using C++, static linking should still be an option. (However, you're much better off compiling on CentOS and using dynamic linking.)
Edit: A virtual machine is the most straightforward way to build on CentOS, but if you want to avoid the memory and CPU overhead of running a VM and don't care about differences between Ubuntu's and CentOS's kernel, then you can create a subdirectory containing a CentOS or Fedora filesystem and chroot do that to do your builds. This blog posting has details.