Copyright © 1999-2012 Erik Andersen
Mail all comments, insults, suggestions and bribes to
Bernhard Reutner-Fischer
Mail all comments, insults, suggestions and bribes to
Bernhard Reutner-Fischer
|
|
| About Documentation Obtain Development Links |
Frequently Asked QuestionsThis is a collection of some of the most frequently asked questions about uClibc. Some of the questions even have answers. If you have additions to this FAQ document, we would love to add them,
Why is it called uClibc?The letter 'u' is short for µ (the greek letter "mu"). µ is commonly used as the abbreviation for the word "micro". The capital "C" is short for "controller". So the name uClibc is sortof an abbreviation for "the microcontroller C library". For simplicity, uClibc is pronounced "yew-see-lib-see". The name is partly historical, since uClibc was originally created to support µClinux, a port of Linux for MMU-less microcontrollers such as the Dragonball, Coldfire, and ARM7TDMI. These days, uClibc also works just fine on normal Linux systems (such as i386, ARM, and PowerPC), but we couldn't think of a better name.
What platforms does uClibc run on?Currently uClibc runs on alpha, amd64, ARM, Blackfin, cris, h8300, hppa, i386, i960, ia64, m68k, mips/mipsel, PowerPC, SH, SPARC, and v850 processors.
Why are you doing this? What's wrong with glibc?Initially, the project began since the GNU C library lacked support for MMU-less systems, and because glibc is very large. The GNU C library is designed with a very different set of goals then uClibc. The GNU C library is a great piece of software, make no mistake. It is compliant with just about every standard ever created, and runs on just about every operating system and architecture -- no small task! But there is a price to be paid for that. It is quite a large library, and keeps getting larger with each release. It does not even pretend to target embedded systems. To quote from Ulrich Drepper, the maintainer of GNU libc: "...glibc is not the right thing for [an embedded OS]. It is designed as a native library (as opposed to embedded). Many functions (e.g., printf) contain functionality which is not wanted in embedded systems." 24 May 1999
So uClibc is smaller then glibc? Doesn't that mean it completely sucks? How could it be smaller and not suck?
uClibc and glibc have different goals. glibc strives for features and performance, and is targeted for desktops and servers with (these days) lots of resources. It also strives for ABI stability. On the other hand, the goal of uClibc is to provide as much functionality as possible in a small amount of space, and it is intended primarily for embedded use. It is also highly configurable in supported features, at the cost of ABI differences for different configurations. uClibc has been designed from the ground up to be a C library for embedded Linux. We don't need to worry about things like MS-DOS support, or BeOS, or AmigaOs any other system. This lets us cut out a lot of complexity and very carefully optimize for Linux. In other cases, uClibc leaves certain features (such as full C99 Math library support, wordexp, IPV6, and RPC support) disabled by default. Those features can be enabled for people that need them, but are otherwise disabled to save space. Some of the space savings in uClibc is obtained at the cost of performance, and some is due to sacrificing features. Much of it comes from aggressive refactoring of code to eliminate redundancy. In regards to locale data, elimination of redundant data storage resulted in substantial space savings. The result is a libc that currently includes the features needed by nearly all applications and yet is considerably smaller than glibc. To compare "apples to apples", if you take uClibc and compile in locale data for about 170 UTF-8 locales, then uClibc will take up about 570k. If you take glibc and add in locale data for the same 170 UTF-8 locales, you will need over 30MB!!! The end result is a C library that will compile just about everything you throw at it, that looks like glibc to application programs when you compile, and is many times smaller.
Why should I use uClibc?I don't know if you should use uClibc or not. It depends on your needs. If you are building an embedded Linux system and you are tight on space, then using uClibc instead if glibc may be a very good idea. If you are building an embedded Linux system and you find that glibc is eating up too much space, you should consider using uClibc. If you are building a huge fileserver with 12 Terabytes of storage, then using glibc may make more sense. Unless, for example, that 12 Terabytes will be Network Attached Storage and you plan to burn Linux into the system's firmware...
If I use uClibc, do I have to release all my source code to the world for free? I want to create a closed source commercial application and I want to protect my intellectual property.No, you do not need to give away your application source code just because you use uClibc and/or run on Linux. uClibc is licensed under the Lesser GPL license, just like the GNU C library (glibc). Please read this license, or have a lawyer read this license if you have any questions. Here is my brief summary... Using shared libraries makes complying with the license easy. You can distribute a closed source application which is linked with an unmodified uClibc shared library. In this case, you do not need to give away any source code for your application. Please consider sharing some of the money you make with us! :-) If you make any changes to uClibc, and distribute uClibc or distribute any applications using your modified version, you must also distribute the source code for uClibc containing all of your changes. If you distribute an application which has uClibc statically linked, you must also make your application available as an object file which can later be re-linked against updated versions of uClibc. This will (in theory) allow your customers to apply uClibc bug fixes to your application. You do not need to make the application object file available to everyone, just to those you gave the fully linked application.
Can I use it on my x86 development system?Sure! In fact, this can be very nice during development. By installing uClibc on your development system, you can be sure that the code you are working on will actually run when you deploy it on your target system.
Does uClibc support shared libraries?Yes. uClibc has native shared library support on i386, ARM, mips, SH, CRIS, and PowerPC processors. Other architectures can use shared libraries but will need to use the GNU libc shared library loader. Shared Libraries are not currently supported by uClibc on MMU-less systems. SnapGear has implemented shared library support for MMU-less systems, however, so if you need MMU-less shared library support they may be able to help.
How do I compile programs with uClibc?You will need to have your own uClibc toolchain. A toolchain consists of GNU binutils, the gcc compiler, and uClibc, all built to produce binaries for your target system linked with uClibc. You can build your own native uClibc toolchain using the uClibc buildroot system. To build your own uClibc toolchain, follow the following simple steps:
Do I really need to build a uClibc toolchain?Yes, you really do need to build a toolchain to produce uClibc binaries. We used to provide a toolchain wrapper, but that has been removed due to numerous problems. The uClibc developers have gone to a lot of trouble to produce the uClibc buildroot system, which makes it easy to build your own uClibc toolchain and/or an initial uClibc based system. Feel free to take the gcc and binutils patches we provide and use them in your own toolchain build system. If you choose to use your own toolchain build system, you will need to use these patches since the upstream GNU binutils and gcc releases do not currently have full support for building a uClibc toolchain.
What happened to the old toolchain wrapper?It is possible in some limited cases to re-use an existing glibc toolchain and subvert it into building uClibc binaries by using gcc commands such as "-nostdlib" and "-nostdinc". In fact, this used to be the recommended method for compiling programs with uClibc, and we made this easy to do by providing a uClibc toolchain wrapper, which attempted to automagically subvert an existing glibc toolchain. This toolchain wrapper was removed from uClibc 0.9.22, and it will not be coming back. This is because it proved impossible to completely subvert an existing toolchain in many cases, and therefore proved to be a real maintainence burder. As uClibc became more capable, the many problems with re-using an existing glibc toolchain led us to conclude that the only safe and sane way to build uClibc binaries was to use a uClibc toolchain. Some discussion on the reasoning behind this decision can be found here: https://www.uClibc.org/lists/uclibc/2003-October/007315.html in the uClibc mailing list archives.
Is a pre-compiled uClibc development system available?If you want to be really lazy and start using uClibc right away without needing to compile your own toolchain or anything, you can grab a pre-compiled uClibc development system. These are currently available for arm, armeb, i386, mips, mipsel, powerpc, and sh4. These are bzip2 compressed ext2 filesystems containing all the development software you need to build your own uClibc applications, including: bash, awk, make, gcc, g++, autoconf, automake, ncurses, zlib, openssl, openssh, gdb, strace, busybox, GNU coreutils, GNU tar, GNU grep, etc. Each of these uClibc development systems was created using buildroot. These development systems should provide pretty much everything you need to get started building your own applications with uClibc. Once you download one of these systems, you can then boot into it, loop mount it, dd it to a spare drive and use a tool such as resize2fs to make it fill a partition... Whatever works best for you. The quickest way to get started using a root_fs image (using the i386 platform as an example) is:
What versions of linux/gcc/etc... are supported?While we never go out of our way to break support for older versions of tools, reality is that slowly older versions fall by the wayside. Certain versions of tools will be actively supported while support for older versions will require interested parties to submit a patch. After all, validation requires a lot of effort, so it is unreasonable to ask the maintainers to validate older versions of tools when we've already said we don't really care if the older versions work. The current versions of things which are actively supported:
We know that people like to run older versions of Linux (such as the 2.2 and 2.4 series). If things do not work for you on those series, you will need to come up with a clean fix that can be merged into uClibc. For threading libraries, linuxthreads is planned to be maintained indefinitely so long as linux-2.4 support is active. After that, it will be the same as everything else: left alone and hopefully left working. The last two GNU binutils releases should work fine. Workarounds for bugs in older versions are not considered. The last official GCC point release is supported. If you encounter a compile error using an older version, then you're out of luck. You'll need to workaround it yourself. For the GNU toolchain releases, only the latest point release is supported. So if you encounter a bug in say GCC-x.y.7, but GCC-x.y.8 works fine, then we will not address it. Upgrading GCC across point releases is trivial. Same goes for binutils.
I think I found a bug in uClibc! What should I do?A well-written bug report should include
For larger applications where it may prove difficult to provide an example application, we recommend that you use a tool such as gdb, strace, ltrace, and or valgrind to create a logfile showing the problem behavior. If you find a problem with uClibc, please submit a detailed bug report to the uClibc bug tracker. If you are uncertain that it is a bug, send the bug report containing the information listed above to the mailing list at uclibc@uclibc.org. Please do not send private email to individual developers asking for private help unless you are planning on paying for consulting services. When we answer questions on the uClibc mailing list, it helps everyone, while private answers help only you... Please read the instructions on how to contribute before submitting patches!
My package builds fine but link fails with errors like "undefined reference to __fputc_unlocked", who do I blame?!This error crops up when a build system mixes system headers (say glibc) with the target headers (say uClibc). Make sure your build system is not including extraneous include options (-I) and double check that it is using the correct compiler. Many build systems incorrectly force things like -I/usr/include or -I/usr/local/include or -I${prefix}/include (which usually just expands to -I/usr/include).
My package builds fine but link fails with errors like "undefined reference to rpl_realloc / rpl_malloc", who do I blame?!This error crops up because you didn't enable MALLOC_GLIBC_COMPAT support. Configure scripts detect whether your libc supports "GNU malloc" features by seeing whether malloc(0) returns NULL (glibc malloc(0) will return a non NULL pointer). uClibc defaults to malloc(0) returning NULL so autoconf will assume malloc from your libc sucks and procede to try and compile the malloc replacement version in the gnulib addon. However, since not many packages include gnulib, they will fail to link with rpl_realloc / rpl_malloc errors. You could force configure to assume working behavior by exporting the variables 'jm_cv_func_working_malloc' (older autoconf) and 'ac_cv_func_malloc_0_nonnull' (newer autoconf) to 'yes' before running `./configure`.
My package links fine, but at runtime it can't find symbols (like "tgetnum"), so why is uClibc broken?!This error crops up when libraries are not directly linked against other libraries that they need, and at runtime the necessary libraries are implicitly provided (due to yet another ELF in the chain being linked against it). The most common case is readline and its implicit curses dependency. Since there are many providers of curses functions out there, and may be swapped out at runtime, readline does not link against any specific library for these functions. Instead, the idea is that the end application gets to pick which library to link against and the runtime linker (ldso) takes care of fixing up the references. Basically, to fix things, make sure your libraries link directly against whatever libraries they need. So if readline needs ncurses, make sure it is linked against it. Do not use `ldd` to verify things, but rather use `readelf -d` (look for DT_NEEDED). The former shows the whole dependency tree while the latter shows the immediate dependencies. You can catch a lot of these errors by using -Wl,-z,defs in your LDFLAGS when compiling packages. The linker will abort the final link step if necessary symbols are not found. The default behavior is to ignore missing symbols when creating shared libraries.
Why do I keep getting "sh: can't access tty; job control turned off" errors? Why doesn't Control-C work within my shell?This isn't really a uClibc question, but I'll answer it here anyways. Job control will be turned off since your shell can not obtain a controlling terminal. This typically happens when you run your shell on /dev/console. The kernel will not provide a controlling terminal on the /dev/console device. Your should run your shell on a normal tty such as tty1 or ttyS0 and everything will work perfectly. If you REALLY want your shell to run on /dev/console, then you can hack your kernel (if you are into that sortof thing) by changing drivers/char/tty_io.c to change the lines where it sets "noctty = 1;" to instead set it to "0". I recommend you instead run your shell on a real console...
How do I make autoconf and automake behave?When you are cross-compiling, autoconf and automake are known to behave badly. This is because a large number of configure scripts (such as the one from openssh) try to actually execute applications that were cross compiled for your target system. This is bad, since of course these won't run, and this will also prevent your programs from compiling. You need to complain to the authors of these programs and ask them to fix their broken configure scripts.
When I run 'ldd' to get a list of the library dependencies for a uClibc binary, ldd segfaults! What should I do?Use the ldd that is built by uClibc, not your system's one. You can build uClibc'd ldd for your host system by going into the uClibc/utils/ directory in the uClibc source and running 'make ldd.host'. When your system's ldd looks for library dependencies, it actually _runs_ that program. This works fine -- usually. It generally will not work at all when you have been cross compiling (which is why ldd segfaults). The ldd program created by uClibc is cross platform and doesn't mind at all if it cannot execute the target program. If you use the uClibc version of 'ldd', it will do the right thing and produce correct results, even when it is used on cross compiled binaries.
Why does localtime() return times in UTC even when I have my timezone set?The uClibc time functions get timezone information from the TZ environment variable, as described in the Single Unix Specification Version 3. See http://www.opengroup.org/onlinepubs/007904975/basedefs/xbd_chap08.html for details on valid settings of TZ. For some additional examples, read https://www.uClibc.org/lists/uclibc/2002-August/004010.html in the uClibc mailing list archive. You can store the value of TZ in the file '/etc/TZ' and uClibc will then automagically use the specified setting.
What is the history of uClibc? Where did it come from?uClibc started off as a fork on the Linux-8086 C library, which is part of the elks project. The Linux-8086 C library was, apparently, largely written from scratch but also borrowed code from libc4, glibc, some Atari library code, with bits and pieces from about 20 other places. I had for some time been despairing over the state of C libraries in Linux. GNU libc, the standard, is very poorly suited to embedded systems and has been getting bigger with every release. I spent quite a bit of time looking over the available Open Source C libraries that I knew of, and none of them really impressed me. I felt there was a real vacancy in the embedded Linux ecology. The closest library to what I imagined an embedded C library should be was uClibc. But it had a lot of problems too -- not the least of which was that, traditionally, uClibc required a complete source tree fork in order to support each and every new platform. This resulted in a big mess of twisty versions, all different. I decided to fix it and the result is what you see here. To start with, (with some initial help from D. Jeff Dionne), I ported uClibc to run on i386. I then grafted in the header files from glibc and cleaned up the resulting breakage. This (plus some additional work) has made it much less dependant on kernel headers, a large departure from its traditional tightly-coupled-to-the-kernel origins. I have written and/or rewritten a number of things that were missing or broken, and sometimes grafted in bits of code from the current glibc and libc5. I have also added a proper configuration system which allows you to easily select your target architecture and enable and disable various features. Many people have helped by testing, contributing ports to new architectures, and adding support for missing features. In particular, around the end of 2000, Manuel Novoa III got involved with uClibc. One of his first contributions was the original gcc wrapper (which has since been removed). Since then, he has written virtually all of the current uClibc stdio, time, string, ctype, locale, and wchar-related code, as well as much of stdlib and various other bits throught the library. These days, uClibc is being developed and enhanced by Erik Andersen and Manuel Novoa III of CodePoet Consulting along with the rest of the embedded Linux community.
I demand that you to add <favorite feature> right now! How come you don't answer all my questions on the mailing list instantly? I demand that you help me with all of my problems Right Now!You have not paid us a single cent and yet you still have the product of several years of work from Erik and Manuel and many other people. We are not your slaves! We work on uClibc because we find it interesting. If you go off flaming us, we will ignore you.
I need help with uClibc! What should I do?If you find that you need help with uClibc, you can ask for help on the uClibc mailing list at uclibc@uclibc.org. In addition to the uClibc mailing list, Erik and Manuel are also known to sometimes hang out on the uClibc IRC channel: #uclibc on irc.freenode.net. Please do not send private email to Erik and/or Manuel asking for private help unless you are planning on paying for consulting services. When we answer questions on the uClibc mailing list, it helps everyone since people with similar problems in the future will be able to get help by searching the mailing list archives. Private help is reserved as a paid service. If you need to use private communication, or if you are serious about getting timely assistance with uClibc, you should seriously consider paying for consulting time.
I need you to add <favorite feature>! Are the uClibc developers willing to be paid in order to fix bugs or add in <favorite feature>? Are you willing to provide support contracts?Sure! Now you have our attention! What you should do is contact Erik Andersen of CodePoet Consulting to bid on your project. If Erik is too busy to personally add your feature, there are several other active uClibc contributors who will almost certainly be able to help you out. Erik can contact them and ask them about their availability.
I think you guys are great and I want to help support your work!Wow, that would be great! You can click here to help support uClibc and/or request features. |
|
Copyright © 1999-2012 Erik Andersen
Mail all comments, insults, suggestions and bribes to Bernhard Reutner-Fischer |
|
|