# # For a description of the syntax of this configuration file, # see extra/config/Kconfig-language.txt # mainmenu "uClibc C Library Configuration" choice prompt "Target Architecture" default TARGET_i386 help Stuff config TARGET_alpha bool "alpha" config TARGET_arm bool "arm" config TARGET_cris bool "cris" config TARGET_e1 bool "e1" config TARGET_h8300 bool "h8300" config TARGET_i386 bool "i386" config TARGET_i960 bool "i960" config TARGET_m68k bool "m68k" config TARGET_microblaze bool "microblaze" config TARGET_mips bool "mips" config TARGET_powerpc bool "powerpc" config TARGET_sh bool "SuperH" config TARGET_sparc bool "sparc" config TARGET_v850 bool "v850" endchoice menu "Target Architecture Features and Options" if TARGET_alpha source "extra/Configs/Config.alpha" endif if TARGET_arm source "extra/Configs/Config.arm" endif if TARGET_cris source "extra/Configs/Config.cris" endif if TARGET_e1 source "extra/Configs/Config.e1" endif if TARGET_h8300 source "extra/Configs/Config.h8300" endif if TARGET_i386 source "extra/Configs/Config.i386" endif if TARGET_i960 source "extra/Configs/Config.i960" endif if TARGET_m68k source "extra/Configs/Config.m68k" endif if TARGET_microblaze source "extra/Configs/Config.microblaze" endif if TARGET_mips source "extra/Configs/Config.mips" endif if TARGET_powerpc source "extra/Configs/Config.powerpc" endif if TARGET_sh source "extra/Configs/Config.sh" endif if TARGET_sparc source "extra/Configs/Config.sparc" endif if TARGET_v850 source "extra/Configs/Config.v850" endif source "extra/Configs/Config.in.arch" endmenu menu "General Library Settings" config HAVE_NO_PIC bool default n config DOPIC bool "Generate Position Independent Code (PIC)" default y depends !HAVE_NO_PIC help If you wish to build uClibc with support for shared libraries then answer Y here. If you only want to build uClibc as a static library, then answer N. config HAVE_NO_SHARED bool default n config HAVE_SHARED bool "Enable support for shared libraries" depends on DOPIC && !HAVE_NO_SHARED default y help If you wish to build uClibc with support for shared libraries then answer Y here. If you only want to build uClibc as a static library, then answer N. config ARCH_HAS_NO_LDSO bool default n config BUILD_UCLIBC_LDSO bool "Compile native shared library loader" depends on HAVE_SHARED && !ARCH_HAS_NO_LDSO default y help uClibc has a native shared library loader for some architectures. If you answer Y here, the uClibc native shared library loader will be built for your target architecture. If this option is available, to you, then you almost certainly want to answer Y. config FORCE_SHAREABLE_TEXT_SEGMENTS bool "Only load shared libraries which can share their text segment" depends on BUILD_UCLIBC_LDSO && UCLIBC_COMPLETELY_PIC default n help If you answer Y here, the uClibc native shared library loader will only load shared libraries, which do not need to modify any non-writable segments. These libraries haven't set the DT_TEXTREL tag in the dynamic section (==> objdump). So all your libraries must be compiled with -fPIC or -fpic, and all assembler function must be written as position independent code (PIC). Enabling this option will makes uClibc's shared library loader a little bit smaller and guarantee that no memory will be wasted by badly coded shared libraries. config UCLIBC_PIE_SUPPORT bool "Support ET_DYN in shared library loader" select FORCE_SHAREABLE_TEXT_SEGMENTS default n help If you answer Y here, the uClibc native shared library loader will support ET_DYN/PIE executables. It requires binutils-2.14.90.0.6 or later and the usage of the -pie option. More about ET_DYN/PIE binaries on . WARNING: This option also enables FORCE_SHAREABLE_TEXT_SEGMENTS, so all libraries have to be built with -fPIC or -fpic, and all assembler functions must be written as position independent code (PIC). config LDSO_LDD_SUPPORT bool "Native shared library loader 'ldd' support" depends on BUILD_UCLIBC_LDSO default y help Enable this to enable all the code needed to support traditional ldd, which executes the shared library loader to resolve all dependencies and then provide a list of shared libraries that are required for an application to function. Disabling this option will makes uClibc's shared library loader a little bit smaller. Most people will answer Y. config UCLIBC_CTOR_DTOR bool "Support global constructors and destructors" default y help If you wish to build uClibc with support for global constructor (ctor) and global destructor (dtor) support, then answer Y here. When ctor/dtor support is enabled, binaries linked with uClibc must also be linked with crtbegin.o and crtend.o which are provided by gcc (the "*startfile:" and "*endfile:" settings in your gcc specs file may need to be adjusted to include these files). This support will also add a small amount of additional size to each binary compiled vs uClibc. If you will be using uClibc with C++, or if you need the gcc __attribute__((constructor)) and __attribute__((destructor)) to work, then you definitely want to answer Y here. If you don't need ctors or dtors and want your binaries to be as small as possible, then answer N. config UCLIBC_PROPOLICE bool "Support for propolice stack protection" default n help Propolice stack protection. More about it on . To be able to use it, you'll also need a propolice patched gcc, supporting the -fstack-protector[-all] options. It is a specially patched gcc version, were __guard and __stack_smash_handler are removed from libgcc. Most people will answer N. config UCLIBC_PROFILING bool "Support gprof profiling" default y help If you wish to build uClibc with support for application profiling using the gprof tool, then you should enable this feature. Then in addition to building uClibc with profiling support, you will also need to recompile all your shared libraries with the profiling enabled version of uClibc. To add profiling support to your applications, you must compile things using the gcc options "-fprofile-arcs -pg". Then when you run your applications, a gmon.out file will be generated which can then be analyzed by 'gprof'. These exist a number of less invasive alternatives that do not require your to specially instrument your application, and recompile and relink everything. Many people have had good results using the combination of Valgrind to generate profiling information and KCachegrind for analysis: http://developer.kde.org/~sewardj/ http://kcachegrind.sourceforge.net/ The OProfile system-wide profiler is another alternative: http://oprofile.sourceforge.net/ Prospect is another alternative based on OProfile: http://prospect.sourceforge.net/ And the Linux Trace Toolkit (LTT) is also a fine tool: http://www.opersys.com/LTT/ If none of these tools do what you need, you can of course enable this option, rebuild everything, and use 'gprof'. There is both a size and performance penalty to profiling your applications this way, so most people should answer N. config HAS_NO_THREADS bool default n config UCLIBC_HAS_THREADS bool "POSIX Threading Support" depends on !HAS_NO_THREADS default y help If you want to compile uClibc with pthread support, then answer Y. This will increase the size of uClibc by adding a bunch of locking to critical data structures, and adding extra code to ensure that functions are properly reentrant. If your applications require pthreads, answer Y. config PTHREADS_DEBUG_SUPPORT bool "Build pthreads debugging support" default n depends on UCLIBC_HAS_THREADS help Say Y here if you wish to be able to debug applications that use uClibc's pthreads library. By enabling this option, a library named libthread_db will be built. This library will be dlopen()'d by gdb and will allow gdb to debug the threads in your application. IMPORTANT NOTE! Because gdb must dlopen() the libthread_db library, you must compile gdb with uClibc in order for pthread debugging to work properly. If you are doing development and want to debug applications using uClibc's pthread library, answer Y. Otherwise, answer N. config UCLIBC_HAS_LFS bool "Large File Support" default y depends on !CONFIG_CRIS help If you wish to build uClibc with support for accessing large files (i.e. files greater then 2 GiB) then answer Y. Do not enable this if you are using an older Linux kernel (2.0.x) that lacks large file support. Enabling this option will increase the size of uClibc. choice prompt "Malloc Implementation" default MALLOC if ! ARCH_HAS_MMU default MALLOC_STANDARD if ARCH_HAS_MMU help "malloc" use mmap for all allocations and so works very well on MMU-less systems that do not support the brk() system call. It is pretty smart about reusing already allocated memory, and minimizing memory wastage. This is the default for uClinux MMU-less systems. "malloc-simple" was written from scratch for uClibc, and is the simplest possible (and therefore smallest) malloc implementation. It is rather dumb, and certainly isn't the fastest. But it is 100% standards compliant, thread safe, and very small. "malloc-standard" is derived from the public domain dlmalloc implementation by Doug Lea. It is quite fast, and is pretty smart about reusing already allocated memory, and minimizing memory wastage. This uses brk() for small allocations, while using mmap() for larger allocations. This is the default malloc implementation for uClibc. If unsure, answer "malloc-standard". config MALLOC bool "malloc" config MALLOC_SIMPLE bool "malloc-simple" config MALLOC_STANDARD bool "malloc-standard" depends on ARCH_HAS_MMU endchoice config MALLOC_GLIBC_COMPAT bool "Malloc returns live pointer for malloc(0)" default n help The behavior of malloc(0) is listed as implementation-defined by SuSv3. Glibc returns a valid pointer to something, while uClibc normally return a NULL. I personally feel glibc's behavior is not particularly safe, and allows buggy applications to hide very serious problems. When this option is enabled, uClibc will act just like glibc, and return a live pointer when someone calls malloc(0). This pointer provides a malloc'ed area with a size of 1 byte. This feature is mostly useful when dealing with applications using autoconf's broken AC_FUNC_MALLOC macro (which redefines malloc as rpl_malloc if it does not detect glibc style returning-a-valid-pointer-for-malloc(0) behavior). Most people can safely answer N. config UCLIBC_DYNAMIC_ATEXIT bool "Dynamic atexit() Support" default y help When this option is enabled, uClibc will support an infinite number, of atexit() and on_exit() functions, limited only by your available memory. This can be important when uClibc is used with C++, since global destructors are implemented via atexit(), and it is quite possible to exceed the default number when this option is disabled. Enabling this option adds a few bytes, and more significantly makes atexit and on_exit depend on malloc, which can be bad when compiling static executables. Unless you use uClibc with C++, you should probably answer N. config HAS_SHADOW bool "Shadow Password Support" default y help Answer N if you do not need shadow password support. Most people will answer Y. config UNIX98PTY_ONLY bool "Support only Unix 98 PTYs" default y help If you want to support only Unix 98 PTYs enable this. Some older applications may need this disabled. For most current programs, you can generally answer Y. config ASSUME_DEVPTS bool "Assume that /dev/pts is a devpts or devfs file system" default y help Enable this if /dev/pts is on a devpts or devfs filesystem. Both these filesystems automatically manage permissions on the /dev/pts devices. You may need to mount your devpts or devfs filesystem on /dev/pts for this to work. Most people should answer Y. config UCLIBC_HAS_TM_EXTENSIONS bool "Support 'struct tm' timezone extension fields" default y help Enabling this option adds fields to 'struct tm' in time.h for tracking the number of seconds east of UTC, and an abbreviation for the current timezone. These fields are not specified by the SuSv3 standard, but they are commonly used in both GNU and BSD application code. To strictly follow the SuSv3 standard, leave this disabled. Most people will probably want to answer Y. config UCLIBC_HAS_TZ_CACHING bool "Enable caching of the last valid timezone 'TZ' string" default y help Answer Y to enable caching of the last valid 'TZ' string describing the timezone setting. This allows a quick string compare to avoid repeated parsing of unchanged 'TZ' strings when tzset() is called. Most people will answer Y. config UCLIBC_HAS_TZ_FILE bool "Enable '/etc/TZ' file support to set a default timezone (uClibc-specific)" default y help Answer Y to enable the setting of a default timezone for uClibc. Ordinarily, uClibc gets the timezone information exclusively from the 'TZ' environment variable. In particular, there is no support for the zoneinfo directory tree or the /etc/timezone file used by glibc. With this option enabled, uClibc will use the value stored in the file '/etc/TZ' (default path) to obtain timezone information if the 'TZ' environment variable is missing or has an invalid value. The file consists of a single line (newline required) of text describing the timezone in the format specified for the TZ environment variable. Simply doing 'echo CST6CDT > /etc/TZ' is enough to create a valid file. See http://www.opengroup.org/onlinepubs/007904975/basedefs/xbd_chap08.html for details on valid settings of 'TZ'. Most people will answer Y. config UCLIBC_HAS_TZ_FILE_READ_MANY bool "Repeatedly read the '/etc/TZ' file" depends on UCLIBC_HAS_TZ_FILE default y help Answer Y to enable repeated reading of the '/etc/TZ' file even after a valid value has been read. This incurs the overhead of an open/read/close for each tzset() call (explicit or implied). However, setting this will allows applications to update their timezone information if the contents of the file change. Most people will answer Y. config UCLIBC_TZ_FILE_PATH string "Path to the 'TZ' file for setting the global timezone" depends on UCLIBC_HAS_TZ_FILE default "/etc/TZ" help This is the path to the 'TZ' file. Most people will use the default of '/etc/TZ'. endmenu menu "Networking Support" config UCLIBC_HAS_IPV6 bool "IP version 6 Support" default n help If you want to include support for the next version of the Internet Protocol (IP version 6) then answer Y. Most people should answer N. config UCLIBC_HAS_RPC bool "Remote Procedure Call (RPC) support" default n help If you want to include RPC support, enable this. RPC is rarely used for anything except for the NFS filesystem. Unless you plan to use NFS, you can probably leave this set to N and save some space. If you need to use NFS then you should answer Y. config UCLIBC_HAS_FULL_RPC bool "Full RPC support" depends on UCLIBC_HAS_RPC default y if !HAVE_SHARED help Normally we enable just enough RPC support for things like rshd and nfs mounts to work. If you find you need the rest of the RPC stuff, then enable this option. Most people can safely answer N. endmenu menu "String and Stdio Support" config UCLIBC_HAS_CTYPE_TABLES bool "Use Table Versions Of 'ctype.h' Functions." default y help Answer Y to use table versions of the 'ctype.h' functions. While the non-table versions are often smaller when building staticly linked apps, they work only in stub locale mode. Most people will answer Y. config UCLIBC_HAS_CTYPE_SIGNED bool "Support Signed Characters In 'ctype.h' Functions." depends UCLIBC_HAS_CTYPE_TABLES default y help Answer Y to enable support for passing signed char values to the 'ctype.h' functions. ANSI/ISO C99 and SUSv3 specify that these functions are only defined for unsigned char values and EOF. However, glibc allows negative signed char values as well in order to support 'broken old programs'. Most people will answer Y. choice prompt "ctype argument checking" depends UCLIBC_HAS_CTYPE_TABLES default UCLIBC_HAS_CTYPE_UNSAFE help Please select the invalid arg behavior you want for the 'ctype' functions. The 'ctype' functions are now implemented using table lookups, with the arg being the index. This can result in incorrect memory accesses or even segfaults for args outside of the allowed range. NOTE: This only affects the 'ctype' _functions_. It does not affect the macro implementations. config UCLIBC_HAS_CTYPE_UNSAFE bool "Do not check -- unsafe" config UCLIBC_HAS_CTYPE_CHECKED bool "Detect and handle appropriately" config UCLIBC_HAS_CTYPE_ENFORCED bool "Issue a diagnostic and abort()" endchoice config UCLIBC_HAS_WCHAR bool "Wide Character Support" default n help Answer Y to enable wide character support. This will make uClibc much larger. It is also currently required for locale support. Most people will answer N. config UCLIBC_HAS_LOCALE bool "Locale Support" select UCLIBC_HAS_WCHAR select UCLIBC_HAS_CTYPE_TABLES default n help uClibc now has full ANSI/ISO C99 locale support (except for wcsftime() and collating items in regex). Be aware that enabling this option will make uClibc much larger. Enabling UCLIBC_HAS_LOCALE with the default set of supported locales (169 UTF-8 locales, and 144 locales for other codesets) will enlarge uClibc by around 300k. You can reduce this size by building your own custom set of locate data (see extra/locale/LOCALES for details). uClibc's locale support is still under development. For example, codesets using shift states are not currently supported. Support is planned in the next iteration of locale support. Answer Y to enable locale support. Most people will answer N. config UCLIBC_PREGENERATED_LOCALE_DATA bool "Use Pre-generated Locale Data" depends on UCLIBC_HAS_LOCALE default n help If you are selective and only want locale data for a few particular locales, or you enjoy pain, or you are a rabid do-it-yourself sort of person, you can turn this option off and manually walk through the mostly undocumented procedure needed to generate your own locale data. Mere mortals will answer Y and use the default set of pregenerated locale data, which supports 169 UTF-8 locales, and 144 locales for other codesets (for the complete list see extra/locale/LOCALES). config UCLIBC_DOWNLOAD_PREGENERATED_LOCALE_DATA bool "Automagically Download the Pre-generated Locale Data (if necessary)" depends on UCLIBC_PREGENERATED_LOCALE_DATA default n help If you would like the build process to use 'wget' to automatically download the pregenerated locale data, enable this option. Otherwise you will need to obtain the locale data yourself from: http://www.uclibc.org/downloads/uClibc-locale-030818.tgz and place the uClibc-locale-030818.tgz tarball in the extra/locale/ directory. Go ahead and make life easy for yourself... Answer Y. config UCLIBC_HAS_XLOCALE bool "Extended Locale Support (experimental/incomplete)" depends on UCLIBC_HAS_LOCALE default n help Answer Y to enable extended locale support similar to that provided by glibc. This is primarily intended to support libstd++ functionality. However, it also allows thread-specific locale selection via uselocale(). Most people will answer N. config UCLIBC_HAS_HEXADECIMAL_FLOATS bool "Support hexadecimal float notation" depends UCLIBC_HAS_CTYPE_TABLES depends on UCLIBC_HAS_FLOATS default n help Answer Y to enable support for hexadecimal float notation in the (wchar and) char string to floating point conversion functions, as well as support for the %a and %A conversion specifiers in the *printf() and *scanf() functions. Most people will answer N. config UCLIBC_HAS_GLIBC_DIGIT_GROUPING bool "Support glibc's \"'\" flag for allowing locale-specific digit grouping" depends on UCLIBC_HAS_LOCALE depends on UCLIBC_HAS_FLOATS default n help Answer Y to enable support for glibc's \"'\" flag for allowing locale-specific digit grouping in base 10 integer conversions and appropriate floating point conversions in the *printf() and *scanf() functions. Most people will answer N. config UCLIBC_HAS_SCANF_LENIENT_DIGIT_GROUPING bool "Do not require digit grouping when the \"'\" flag is specified" depends on UCLIBC_HAS_GLIBC_DIGIT_GROUPING default y help Answer Y to make digit grouping optional when the \"'\" flag is specified. This is the standard glibc behavior. If the initial string of digits exceeds the maximum group number, the input will be treated as a normal non-grouped number. Most people will answer N. config UCLIBC_HAS_GLIBC_CUSTOM_PRINTF bool "Support glibc's register_printf_function() (glibc-compat)" depends on !USE_OLD_VFPRINTF default n help Answer Y to support glibc's register_printf_function() to allow an application to add its own printf conversion specifiers. NOTE: This implementation limits the number or registered specifiers to 10. NOTE: This implementation requires new conversion specifiers to be ASCII characters (0-0x7f). This is to avoid problems with processing format strings in locales with different multibyte conversions. Most people will answer N. config USE_OLD_VFPRINTF bool "Use the old vfprintf implementation" depends on !UCLIBC_HAS_WCHAR default n help Set to true to use the old vfprintf instead of the new. This is roughly C89 compliant with some extensions, and is much smaller. However, it does not support wide chars, positional args, or glibc custom printf specifiers. Most people will answer N. config UCLIBC_PRINTF_SCANF_POSITIONAL_ARGS int "Maximum number of positional args. Either 0 or >= 9." depends on !USE_OLD_VFPRINTF default 9 help Set the maximum number of positional args supported by the printf/scanf functions. The Single Unix Specification Version 3 requires a minimum value of 9. Setting this to a value lower than 9 will disable positional arg support and cause the NL_ARGMAX macro in limits.h to be #undef'd. WARNING! The workspace to support positional args is currently allocated on the stack. You probably don't want to set this to too high a value. Most people will answer 9. config UCLIBC_HAS_SCANF_GLIBC_A_FLAG bool "Support glibc's 'a' flag for scanf string conversions" default n help NOTE!!! Currently Not Implemented!!! Just A Place Holder!! NOTE!!! Answer Y to enable support for glibc's 'a' flag for the scanf string conversions '%s', '%[', '%ls', '%l[', and '%S'. This is used to auto-allocate sufficient memory to hold the data retrieved. Most people will answer N. choice prompt "Stdio buffer size" default UCLIBC_HAS_STDIO_BUFSIZ_4096 help Please select a value for BUFSIZ. This will be used by the stdio subsystem as the default buffer size for a file, and affects fopen(), setvbuf(), etc. NOTE: Setting this to 'none' will disable buffering completely. However, BUFSIZ will still be defined in stdio.h as 256 because many applications use this value. config UCLIBC_HAS_STDIO_BUFSIZ_NONE bool "none (WARNING - BUFSIZ will be 256 in stdio.h)" depends !UCLIBC_HAS_WCHAR config UCLIBC_HAS_STDIO_BUFSIZ_256 bool "256 (minimum ANSI/ISO C99 value)" config UCLIBC_HAS_STDIO_BUFSIZ_512 bool "512" config UCLIBC_HAS_STDIO_BUFSIZ_1024 bool "1024" config UCLIBC_HAS_STDIO_BUFSIZ_2048 bool "2048" config UCLIBC_HAS_STDIO_BUFSIZ_4096 bool "4096" config UCLIBC_HAS_STDIO_BUFSIZ_8192 bool "8192" # If you add more choices, you will need to update uClibc_stdio.h. endchoice choice prompt "Stdio builtin buffer size (uClibc-specific)" depends !UCLIBC_HAS_STDIO_BUFSIZ_NONE default UCLIBC_HAS_STDIO_BUILTIN_BUFFER_NONE help When a FILE is created with fopen(), an attempt is made to allocate a BUFSIZ buffer for it. If the allocation fails, fopen() will still succeed but the FILE will be unbuffered. This option adds a small amount of space to each FILE to act as an emergency buffer in the event of a buffer allocation failure. Most people will answer None. config UCLIBC_HAS_STDIO_BUILTIN_BUFFER_NONE bool "None" config UCLIBC_HAS_STDIO_BUILTIN_BUFFER_4 bool "4" config UCLIBC_HAS_STDIO_BUILTIN_BUFFER_8 bool "8" # If you add more choices, you will need to update uClibc_stdio.h. endchoice config UCLIBC_HAS_STDIO_GETC_MACRO bool "Provide a macro version of getc()" depends !UCLIBC_HAS_STDIO_BUFSIZ_NONE default y help Provide a macro version of getc(). Most people will answer Y. config UCLIBC_HAS_STDIO_PUTC_MACRO bool "Provide a macro version of putc()" depends !UCLIBC_HAS_STDIO_BUFSIZ_NONE default y help Provide a macro version of putc(). Most people will answer Y. config UCLIBC_HAS_STDIO_AUTO_RW_TRANSITION bool "Support auto-r/w transition" default y help Answer Y to enable the stdio subsystem to automaticly transition between reading and writing. This relaxes the ANSI/ISO C99 requirement: When a file is opened with update mode ('+' as the second or third character in the list of mode argument values), both input and output may be performed on the associated stream. However, output shall not be directly followed by input without an intervening call to the fflush function or to a file positioning function (fseek, fsetpos, or rewind), and input shall not be directly followed by output without an intervening call to a file positioning function, unless the input operation encounters end­of­file. Most people will answer Y. config UCLIBC_HAS_FOPEN_LARGEFILE_MODE bool "Support an fopen() 'F' flag for large file mode (uClibc-specific)" depends on UCLIBC_HAS_LFS default n help Answer Y to enable a uClibc-specific extension to allow passing an additional 'F' flag in the mode string for fopen() to specify that the file should be open()ed with the O_LARGEFILE flag set. Most people will answer N. config UCLIBC_HAS_FOPEN_EXCLUSIVE_MODE bool "Support an fopen() 'x' flag for exclusive mode (glibc-compat)" default n help Answer Y to support a glibc extension to allow passing additional 'x' flag in the mode string for fopen() to specify that the file should be open()ed with the O_EXCL flag set. Most people will answer N. config UCLIBC_HAS_GLIBC_CUSTOM_STREAMS bool "Support fmemopen(), open_memstream(), and fopencookie() (glibc-compat)" default n help Answer Y to support the glibc 'custom stream' extension functions fmemopen(), open_memstream(), and fopencookie(). NOTE: There are some minor differences regarding seeking behavior. Most people will answer N. config UCLIBC_HAS_PRINTF_M_SPEC bool "Support the '%m' specifier in printf format strings (glibc-compat)" default n help Answer Y to support a glibc extension to interpret '%m' in printf format strings as an instruction to output the error message string (as generated by strerror) corresponding to the current value of 'errno'. Most people will answer N. config UCLIBC_HAS_ERRNO_MESSAGES bool "Include the errno message text in the library" default y help Answer Y if you want to include the errno message text in the library. This adds about 3K to the library, but enables strerror() to generate text other than 'Unknown error '. Most people will answer Y. config UCLIBC_HAS_SYS_ERRLIST bool "Support sys_errlist[] (obsolete-compat)" depends on UCLIBC_HAS_ERRNO_MESSAGES default n help Answer Y if you want to support the obsolete sys_errlist[]. This adds about 0.5k to the library, except for the mips arch where it adds over 4K. WARNING! In the future, support for sys_errlist[] may be unavailable in at least some configurations. In fact, it may be removed altogether. Most people will answer N. config UCLIBC_HAS_SIGNUM_MESSAGES bool "Include the signum message text in the library" default y help Answer Y if you want to include the signum message text in the library. This adds about 0.5K to the library, but enables strsignal() to generate text other than 'Unknown signal '. Most people will answer Y. config UCLIBC_HAS_SYS_SIGLIST bool "Support sys_siglist[] (bsd-compat)" depends on UCLIBC_HAS_SIGNUM_MESSAGES default n help Answer Y if you want to support sys_siglist[]. WARNING! In the future, support for sys_siglist[] may be unavailable in at least some configurations. In fact, it may be removed altogether. Most people will answer N. config UCLIBC_HAS_GETTEXT_AWARENESS bool "Include gettext awareness" depends on UCLIBC_HAS_LOCALE default n help NOTE!!! Not yet integrated with strerror and strsignal. NOTE!!! Answer Y if you want to include weak stub gettext support and make the *strerror*() and strsignal() functions gettext-aware. Currently, to get functional gettext functionality you will need to use gnu gettext. Most people will answer N. config UCLIBC_HAS_GNU_GETOPT bool "Support gnu getopt" default y help Answer Y if you want to include full gnu getopt() instead of a (much smaller) SUSv3 compatible getopt(). Most people will answer Y. endmenu menu "Big and Tall" config UCLIBC_HAS_REGEX bool "Regular Expression Support" default y help POSIX regular expression code is really big -- 27k all by itself. If you don't use regular expressions, turn this off and save space. Of course, if you only staticly link, leave this on, since it will only be included in your apps if you use regular expressions. config UCLIBC_HAS_WORDEXP bool "Support the wordexp() interface" default n help The SuSv3 wordexp() interface performs word expansions per the Shell and Utilities volume of IEEE Std 1003.1-2001, Section 2.6. It is intended for use by applications that want to implement all of the standard Bourne shell expansions on input data. This interface is rarely used, and very large. Unless you have a pressing need for wordexp(), you should probably answer N. config UCLIBC_HAS_FTW bool "Support the ftw() and nftw() interfaces" default n help The SuSv3 ftw() and nftw() interfaces are used to recursively descend directory paths while repeatedly calling a function. This interface is rarely used, and adds around 4.5k. Unless you have a pressing need for ftw() or nftw(), you should probably answer N. config UCLIBC_HAS_GLOB bool "Support the glob() interface" default y help The glob interface is somewhat large (weighing in at about 4k). It is used fairly often, but is an option since people wanting to go for absolute minimum size may wish to omit it. Most people will answer Y. endmenu menu "Library Installation Options" config SHARED_LIB_LOADER_PREFIX string "Shared library loader path" depends on BUILD_UCLIBC_LDSO default "$(DEVEL_PREFIX)/lib" help When using shared libraries, this path is the location where the shared library will be invoked. This value will be compiled into every binary compiled with uClibc. For a typical target system this should be set to "/lib", such that 'make install' will install /lib/ld-uClibc.so.0. BIG FAT WARNING: If you do not have a shared library loader with the correct name sitting in the directory this points to, your binaries will not run. config SYSTEM_LDSO string "System shared library loader" depends on HAVE_SHARED && !BUILD_UCLIBC_LDSO default "/lib/ld-linux.so.2" help If you are using shared libraries, but do not want/have a native uClibc shared library loader, please specify the name of your target system's shared library loader here... BIG FAT WARNING: If you do not have a shared library loader with the correct name sitting in the directory this points to, your binaries will not run. config RUNTIME_PREFIX string "uClibc runtime library directory" default "/usr/$(TARGET_ARCH)-linux-uclibc/" help RUNTIME_PREFIX is the directory into which the uClibc runtime libraries will be installed. The result will look something like the following: $(RUNTIME_PREFIX)/ lib/ usr/bin/ldd sbin/ldconfig This value is used by the 'make install' Makefile target. Since this directory is compiled into the shared library loader, you will need to recompile uClibc if you change this value... For a typical target system this should be set to "/", such that 'make install' will install /lib/libuClibc-.so config DEVEL_PREFIX string "uClibc development environment directory" default "/usr/$(TARGET_ARCH)-linux-uclibc/usr/" help DEVEL_PREFIX is the directory into which the uClibc development environment will be installed. The result will look something like the following: $(DEVEL_PREFIX)/ lib/ include/ This value is used by the 'make install' Makefile target when installing a uClibc development environment. For a typical target system this should be set to "/usr", such that 'make install' will install /usr/include/
. endmenu menu "uClibc development/debugging options" config DODEBUG bool "Build uClibc with debugging symbols" default n help Say Y here if you wish to compile uClibc with debugging symbols. This will allow you to use a debugger to examine uClibc internals while applications are running. This increases the size of the library considerably and should only be used when doing development. If you are doing development and want to debug uClibc, answer Y. Otherwise, answer N. config DOASSERTS bool "Build uClibc with run-time assertion testing" default n help Say Y here to include runtime assertion tests. This enables runtime assertion testing in some code, which can increase the size of the library and incur runtime overhead. If you say N, then this testing will be disabled. config SUPPORT_LD_DEBUG bool "Build the shared library loader with debugging support" depends on BUILD_UCLIBC_LDSO default n help Answer Y here to enable all the extra code needed to debug the uClibc native shared library loader. The level of debugging noise that is generated depends on the LD_DEBUG environment variable... Just set LD_DEBUG to something like: 'LD_DEBUG=token1,token2,.. prog' to debug your application. Diagnostic messages will then be printed to the stderr. For now these debugging tokens are available: detail provide more information for some options move display copy processing symbols display symbol table processing reloc display relocation processing; detail shows the relocation patch nofixups never fixes up jump relocations bindings displays the resolve processing (function calls); detail shows the relocation patch all Enable everything! The additional environment variable: LD_DEBUG_OUTPUT=file redirects the diagnostics to an output file created using the specified name and the process id as a suffix. An excellent start is simply: $ LD_DEBUG=binding,move,symbols,reloc,detail ./appname or to log everything to a file named 'logfile', try this $ LD_DEBUG=all LD_DEBUG_OUTPUT=logfile ./appname If you are doing development and want to debug uClibc's shared library loader, answer Y. Mere mortals answer N. config SUPPORT_LD_DEBUG_EARLY bool "Build the shared library loader with early debugging support" depends on BUILD_UCLIBC_LDSO default n help Answer Y here to if you find the uClibc shared library loader is crashing or otherwise not working very early on. This is typical only when starting a new port when you haven't figured out how to properly get the values for argc, argv, environ, etc. This method allows a degree of visibility into the very early shared library loader initialization process. If you are doing development and want to debug the uClibc shared library loader early initialization, answer Y. Mere mortals answer N. config UCLIBC_MALLOC_DEBUGGING bool "Build malloc with debugging support" depends MALLOC default n help Answer Y here to compile extra debugging support code into malloc. Malloc debugging output may then be enabled at runtime using the MALLOC_DEBUG environment variable. The value of MALLOC_DEBUG should be an integer, which is interpreted as a bitmask with the following bits: 1 - do extra consistency checking 2 - output messages for malloc/free calls and OS allocation calls 4 - output messages for the `MMB' layer 8 - output messages for internal malloc heap manipulation calls Because this increases the size of malloc appreciably (due to strings etc), you should say N unless you need to debug a malloc problem. config UCLIBC_MJN3_ONLY bool "Manuel's hidden warnings" default n help Answer Y here to see all Manuel's personal notes, warnings, and todos. Most people will answer N. endmenu