228 lines · plain
1.. _printf_behavior:2 3====================================4Printf Behavior Under All Conditions5====================================6 7Introduction:8=============9On the "defining undefined behavior" page, I said you should write down your10decisions regarding undefined behavior in your functions. This is that document11for my printf implementation.12 13Unless otherwise specified, the functionality described is aligned with the ISO14C standard and POSIX standard. If any behavior is not mentioned here, it should15be assumed to follow the behavior described in those standards.16 17The LLVM-libc codebase is under active development, and may change. This18document was last updated [January 8, 2024] by [michaelrj] and may19not be accurate after this point.20 21The behavior of LLVM-libc's printf is heavily influenced by compile-time flags.22Make sure to check what flags are defined before filing a bug report. It is also23not relevant to any other libc implementation of printf, which may or may not24share the same behavior.25 26This document assumes familiarity with the definition of the printf function and27is intended as a reference, not a replacement for the original standards.28 29--------------30General Flags:31--------------32These compile-time flags will change the behavior of LLVM-libc's printf when it33is compiled. Combinations of flags that are incompatible will be marked.34 35LIBC_COPT_STDIO_USE_SYSTEM_FILE36-------------------------------37When set, this flag changes fprintf and printf to use the FILE API from the38system's libc, instead of LLVM-libc's internal FILE API. This is set by default39when LLVM-libc is built in overlay mode.40 41LIBC_COPT_PRINTF_DISABLE_INDEX_MODE42-----------------------------------43When set, this flag disables support for the POSIX "%n$" format, hereafter44referred to as "index mode"; conversions using the index mode format will be45treated as invalid. This reduces code size.46 47LIBC_COPT_PRINTF_INDEX_ARR_LEN48------------------------------49This flag takes a positive integer value, defaulting to 128. This flag50determines the number of entries the parser's type descriptor array has. This is51used in index mode to avoid re-parsing the format string to determine types when52an index lower than the previously specified one is requested. This has no53effect when index mode is disabled.54 55LIBC_COPT_PRINTF_DISABLE_WRITE_INT56----------------------------------57When set, this flag disables support for the C Standard "%n" conversion; any58"%n" conversion will be treated as invalid. This is set by default to improve59security.60 61LIBC_COPT_PRINTF_DISABLE_FLOAT62------------------------------63When set, this flag disables support for floating point numbers and all their64conversions (%a, %f, %e, %g); any floating point number conversion will be65treated as invalid. This reduces code size.66 67LIBC_COPT_PRINTF_DISABLE_FIXED_POINT68------------------------------------69When set, this flag disables support for fixed point numbers and all their70conversions (%r, %k); any fixed point number conversion will be treated as71invalid. This reduces code size. This has no effect if the current compiler does72not support fixed point numbers.73 74LIBC_COPT_PRINTF_NO_NULLPTR_CHECKS75----------------------------------76When set, this flag disables the nullptr checks in %n and %s.77 78LIBC_COPT_PRINTF_CONV_ATLAS79---------------------------80When set, this flag changes the include path for the "converter atlas" which is81a header that includes all the files containing the conversion functions. This82is not recommended to be set without careful consideration.83 84LIBC_COPT_PRINTF_HEX_LONG_DOUBLE85--------------------------------86When set, this flag replaces all decimal long double conversions (%Lf, %Le, %Lg)87with hexadecimal long double conversions (%La). This will improve performance88significantly, but may cause some tests to fail. This has no effect when float89conversions are disabled.90 91--------------------------------92Float Conversion Internal Flags:93--------------------------------94The following floating point conversion flags are provided for reference, but95are not recommended to be adjusted except by persons familiar with the Printf96Ryu Algorithm. Additionally they have no effect when float conversions are97disabled.98 99LIBC_COPT_FLOAT_TO_STR_NO_SPECIALIZE_LD100---------------------------------------101This flag disables the separate long double conversion implementation. It is102not based on the Ryu algorithm, instead generating the digits by103multiplying/dividing the written-out number by 10^9 to get blocks. It's104significantly faster than INT_CALC, only about 10x slower than MEGA_TABLE,105and is small in binary size. Its downside is that it always calculates all106of the digits above the decimal point, making it slightly inefficient for %e107calls with large exponents. This is the default. This specialization overrides108other flags, so this flag must be set for other flags to effect the long double109behavior.110 111LIBC_COPT_FLOAT_TO_STR_USE_MEGA_LONG_DOUBLE_TABLE112-------------------------------------------------113When set, the float to string decimal conversion algorithm will use a larger114table to accelerate long double conversions. This larger table is around 5MB of115size when compiled.116 117LIBC_COPT_FLOAT_TO_STR_USE_DYADIC_FLOAT118---------------------------------------119When set, the float to string decimal conversion algorithm will use dyadic120floats instead of a table when performing floating point conversions. This121results in ~50 digits of accuracy in the result, then zeroes for the remaining122values. This may improve performance but may also cause some tests to fail. The123flag ending in _LD is the same, but only applies to long double decimal124conversions.125 126LIBC_COPT_FLOAT_TO_STR_USE_INT_CALC127-----------------------------------128When set, the float to string decimal conversion algorithm will use wide129integers instead of a table when performing floating point conversions. This130gives the same results as the table, but is very slow at the extreme ends of131the long double range.132 133LIBC_COPT_FLOAT_TO_STR_NO_TABLE134-------------------------------135When set, the float to string decimal conversion algorithm will not use either136the mega table or the normal table for any conversions. Instead it will set137algorithmic constants to improve performance when using calculation algorithms.138If this flag is set without any calculation algorithm flag set, an error will139occur.140 141--------142Parsing:143--------144 145When printf encounters an invalid conversion specification, the entire146conversion specification will be passed literally to the output string.147As an example, printf("%Z") would display "%Z".148 149If an index mode conversion is requested for index "n" and there exists a number150in [1,n) that does not have a conversion specified in the format string, then151the conversion for index "n" is considered invalid.152 153If a non-index mode (also referred to as sequential mode) conversion is154specified after an index mode conversion, the next argument will be read but the155current index will not be incremented. From this point on, the arguments156selected by each conversion may or may not be correct. This is considered157dangerously undefined and may change without warning.158 159If a conversion specification is provided an invalid type modifier, that type160modifier will be ignored, and the default type for that conversion will be used.161In the case of the length modifier "L" and integer conversions, it will be162treated as if it was "ll" (lowercase LL). For this purpose the list of integer163conversions is d, i, u, o, x, X, b, B, n.164 165If a conversion specification ending in % has any options that consume arguments166(e.g. "%*.*%") those arguments will be consumed as normal, but their values will167be ignored.168 169If a conversion specification ends in a null byte ('\0') then it shall be170treated as an invalid conversion followed by a null byte.171 172If a number passed as a field width or precision value is out of range for an173int, then it will be treated as the largest value in the int range174(e.g. "%-999999999999.999999999999s" is the same as "%-2147483647.2147483647s").175 176If the field width is set to INT_MIN by using the '*' form,177e.g. printf("%*d", INT_MIN, 1), it will be treated as INT_MAX, since -INT_MIN is178not representable as an int.179 180If a number passed as a bit width is less than or equal to zero, the conversion181is considered invalid. If the provided bit width is larger than the width of182uintmax_t, it will be clamped to the width of uintmax_t.183 184----------185Conversion186----------187Any conversion specification that contains a flag or option that it does not188have defined behavior for will ignore that flag or option (e.g. %.5c is the same189as %c).190 191If a conversion specification ends in %, then it will be treated as if it is192"%%", ignoring all options.193 194If a null pointer is passed to a %s conversion specification and null pointer195checks are enabled, it will be treated as if the provided string is "null".196 197If a null pointer is passed to a %n conversion specification and null pointer198checks are enabled, the conversion will fail and printf will return a negative199value.200 201If a null pointer is passed to a %p conversion specification, the string202"(nullptr)" will be returned instead of an integer value.203 204The %p conversion will display any non-null pointer as if it was a uintptr value205passed to a "%#tx" conversion, with all other options remaining the same as the206original conversion.207 208The %p conversion will display a null pointer as if it was the string209"(nullptr)" passed to a "%s" conversion, with all other options remaining the210same as the original conversion.211 212The %r, %R, %k, and %K fixed point number format specifiers are accepted as213defined in ISO/IEC TR 18037 (the fixed point number extension). These are214available when the compiler is detected as having support for fixed point215numbers and the LIBC_COPT_PRINTF_DISABLE_FIXED_POINT flag is not set.216 217The %m conversion will behave as specified by POSIX for syslog: It takes no218arguments, and outputs the result of strerror(errno). Additionally, to match219existing printf behaviors, it will behave as if it is a %s string conversion for220the purpose of all options, except for the alt form flag. If the alt form flag221is specified, %m will instead output a string matching the macro name of the222value of errno (e.g. "ERANGE" for errno = ERANGE), again treating it as a string223conversion. If there is no corresponding macro, then alt form %m will print the224value of errno as an integer with the %d format, including all options. If225errno = 0 and alt form is specified, the conversion will be a string conversion226on "0" for simplicity of implementation. This matches what other libcs227implementing this feature have done.228