GeneralPurpose-FormatSpecifier

Validates the use of format specifiers

Required inputs: IR

This rule checks for incorrect use of format specifiers with the printf/scanf family of functions.

Due to these functions using C varargs, compilers cannot check that the parameter types expected by the function. In cases where the format argument to these function is a string literal, this rule will verify that the remaining arguments are compatible with those expected by the format string. This rule also checks for invalid combination of flags/modifiers and conversion specifiers.

Bad code:
#include <stdio.h>

void f() {
    const char *str = "Hello, World!";
    int num = 42;
    int count;
    char buf1[10];
    char *buf2;

    // printf
    printf("%s: %d\n", str); // ERROR: format '%d' expects a matching 'signed int' argument
    printf("%s: %d\n", str, num, num); // ERROR: Too many arguments for format.
    printf("%ld\n", num); // ERROR: format '%ld' expects argument of type 'signed long', but argument 1 has type 'signed int'
    printf("%Ld\n", num); // ERROR: %d does not support the 'L' length modifier
    printf("%0\n", num); // ERROR: Invalid or non-standard conversion specification '%0'
    printf("%.*p\n", num, str); // ERROR: Precision must not be used with %p conversion specifier
    printf("%#s\n", str); // ERROR: %s does not support the '#' flag
    printf("%0s\n", str); // ERROR: %s does not support the '0' flag
    printf("%1%"); // ERROR: Cannot use any flags or modifiers with '%%'

    printf("%s%+n", str, &count); // ERROR: %n does not support flags
    printf("%s%1n", str, &count); // ERROR: %n does not support field width

    // printf GNU extensions:
    printf("%'s\n", str); // ERROR: %s does not support the "'" flag
    printf("%Is\n", str); // ERROR: %s does not support the 'I' flag

    // scanf
    scanf("%s", buf1); // ERROR: Potential buffer overflow: format '%s' has no limit on amount of characters read.
    scanf("%10s", buf1); // ERROR: format '%s' may write up to 11 characters to buffer of size 10.
    scanf("%9s", buf2); // ERROR: Potential buffer overflow: format '%s' used with buffer of unknown size.
    scanf("%d%*n", &num, &count); // ERROR: %n does not support assignment suppression
}
Good code:
#include <stdio.h>

void f() {
    const char *str = "Hello, World!";
    int num = 42;
    double pi = 3.14159;
    int count;
    char buf1[10];

    // printf
    printf("%s: %d\n", str, num); // OK: format and argument count match
    printf("%d\n", num); // OK: format '%d' matches 'signed int' argument
    printf("%5d\n", num); // OK: field width of 5
    printf("%05d\n", num); // OK: zero-padded to width 5
    printf("%-5d\n", num); // OK: left-aligned to width 5
    printf("%+d\n", num); // OK: show sign
    printf("%.2f\n", pi); // OK: precision of 2 decimal places
    printf("%8.2f\n", pi); // OK: width 8, precision 2
    printf("%p\n", (void*)str); // OK: no precision with %p
    printf("%%\n"); // OK: no flags with '%%'
    printf("%s%n\n", str, &count); // OK: %n without flags

    // printf GNU extensions:
    printf("%'d\n", num); // OK: thousand separator
    printf("%Id\n", num); // OK: locale specific digits

    // scanf
    scanf("%9s", buf1); // OK: limited to 9 characters for 10-byte buffer
    scanf("%d%n", &num, &count); // OK: %n without assignment suppression
}

Possible Messages

Key

Text

Severity

Disabled

arg_type_mismatch

{} expects argument of type ‘{}’, but argument {} has type ‘{}’

None

False

buffer_too_small

{} may write up to {} characters to buffer of size {}.

None

False

invalid_conversion

Invalid or non-standard conversion specification

None

False

matching_arg_expected

{} expects a matching ‘{}’ argument

None

False

precision_for_conversion

Precision must not be used with %{} conversion specifier

None

False

too_many_args

Too many arguments for format.

None

False

unlimited_read

Potential buffer overflow: {} has no limit on amount of characters read.

None

False

unsupported_assignment_suppression

%n does not support assignment suppression

None

False

unsupported_field_width

%n does not support field width

None

False

unsupported_flags

%n does not support flags

None

False

unsupported_flags_modifiers

Cannot use any flags or modifiers with ‘%%’

None

False

unsupported_hash

%{} does not support the ‘#’ flag

None

False

unsupported_i_flag

%{} does not support the ‘I’ flag

None

False

unsupported_length_modifier

%{} does not support the ‘{}’ length modifier

None

False

unsupported_tick

%{} does not support the “’” flag

None

False

unsupported_zero

%{} does not support the ‘0’ flag

None

False

Options

allow_extra_args

allow_extra_args : bool = False

Whether to allow additional arguments that are not used by the format string.
 

allow_gnu_extensions

allow_gnu_extensions : bool = True

Whether to allow the GNU extensions to format specifications.
 

allow_unknown_specs

allow_unknown_specs : bool = False

Whether to allow unknown format specifications. It may be necessary to set this option when using implementation-specific extensions. Arguments are not checked when the format string contains unknown format specifications.
 

functions

functions

Type: dict[bauhaus.analysis.config.QualifiedName, typing.Tuple[str, int, typing.Optional[int]]]

Default:

{
   '_printf_l': ('printf', 1, 3),
   'fprintf': ('printf', 1, 2),
   'fscanf': ('scanf', 1, 2),
   'printf': ('printf', 0, 1),
   'scanf': ('scanf', 0, 1),
   'snprintf': ('printf', 2, 3),
   'sprintf': ('printf', 1, 2),
   'sscanf': ('scanf', 1, 2),
   'vfprintf': ('printf', 1, None),
   'vfscanf': ('scanf', 1, None),
   'vprintf': ('printf', 0, None),
   'vscanf': ('scanf', 0, None),
   'vsnprintf': ('printf', 2, None),
   'vsprintf': ('printf', 1, None),
   'vsscanf': ('scanf', 1, None)
}
A dictionary mapping the names of the functions to check, to a triple (function_kind, fmt_param_index, arg_start_index) where function_kind is either printf or scanf, fmt_param_index is the index of the format-string parameter, and arg_start_index is the index of the first variadic argument.
 

use_static_semantic_analysis

use_static_semantic_analysis : bool = True

Whether the rule should use the results of the StaticSemanticAnalysis to check buffer sizes. This can produce different findings, as the size of buffer arguments are computed more accurately. This will not enforce StaticSemanticAnalysis to be enabled, but will produce less accurate results if it is not.