warning "operation on <variable> may be undefined" - g++

When I use overloading [][] operators in c++ to create a minimal matrix class
class matrix {
private:
vector<T> elems_;
size_t nrows_;
size_t ncols_;
public:
T const* operator[] ( size_t const r ) const { return &elems_[r * ncols_]; }
T* operator[] ( size_t const r ) { return &elems_[r * ncols_]; }
matrix ();
matrix ( size_t const nr, size_t const nc )
: elems_( nr * nc ), nrows_( nr ), ncols_( nc )
{ }
matrix ( size_t const nr, size_t const nc, T const *data)
: elems_( nr * nc ), nrows_( nr ), ncols_( nc )
{ size_t ptr=0;
for (int i=0;i<nr;i++)
for (int j=0;j<nc;j++)
elems_[ptr] = data[ptr++];
}
}
g++ returns the warning operation on ‘ptr’ may be undefined [-Wsequence-point]. In previous post
Why I got "operation may be undefined" in Statement Expression in C++? it is explained that the last thing in the compound statement should be an expression followed by a semicolon but not which purpose it serves. Does it mean that g++ will always throw this warning for compound statements that have a for-loop at the end?
If so can anyone explain why this is useful? I cannot think of any reason why anyone should be advised against ending a compound statement with a for loop.

In elems_[ptr] = data[ptr++], it is undefined whether elems_[ptr] is evaluated first or data[ptr++] is evaluated first. This is so, because = does not introduce a sequence point.
Depending on the order, elems_[ptr] = data[ptr++] yields different results. Hence the warning.

Related

SMHasher setup?

The SMHasher test suite for hash functions is touted as the best of the lot. But the latest version I've got (from rurban) gives absolutely no clue on how to check your proposed hash function (it does include an impressive battery of hash functions, but some of interest --if only for historic value-- are missing). Add that I'm a complete CMake newbie.
It's actually quite simple. You just need to install CMake.
Building SMHasher
To build SMHasher on a Linux/Unix machine:
git clone https://github.com/rurban/smhasher
cd smhasher/
git submodule init
git submodule update
cmake .
make
Adding a new hash function
To add a new function, you can edit just three files: Hashes.cpp, Hashes.h and main.cpp.
For example, I will add the ElfHash:
unsigned long ElfHash(const unsigned char *s)
{
unsigned long h = 0, high;
while (*s)
{
h = (h << 4) + *s++;
if (high = h & 0xF0000000)
h ^= high >> 24;
h &= ~high;
}
return h;
}
First, need to modify it slightly to take a seed and length:
uint32_t ElfHash(const void *key, int len, uint32_t seed)
{
unsigned long h = seed, high;
const uint8_t *data = (const uint8_t *)key;
for (int i = 0; i < len; i++)
{
h = (h << 4) + *data++;
if (high = h & 0xF0000000)
h ^= high >> 24;
h &= ~high;
}
return h;
}
Add this function definition to Hashes.cpp. Also add the following to Hashes.h:
uint32_t ElfHash(const void *key, int len, uint32_t seed);
inline void ElfHash_test(const void *key, int len, uint32_t seed, void *out) {
*(uint32_t *) out = ElfHash(key, len, seed);
}
In file main.cpp add the following line into array g_hashes:
{ ElfHash_test, 32, 0x0, "ElfHash", "ElfHash 32-bit", POOR, {0x0} },
(The third value is self-verification. You will learn this only after running the test once.)
Finally, rebuild and run the test:
make
./SMHasher ElfHash
It will show you all the tests that this hash function fails. (It is very bad.)

Basic group arithmetic in libsodium

I am trying to implement a simple cryptographic primitive.
Under the following code: given sa, sk, hn, I want to compute sb: such that sg*G = (sb + sk . hn)*G.
However, after finding sb, the following equality does not hold: sb*G + (sk.hn)G = saG.
My understand stand is that in the exponent is arithmetic modulo the order of group instead of L.
However, I have a few questions relating to their implementation:
why the scalar has to be chosen from [0,L] where L is the order of the subgroup?
is there a "helper" function that multiplies two large scalar without performing modulo L?
int main(void)
{
if (sodium_init() < 0) {
/* panic! the library couldn't be initialized, it is not safe to use */
return -1;
}
uint8_t sb[crypto_core_ed25519_SCALARBYTES];
uint8_t sa[crypto_core_ed25519_SCALARBYTES];
uint8_t hn[crypto_core_ed25519_SCALARBYTES];
uint8_t sk[crypto_core_ed25519_SCALARBYTES];
crypto_core_ed25519_scalar_random(sa); // s_a <- [0,l]
crypto_core_ed25519_scalar_random(sk); // sk <- [0,l]
crypto_core_ed25519_scalar_random(hn); // hn <- [0,l]
uint8_t product[crypto_core_ed25519_SCALARBYTES];
crypto_core_ed25519_scalar_mul(product, sk,hn); // sk*hn
crypto_core_ed25519_scalar_sub(sb, sa, product); // sb = sa-hn*sk
uint8_t point1[crypto_core_ed25519_BYTES];
crypto_scalarmult_ed25519_base(point1, sa);
uint8_t point2[crypto_core_ed25519_BYTES];
uint8_t sum[crypto_core_ed25519_BYTES];
// equal
// crypto_core_ed25519_scalar_add(sum, sb, product);
// crypto_scalarmult_ed25519_base(point2, sum);
// is not equal
uint8_t temp1[crypto_core_ed25519_BYTES];
uint8_t temp2[crypto_core_ed25519_BYTES];
crypto_scalarmult_ed25519_base(temp1, sb); // sb*G
crypto_scalarmult_ed25519_base(temp2, product); //
crypto_core_ed25519_add(point2, temp1, temp2);
if(memcmp(point1, point2, 32) != 0)
{
printf("[-] Not equal ");
return -1;
}
printf("[+] equal");
return 0;
}
I got the answer from jedisct1 , the author of libsodium and I will post it here:
crypto_scalarmult_ed25519_base() clamps the scalar (clears the 3 lower bits, set the high bit) before performing the multiplication.
Use crypto_scalarmult_ed25519_base_noclamp() to prevent this.
Or, even better, use the Ristretto group instead.

Determine types from a variadic function's arguments in C

I'd like a step by step explanation on how to parse the arguments of a variadic function
so that when calling va_arg(ap, TYPE); I pass the correct data TYPE of the argument being passed.
Currently I'm trying to code printf.
I am only looking for an explanation preferably with simple examples but not the solution to printf since I want to solve it myself.
Here are three examples which look like what I am looking for:
https://stackoverflow.com/a/1689228/3206885
https://stackoverflow.com/a/5551632/3206885
https://stackoverflow.com/a/1722238/3206885
I know the basics of what typedef, struct, enum and union do but can't figure out some practical application cases like the examples in the links.
What do they really mean? I can't wrap my brain around how they work.
How can I pass the data type from a union to va_arg like in the links examples? How does it match?
with a modifier like %d, %i ... or the data type of a parameter?
Here's what I've got so far:
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include "my.h"
typedef struct s_flist
{
char c;
(*f)();
} t_flist;
int my_printf(char *format, ...)
{
va_list ap;
int i;
int j;
int result;
int arg_count;
char *cur_arg = format;
char *types;
t_flist flist[] =
{
{ 's', &my_putstr },
{ 'i', &my_put_nbr },
{ 'd', &my_put_nbr }
};
i = 0;
result = 0;
types = (char*)malloc( sizeof(*format) * (my_strlen(format) / 2 + 1) );
fparser(types, format);
arg_count = my_strlen(types);
while (format[i])
{
if (format[i] == '%' && format[i + 1])
{
i++;
if (format[i] == '%')
result += my_putchar(format[i]);
else
{
j = 0;
va_start(ap, format);
while (flist[j].c)
{
if (format[i] == flist[j].c)
result += flist[i].f(va_arg(ap, flist[i].DATA_TYPE??));
j++;
}
}
}
result += my_putchar(format[i]);
i++;
}
va_end(ap);
return (result);
}
char *fparser(char *types, char *str)
{
int i;
int j;
i = 0;
j = 0;
while (str[i])
{
if (str[i] == '%' && str[i + 1] &&
str[i + 1] != '%' && str[i + 1] != ' ')
{
i++;
types[j] = str[i];
j++;
}
i++;
}
types[j] = '\0';
return (types);
}
You can't get actual type information from va_list. You can get what you're looking for from format. What it seems you're not expecting is: none of the arguments know what the actual types are, but format represents the caller's idea of what the types should be. (Perhaps a further hint: what would the actual printf do if a caller gave it format specifiers that didn't match the varargs passed in? Would it notice?)
Your code would have to parse the format string for "%" format specifiers, and use those specifiers to branch into reading the va_list with specific hardcoded types. For example, (pseudocode) if (fspec was "%s") { char* str = va_arg(ap, char*); print out str; }. Not giving more detail because you explicitly said you didn't want a complete solution.
You will never have a type as a piece of runtime data that you can pass to va_arg as a value. The second argument to va_arg must be a literal, hardcoded specification referring to a known type at compile time. (Note that va_arg is a macro that gets expanded at compile time, not a function that gets executed at runtime - you couldn't have a function taking a type as an argument.)
A couple of your links suggest keeping track of types via an enum, but this is only for the benefit of your own code being able to branch based on that information; it is still not something that can be passed to va_arg. You have to have separate pieces of code saying literally va_arg(ap, int) and va_arg(ap, char*) so there's no way to avoid a switch or a chain of ifs.
The solution you want to make, using the unions and structs, would start from something like this:
typedef union {
int i;
char *s;
} PRINTABLE_THING;
int print_integer(PRINTABLE_THING pt) {
// format and print pt.i
}
int print_string(PRINTABLE_THING pt) {
// format and print pt.s
}
The two specialized functions would work fine on their own by taking explicit int or char* params; the reason we make the union is to enable the functions to formally take the same type of parameter, so that they have the same signature, so that we can define a single type that means pointer to that kind of function:
typedef int (*print_printable_thing)(PRINTABLE_THING);
Now your code can have an array of function pointers of type print_printable_thing, or an array of structs that have print_printable_thing as one of the structs' fields:
typedef struct {
char format_char;
print_printable_thing printing_function;
} FORMAT_CHAR_AND_PRINTING_FUNCTION_PAIRING;
FORMAT_CHAR_AND_PRINTING_FUNCTION_PAIRING formatters[] = {
{ 'd', print_integer },
{ 's', print_string }
};
int formatter_count = sizeof(formatters) / sizeof(FORMAT_CHAR_AND_PRINTING_FUNCTION_PAIRING);
(Yes, the names are all intentionally super verbose. You'd probably want shorter ones in the real program, or even anonymous types where appropriate.)
Now you can use that array to select the correct formatter at runtime:
for (int i = 0; i < formatter_count; i++)
if (current_format_char == formatters[i].format_char)
result += formatters[i].printing_function(current_printable_thing);
But the process of getting the correct thing into current_printable_thing is still going to involve branching to get to a va_arg(ap, ...) with the correct hardcoded type. Once you've written it, you may find yourself deciding that you didn't actually need the union nor the array of structs.

Static Parameter Function Specialization in D

I've read somewhere that D supports specialization of functions to calls where arguments are compile-time constants. Typical use of this is in matrix power functions (if exponent is 2 x*x is often faster than the general case).
I want this in my member function
bool opIndexAssign(bool b, size_t i) #trusted pure nothrow in {
assert(i < len); // TODO: Add static assert(i < len) when i is constant
} body {
b ? bts(ptr, i) : btr(ptr, i);
return b;
}
of a statically sized BitSet struct I'm writing. This in order to, when possible, get compile-time bounds checking on the index variable i. I thought
bool opIndexAssign(bool b, const size_t i) #trusted pure nothrow in {
static assert(i < len);
} body {
b ? bts(ptr, i) : btr(ptr, i);
return b;
}
would suffice but then DMD complains as follows
dmd -debug -gc -gs -unittest -D -Dd/home/per/.emacs.d/auto-builds/dmd/Debug-Boundscheck-Unittest/home/per/Work/justd/ -w -main ~/Work/justd/bitset.d /home/per/Work/justd/assert_ex.d -of/home/per/.emacs.d/auto-builds/dmd/Debug-Boundscheck-Unittest/home/per/Work/justd/bitset
/home/per/Work/justd/bitset.d(58): Error: bitset.BitSet!2.BitSet.opIndexAssign called with argument types (bool, int) matches both:
/home/per/Work/justd/bitset.d(49): opIndexAssign(bool b, ulong i)
and:
/home/per/Work/justd/bitset.d(65): opIndexAssign(bool b, const(ulong) i)
/home/per/Work/justd/bitset.d(66): Error: variable i cannot be read at compile time
/home/per/Work/justd/bitset.d(66): while evaluating: static assert(i < 2LU)
/home/per/Work/justd/bitset.d(58): Error: bitset.BitSet!2.BitSet.opIndexAssign called with argument types (bool, int) matches both:
/home/per/Work/justd/bitset.d(49): opIndexAssign(bool b, ulong i)
Do I have to make parameter i a template parameter, say using type U, and then use static if someTypeTrait!U. I tried this but isMutable!Index always evaluates to true.
import std.traits: isIntegral;
bool opIndexAssign(Index)(bool b, Index i) #trusted pure nothrow if (isIntegral!Index) in {
import std.traits: isMutable;
// See also: http://stackoverflow.com/questions/19906516/static-parameter-function-specialization-in-d
static if (isMutable!Index) {
assert(i < len);
} else {
import std.conv: to;
static assert(i < len,
"Index " ~ to!string(i) ~ " must be smaller than BitSet length " ~ to!string(len));
}
} body {
b ? bts(ptr, i) : btr(ptr, i);
return b;
}
What you're trying to do doesn't really work. You can do a template value parameter:
void foo(int i)() { /* use i at compile time */ }
but then you can't pass a runtime value to it, and it has different call syntax: foo!2 vs foo(2).
The closest you can get is is CTFE:
int foo(int i) { return i; }
enum something = foo(2); // works, evaluated at compile time
int s = foo(2); // also works, but runs at runtime.
Inside the function, there is a magic if(__ctfe) { running at compile time } else { at runtime}, but again, this isn't if there's a literal, it is if the function is run in a CT context, e.g., assigning the result to an enum constant.
But, otherwise, an int literal is still a mutable int as far as the function is concerned. So what you're specifically trying to do won't work in D as it is right now. (There's been some talk about wanting a way to tell if it is a literal, but as far as I know, there's no plan to actually do it.)

How to interpret objective-c type specifier (e.g. returned by method_copyReturnType())?

Given I have a type specifier as returned by method_copyReturnType(). In the GNU runtime delivered with the GCC there are various methods to work with such a type specifier like objc_sizeof_type(), objc_alignof_type() and others.
When using the Apple runtime there are no such methods.
How can I interpret a type specifier string (e.g. get the size of a type) using the Apple runtime without implementing an if/else or case switch for myself?
[update]
I am not able to use the Apple Foundation.
I believe that you're looking for NSGetSizeAndAlignment:
Obtains the actual size and the aligned size of an encoded type.
const char * NSGetSizeAndAlignment (
const char *typePtr,
NSUInteger *sizep,
NSUInteger *alignp
);
Discussion
Obtains the actual size and the aligned size of the first data type represented by typePtr and returns a pointer to the position of the next data type in typePtr.
This is a Foundation function, not part of the base runtime, which is probably why you didn't find it.
UPDATE: Although you didn't initially mention that you're using Cocotron, it is also available there. You can find it in Cocotron's Foundation, in NSObjCRuntime.m.
Obviously, this is much better than rolling your own, since you can trust it to always correctly handle strings generated by its own runtime in the unlikely event that the encoding characters should change.
For some reason, however, it's unable to handle the digit elements of a method signature string (which presumably have something to do with offsets in memory). This improved version, by Mike Ash will do so:
static const char *SizeAndAlignment(const char *str, NSUInteger *sizep, NSUInteger *alignp, int *len)
{
const char *out = NSGetSizeAndAlignment(str, sizep, alignp);
if(len)
*len = out - str;
while(isdigit(*out))
out++;
return out;
}
afaik, you'll need to bake that info into your binary. just create a function which returns the sizeof and alignof in a struct, supports the types you must support, then call that function (or class method) for the info.
The program below shows you that many of the primitives are just one character. So the bulk of the function's implementation could be a switch.
static void test(SEL sel) {
Method method = class_getInstanceMethod([NSString class], sel);
const char* const type = method_copyReturnType(method);
printf("%s : %s\n", NSStringFromSelector(sel).UTF8String, type);
free((void*)type);
}
int main(int argc, char *argv[]) {
#autoreleasepool {
test(#selector(init));
test(#selector(superclass));
test(#selector(isEqual:));
test(#selector(length));
return 0;
}
}
and you could then use this as a starting point:
typedef struct t_pair_alignof_sizeof {
size_t align;
size_t size;
} t_pair_alignof_sizeof;
static t_pair_alignof_sizeof MakeAlignOfSizeOf(size_t align, size_t size) {
t_pair_alignof_sizeof ret = {align, size};
return ret;
}
static t_pair_alignof_sizeof test2(SEL sel) {
Method method = class_getInstanceMethod([NSString class], sel);
const char* const type = method_copyReturnType(method);
const size_t length = strlen(type);
if (1U == length) {
switch (type[0]) {
case '#' :
return MakeAlignOfSizeOf(__alignof__(id), sizeof(id));
case '#' :
return MakeAlignOfSizeOf(__alignof__(Class), sizeof(Class));
case 'c' :
return MakeAlignOfSizeOf(__alignof__(signed char), sizeof(signed char));
...