printf() less __assert_fail()

printf() less __assert_fail()

[Daniel Cegiełka]

[-- Attachment #1: Type: text/plain, Size: 81 bytes --]

It makes no sense to use printf() in this function. Is this a good idea?

Daniel

[-- Attachment #2: assert.c --]
[-- Type: text/x-csrc, Size: 635 bytes --]

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <string.h>

static void print(const char *s)
{
	write(2, s, strlen(s));
}

static unsigned int fmt_uint(char *s, int n)
{
	unsigned int l = 1, q = n;

	while (q > 9) { ++l; q /= 10; }
	if (s) {
		s += l;
		do { *--s = '0' + (n % 10); n /= 10; } while (n);
	}
	return l;
}

void __assert_fail(const char *expr, const char *file, int line, const char *func)
{
	char buf[10];

	print("Assertion failed: ");
	print(expr);
	print(" (");
	print(file);
	print(": ");
	print(func);
	print(": ");
	write(2, buf, fmt_uint(buf, line));
	print(")\n");
	fflush(NULL);
	abort();
}

Re: printf() less __assert_fail()

[Rich Felker]

On Wed, Oct 01, 2014 at 04:45:44PM +0200, Daniel Cegiełka wrote:
> It makes no sense to use printf() in this function. Is this a good idea?

I'm not sure why it "makes no sense". There are a few minor
differences in behavior with your version, the main ones I see being
that yours is non-atomic but async-signal-safe. Are there major
reasons you want to change it?

Rich

Re: printf() less __assert_fail()

[Daniel Cegiełka]

2014-10-01 16:54 GMT+02:00 Rich Felker <dalias@libc.org>:
> On Wed, Oct 01, 2014 at 04:45:44PM +0200, Daniel Cegiełka wrote:
>> It makes no sense to use printf() in this function. Is this a good idea?
>
> I'm not sure why it "makes no sense". There are a few minor
> differences in behavior with your version, the main ones I see being
> that yours is non-atomic but async-signal-safe. Are there major
> reasons you want to change it?

Atomic version, yes, I thought about it. It's just an idea. This
version gives a much smaller binary. Similar solutions (WSTR macro)
you use in noxcuse for a small size.

Daniel

> Rich