Thread: adjusting character counts for utf8

Originally Posted by tabstop

I suppose you could always use the -E flag to make sure strcasestr is actually there (as this is the warning I would expect if the prototype was missing). (Are you using -std or -ansi flags?)

Oh it's there. This works:

Code:

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

int main() {
	char *ptr=strstr("this and that","and");
	printf("%p %s\n",ptr,ptr);
	ptr=strstr("this and that","and");
	printf("%p %s\n",ptr,ptr);
	return 0;
}

But obviously it's useless for illustrating my point. I have the same problem on my other computer, which is running a slightly older glibc/gcc (FC7-32 vs. FC10-64).

I'm not using the flags you mention, but I see no good reason for strcasestr to cause a segfault like that.

What do you think about what I was trying to say about the pointer addressess? How could two pointers with two values which appear different to me effectively point to the same place (like I said, the first find would be fine, but then the subsequent search would go awry because of the address returned from the first find -- meaning that pointer was no good for arithmetic)??

Originally Posted by tabstop

Maybe I got lost with what you were asking on the pointer question -- just because two pointers point to the same content doesn't mean they point to the same place.

!!
huh...I had presumed that there is one copy of the buffer in memory, and that any pointer into that content must point into somewhere in that one copy.

I can't find anything weird about strcasestr; I get normal things, but I'm not using utf8 characters.

So the incredibly simple example above, ptr=strcasestr(haystack,needle), does not segfault for you? There are not utf8 characters in that.

I'd be curious to know, since I'm writing for other linux systems, and I would say FC7 and FC10 are very common. If it works on other systems, then our strcasestr is a dud.

>> test.c:8: warning: assignment makes pointer from integer without a cast
That's bad news. Compile with '-Wimplicit'. I think tabstop was right-on about the prototype being missing. Some online man pages say _GNU_SOURCE is required.

gg

Well, I tried _GNU_SOURCE, -Wimplicit, even include <ctype.h> etc. No dice. Then I just added my own prototype at the top:

Code:

char *strcasestr (const char *haystack, const char *needle);

and it works. I'm kind of surprised; this has obviously been the case for several years.

Anyway, I won't be using strcasestr because, as tabstop confirms, it may return a pointer into a copy of (part of) haystack rather than into haystack itself.

>> it may return a pointer into a copy
I highly doubt it. All documentation suggests otherwise. Not to mention, there's no reason what-so-ever to make a copy.

gg

Originally Posted by Codeplug

>> it may return a pointer into a copy
I highly doubt it. All documentation suggests otherwise. Not to mention, there's no reason what-so-ever to make a copy.
gg

So would I. But then, I would include a prototype if I was writing a function for the C library! So who knows?

Plus there is this, tho it may be just hearsay:

Originally Posted by me

What do you think about what I was trying to say about the pointer addressess? How could two pointers with two values which appear different to me effectively point to the same place (like I said, the first find would be fine, but then the subsequent search would go awry because of the address returned from the first find -- meaning that pointer was no good for arithmetic)??

In fact, I had to put a control in so that if the return pointer was beyond the end of the search, the search would end (otherwise there was another seg fault for submitting this mystery pointer back to strcasestr). There's a toggle on the GUI to determine the case sensitivity, and lots of debugging to the console, so I could sit and just toggle the button on and off and hit return to do the exact same search on the exact same buffer, over and over. In both cases, the result in the GUI was the same -- str(case)str found the first term. But the pointer value in the debug output was, as I said, outside the buffer for strcasestr. Because I was just one substituting one function for another, there was nothing in my code which could account for the discrepancy.
Perhaps intentionally leaving the prototype out is a coy way of indicating something...

An implicitly declared function is treated as returning int. That would be bad news on a 64bit system.

gg

Originally Posted by tabstop

But ... but ... when I specifically asked whether a prototype was included in the preprocessor output (what you see with -E) you said "Oh it's there." So which is it?

What you actually wrote, being a smart, but perhaps a little inscrutable or cryptic person, who occasionally seems too lazy to completely explain him or herself to ignoramuses like me, is:

I suppose you could always use the -E flag to make sure strcasestr is actually there (as this is the warning I would expect if the prototype was missing).

Using the -E flag just made me roll my eyes. If it knew what it was for, I suppose that might have meant the same thing as what you just wrote. Sorry! You know there is just too much information, and people like myself sometimes have to wear blinders so we don't get distracted by too many new things.

Anyway, I was of the belief that if a function call actually works, then the function must be somewhere. That someone could include a function in a library somewhere without a prototype (as I guess is the case) wouldn't have occurred to me, especially since I don't think gcc would let me do that.

But thanks for your help, tabstop. I can live with "somewhat cryptic".

Originally Posted by Codeplug

An implicitly declared function is treated as returning int. That wouldn't be bad news on a 64bit system.
gg

Could this explain the pointer wierdness? I would think the value would have to be the same anyway...although I notice on my 64 bit system, an int is actually twice as big as a pointer. Hmmm. Only with a very large buffer. Hmmm.

If I can find an older version of the code (way back when I was still using strcasestr) I will try it on the 32.

Originally Posted by MK27

What you actually wrote, being a smart, but perhaps a little inscrutable or cryptic person, who occasionally seems too lazy to completely explain him or herself...

You almost got it 100% right.

But that's what makes tabstop 85%-90% awesome.