Is this code (simplified from a real project) correct? Will the message always print?
char *cp = NULL;
char **cpp = &cp;
if(*cpp == NULL) {
printf("I believe this will this always print. Does it?\n");
}
Thanks!
Add a comment
|
3 Answers
There is nothing wrong with the code you have shown. Your char ** pointer is pointing to a valid variable, thus it's always proper to dereference it.
P.S. Yes, it will always print.
5 Comments
Mark Ransom
@Heath Hunnicutt - Is there something wrong with the answer? I was reading between the lines of the question and guessing there was some confusion about how pointers to pointers actually work.
pmg
@Heath: I can't speak for Mark, of course ... but a) of course we want the points (how do you think we got to 40k or 20K or 6K); b) not really, points are overrated; c) I don't care for points: my answer wasn't suitable for a comment; d) if SO shuts down, I'll sell my points to the highest bidder
Mark Ransom
@pmg, thanks for the elaboration. Points aren't the end-all, but they're a good surrogate for measuring how helpful you've been. That's my primary motivation. As time goes on I find myself using comments as much or more than answers, even though comments don't affect your point scores.
Mark Ransom
P.S. I'll just say that some days it's hard to find a question you can answer! Not every song has to be a hit.
pmg
indeed; I also find myself doing more and more comments nowadays. I think @Heath was wondering why the community wiki check for answers is not used more :)
Yes, it will always print.
You can safely assume that your base stack pointer never points to 0x0, so &cp will always not equal NULL.
In fact the compiler will eliminate the check at compile time, because it knows &cp != NULL.
See for yourself:
Compiled with -O1:
$ objdump -dC a.out
a.out: file format elf64-x86-64
...
00000000004004f4 <main>:
4004f4: 48 83 ec 08 sub $0x8,%rsp
4004f8: bf 00 06 40 00 mov $0x400600,%edi
4004fd: e8 ee fe ff ff callq 4003f0 <puts@plt>
400502: b8 00 00 00 00 mov $0x0,%eax
400507: 48 83 c4 08 add $0x8,%rsp
40050b: c3 retq
(With -O0 there will be a test, though:)
00000000004004f4 <main>:
4004f4: 55 push %rbp
4004f5: 48 89 e5 mov %rsp,%rbp
4004f8: 48 83 ec 10 sub $0x10,%rsp
4004fc: 48 c7 45 f0 00 00 00 movq $0x0,-0x10(%rbp)
400503: 00
400504: 48 8d 45 f0 lea -0x10(%rbp),%rax
400508: 48 89 45 f8 mov %rax,-0x8(%rbp)
40050c: 48 8b 45 f8 mov -0x8(%rbp),%rax
400510: 48 8b 00 mov (%rax),%rax
400513: 48 85 c0 test %rax,%rax
400516: 75 0a jne 400522 <main+0x2e>
400518: bf 20 06 40 00 mov $0x400620,%edi
40051d: e8 ce fe ff ff callq 4003f0 <puts@plt>
400522: b8 00 00 00 00 mov $0x0,%eax
400527: c9 leaveq
400528: c3 retq
Comments
Of course that always prints, why would it not?
The pointer you are dereferencing in your if() statement is cpp. You only dereference one level, by using one * asterisk. cpp may point to a NULL pointer, but cpp is not itself NULL.
On the other hand writing char k = **cpp should result in a crash of some sort.
14 Comments
aschepler
Undefined Behavior "should" not necessarily do anything. It might crash, it might "work", it might spew acid.
Heath Hunnicutt
Blah, blah, undefined behavior. Seriously irrelevant IRL. Often used as a pretense for representing C as less predictable than it really is. Nothing will be spewing acid, as an example. I wager the OP is not coding on DOS, either.
Oliver Charlesworth
@Heath: It's most definitely not irrelevant (well, acid probably is, but not the "might crash, might work").
R.. GitHub STOP HELPING ICE
UB is very relevant IRL here. If the compiler can determine statically that
**cpp will always be a null pointer dereference, it can optimize out any code after it and any prior code paths that necessarily lead up to the dereference, on the assumption that they cannot happen.Oliver Charlesworth
@Heath: Not everyone's writing code for an x86 with optimizations disabled! Many embedded controllers/DSPs (for example) have no hardware trap for acessing 0x0000.
|