1
00:00:00,000 --> 00:00:00,180

2
00:00:00,180 --> 00:00:02,920
>> SPEAKER 1: Let's write a program that
gets a string from the user without

3
00:00:02,920 --> 00:00:05,700
using the CS50 Library's
function GetString.

4
00:00:05,700 --> 00:00:08,720
To do this, we'll go ahead and use
scanf, the function that the GetString

5
00:00:08,720 --> 00:00:10,950
function actually uses
underneath the hood.

6
00:00:10,950 --> 00:00:13,780
But I'm going to do this deliberately
in a buggy way.

7
00:00:13,780 --> 00:00:17,230
I'm going to do in a way that I think
would be right, but it turns out that

8
00:00:17,230 --> 00:00:19,380
my assumption's going to
be quite, quite flawed.

9
00:00:19,380 --> 00:00:20,800
And in fact, quite dangerous.

10
00:00:20,800 --> 00:00:24,750
Because bugs like the one I'm about to
make can be exploited by adversaries

11
00:00:24,750 --> 00:00:28,870
such that your machine or your program
can be taken over potentially.

12
00:00:28,870 --> 00:00:30,200
>> Let's begin as follows.

13
00:00:30,200 --> 00:00:33,540
First let's declare our string,
otherwise known now as a char star,

14
00:00:33,540 --> 00:00:34,750
and call it s.

15
00:00:34,750 --> 00:00:39,400
Lets next prompt the user for a string,
as with "string please." And

16
00:00:39,400 --> 00:00:44,250
let's now get the string from the user
using scanf, quote unquote, "%s." In

17
00:00:44,250 --> 00:00:47,760
other words, let's inform scanf that we
do in fact expect to get a string

18
00:00:47,760 --> 00:00:48,630
from the user.

19
00:00:48,630 --> 00:00:50,810
>> But now we need to tell scanf
one other thing--

20
00:00:50,810 --> 00:00:53,350
where to put the string that
the user provides.

21
00:00:53,350 --> 00:00:57,840
Well, I'm going to quite simply start
with comma s, specifying that I'd like

22
00:00:57,840 --> 00:00:59,320
scanf to put the string there.

23
00:00:59,320 --> 00:01:04,818
I'm next going to print out something
like printf "thanks for the %s

24
00:01:04,818 --> 00:01:10,670
backslash n comma." And as always, I'm
going to pass in the string, s.

25
00:01:10,670 --> 00:01:14,920
Now let's save, compile, and run this
program, and see if we can't induce

26
00:01:14,920 --> 00:01:16,590
the problem I predicted.

27
00:01:16,590 --> 00:01:18,650
>> Make scanf-1.

28
00:01:18,650 --> 00:01:20,960
./scanf-1.

29
00:01:20,960 --> 00:01:21,830
String please.

30
00:01:21,830 --> 00:01:25,540
Let's provide something like, "hello."
"Thanks for the null." Hmm, that's not

31
00:01:25,540 --> 00:01:26,750
what I was expecting.

32
00:01:26,750 --> 00:01:28,240
So what's going on here?

33
00:01:28,240 --> 00:01:32,040
>> Well, it turns out because we declared
s as a char star but we didn't

34
00:01:32,040 --> 00:01:36,120
actually stored in s the address of an
actual chunk of memory, scanf didn't

35
00:01:36,120 --> 00:01:38,940
have anywhere to put the string
that the user typed in.

36
00:01:38,940 --> 00:01:42,510
Indeed, if the user were to now type in
a much longer string than "hello,"

37
00:01:42,510 --> 00:01:46,780
for instance several lines of text or
several paragraphs of text, it's quite

38
00:01:46,780 --> 00:01:50,280
possible that we might induce a
so-called segmentation fault.

39
00:01:50,280 --> 00:01:53,570
>> Because scanf isn't going to know that
I haven't actually put an address

40
00:01:53,570 --> 00:01:54,610
inside of s.

41
00:01:54,610 --> 00:01:58,000
Rather, it's going to see some value
in s, some pattern of bits that may

42
00:01:58,000 --> 00:02:00,910
very well be a garbage value,
there just by chance.

43
00:02:00,910 --> 00:02:04,600
And scanf is still going to try to write
the user string to that address,

44
00:02:04,600 --> 00:02:08,789
even if it is a garbage value, which
could indeed induce a crash.

45
00:02:08,789 --> 00:02:10,130
>> So how are we going to fix this?

46
00:02:10,130 --> 00:02:12,523