tests/example segfaults on add user

[wittend]

On Ubuntu 18.04 example segfaults on option 1 or 2

$ git clone
$ cd uthash/tests

$ make
all tests pass

$ make clean
$ make example
cc -I../src -g -Wall -o example example.c
(warnings concerning gets, add_user, atoi...)

$.../uthash/tests$ ./example

1. add user
2. add/rename user by id
3. find user
4. delete user
5. delete all users
6. sort items by name
7. sort items by id
8. print users
9. count users
10. quit
    1
    name?
    Gertrude
    Segmentation fault (core dumped)

Option 2 behaves similarly.

[Quuxplusone]

Hm. Can't reproduce on RHEL5 or Mac OSX. It would be useful to see (1) a transcript of exactly what you entered (e.g. if it wasn't Gertrude but maybe Gertrude-followed-by-several-space-characters?), and/or (2) a stack trace. Obviously example.c isn't "safe," in the sense that it does a gets into a 10-byte buffer, so if you overflow the buffer all bets are off.

Oh, and if changing the buffer size from [10] to [100] fixes your segfault, that's useful to know too.

[wittend]

Quuxplusone, I can post the complete output tomorrow. I suspect that is something simple. But I thought that it would be good to know that it doesn't behave 'right out of the box' on a major platform. And no, it isn't specific to 'Gertrude'. The first few times I tried 'Joe' 'bob' and 'Dave'. I thought that might be too short, therefore 'Gertrude' Dave

…

On Tue, Mar 9, 2021 at 8:31 PM Quuxplusone ***@***.***> wrote: Hm. Can't reproduce on RHEL5 or Mac OSX. It would be useful to see (1) a transcript of exactly what you entered (e.g. if it wasn't Gertrude but maybe Gertrude-followed-by-several-space-characters?), and/or (2) a stack trace. Obviously example.c isn't "safe," in the sense that it does a gets into a 10-byte buffer, so if you overflow the buffer all bets are off. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#228 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACSMISAFNLXFW2RLAEBBSDTC3DWDANCNFSM4Y4Z2EAA> .

[wittend]

Quuxplusone, OK, this is all really about the use of gets() in example.c. Nothing about uthash itself. I had forgotten gets() even exists. Just as well: https://stackoverflow.com/questions/1694036/why-is-the-gets-function-so-dangerous-that-it-should-not-be-used Dave

On Tue, Mar 9, 2021 at 10:08 PM David Witten ***@***.***> wrote: Quuxplusone, I can post the complete output tomorrow. I suspect that is something simple. But I thought that it would be good to know that it doesn't behave 'right out of the box' on a major platform. And no, it isn't specific to 'Gertrude'. The first few times I tried 'Joe' 'bob' and 'Dave'. I thought that might be too short, therefore 'Gertrude' Dave On Tue, Mar 9, 2021 at 8:31 PM Quuxplusone ***@***.***> wrote: > Hm. Can't reproduce on RHEL5 or Mac OSX. It would be useful to see (1) a > transcript of exactly what you entered (e.g. if it wasn't Gertrude but > maybe Gertrude-followed-by-several-space-characters?), and/or (2) a > stack trace. Obviously example.c isn't "safe," in the sense that it does > a gets into a 10-byte buffer, so if you overflow the buffer all bets are > off. > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#228 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AACSMISAFNLXFW2RLAEBBSDTC3DWDANCNFSM4Y4Z2EAA> > . >

$ make example cc -I../src -g -Wall -o example example.c example.c: In function ‘main’: example.c:98:9: warning: implicit declaration of function ‘gets’; did you mean ‘fgets’? [-Wimplicit-function-declaration] gets(in); ^~~~ fgets example.c:102:32: warning: passing argument 2 of ‘add_user’ makes pointer from integer without a cast [-Wint-conversion] add_user(id++, gets(in)); ^~~~ example.c:14:6: note: expected ‘char *’ but argument is of type ‘int’ void add_user(int user_id, char *name) ^~~~~~~~ example.c:109:30: warning: passing argument 2 of ‘add_user’ makes pointer from integer without a cast [-Wint-conversion] add_user(id, gets(in)); ^~~~ example.c:14:6: note: expected ‘char *’ but argument is of type ‘int’ void add_user(int user_id, char *name) ^~~~~~~~ example.c:113:36: warning: passing argument 1 of ‘atoi’ makes pointer from integer without a cast [-Wint-conversion] s = find_user(atoi(gets(in))); ^~~~ In file included from example.c:2:0: /usr/include/stdlib.h:104:12: note: expected ‘const char *’ but argument is of type ‘int’ extern int atoi (const char *__nptr) ^~~~ example.c:118:36: warning: passing argument 1 of ‘atoi’ makes pointer from integer without a cast [-Wint-conversion] s = find_user(atoi(gets(in))); ^~~~ In file included from example.c:2:0: /usr/include/stdlib.h:104:12: note: expected ‘const char *’ but argument is of type ‘int’ extern int atoi (const char *__nptr) ^~~~ /tmp/cc6zxO7k.o: In function `main': /mnt/2de4d611-1c1b-4c79-9d67-9d8283763f42/Projects/git/uthash/tests/example.c:98: warning: the `gets' function is dangerous and should not be used.

[pdebuyl]

Hi @Quuxplusone are you interested in a pull request for this?

gets() is not recommended, so it could be useful to fix this.

I am using uthash since a few weeks and find it very useful, so I would be willing to propose something according to your criteria.

[Quuxplusone]

@pdebuyl: Sure, if you want to make a PR, I'll look at it. No promises, though. I wouldn't like to complicate the example code too much; I agree it's really bad that it currently relies on gets, but I've also already looked at the naïve "rewrite to use fgets" version and it looked a lot uglier.

It might even be better to throw out the example code and write a completely different example program without gets or fgets, as long as it was something interactive that exercised all the same API entry points.

[Quuxplusone]

Fixed by a109c6b !