Re: why does bsearch segfault on custom strcmp when qsort works fine?

Liste des GroupesRevenir à l c 
Sujet : Re: why does bsearch segfault on custom strcmp when qsort works fine?
De : ike (at) *nospam* sdf.org (Ike Naar)
Groupes : comp.lang.c
Date : 15. Aug 2024, 10:55:45
Autres entêtes
Organisation : A noiseless patient Spider
Message-ID : <slrnvbrgkg.qdu.ike@iceland.freeshell.org>
References : 1
User-Agent : slrn/1.0.3 (Patched for libcanlock3) (NetBSD)
On 2024-08-15, Mark Summerfield <mark@qtrac.eu> wrote:
I have a program (complete source at the end) which correctly outputs this:
>
["charlie" "foxtrot" "oscar" "echo" "alpha" "golf" "tango" "delta" "bravo"]
["alpha" "bravo" "charlie" "delta" "echo" "foxtrot" "golf" "oscar" "tango"]
check_array OK
check_index_found true OK
check_index_found false OK
>
However, if you uncomment the "//#define BUG" line, the output (in gdb) is this:
>
(gdb) run
Starting program: /home/mark/tmp/mycmpstr/mycmpstr
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
["charlie" "foxtrot" "oscar" "echo" "alpha" "golf" "tango" "delta" "bravo"]
["alpha" "bravo" "charlie" "delta" "echo" "foxtrot" "golf" "oscar" "tango"]
check_array OK
>
Program received signal SIGSEGV, Segmentation fault.
__strcmp_avx2 () at ../sysdeps/x86_64/multiarch/strcmp-avx2.S:283
283 ../sysdeps/x86_64/multiarch/strcmp-avx2.S: No such file or directory.
(gdb) bt
#0  __strcmp_avx2 () at ../sysdeps/x86_64/multiarch/strcmp-avx2.S:283
#1  0x00005555555553e0 in mystrcmp (s=0x555555556030, t=0x7fffffffde10) at mycmpstr.c:50
#2  0x00007ffff7e0a53c in __GI_bsearch (__key=0x555555556030, __base=0x7fffffffddf0,
    __nmemb=<optimized out>, __size=8, __compar=0x5555555553b7 <mystrcmp>)
    at ../bits/stdlib-bsearch.h:33
#3  0x0000555555555317 in main () at mycmpstr.c:30
>
The difference is that without BUG defined I use my own binary search,
but with BUG defined I use bsearch.

You're mixing up char* and char** in a few places.

[...]

    // mystrcmp segfaults:
    char* p = bsearch("oscar", words, size, sizeof(char*), mystrcmp);

The elements of the words array have type pointer-to-char.
So the first argument to bsearch should be the address of such an element, that is,
a pointer-to-pointer-to-char and it should contain the adress of a pointer to the
first character of the oscar string.
Also, the value returned from bsearch should be interpreted as a pointer-to-pointer-to-char.

    char * key = "oscar";
    char * * p = bsearch(&key, words, size, sizeof(char*), mystrcmp);

    index = p - words[0];
    found = p != NULL:

Two problems here: first, if bsearch returns NULL, the subtraction is ill-defined.
Second, if bsearch returns non-null the index will be p - words, not p - words[0];

    found = p != NULL:
    if (found) index = p - words;

Date Sujet#  Auteur
15 Aug 24 * why does bsearch segfault on custom strcmp when qsort works fine?5Mark Summerfield
15 Aug 24 +* Re: why does bsearch segfault on custom strcmp when qsort works fine?2Ike Naar
15 Aug 24 i`- Re: why does bsearch segfault on custom strcmp when qsort works fine?1Mark Summerfield
15 Aug 24 `* Re: why does bsearch segfault on custom strcmp when qsort works fine?2Richard Harnden
15 Aug 24  `- Re: why does bsearch segfault on custom strcmp when qsort works fine?1Michael S

Haut de la page

Les messages affichés proviennent d'usenet.

NewsPortal