Path: ...!news.nobody.at!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Ike Naar <ike@sdf.org>
Newsgroups: comp.lang.c
Subject: Re: why does bsearch segfault on custom strcmp when qsort works fine?
Date: Thu, 15 Aug 2024 08:55:45 -0000 (UTC)
Organization: A noiseless patient Spider
Lines: 57
Message-ID: <slrnvbrgkg.qdu.ike@iceland.freeshell.org>
References: <UQWdnfwQxtYABSD7nZ2dnZfqnPadnZ2d@brightview.co.uk>
Injection-Date: Thu, 15 Aug 2024 10:55:45 +0200 (CEST)
Injection-Info: dont-email.me; posting-host="4fc88af0757a85c3325ff7527967d582";
	logging-data="962690"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX18/dl3DAq6vi0IOznQZ5mSA"
User-Agent: slrn/1.0.3 (Patched for libcanlock3) (NetBSD)
Cancel-Lock: sha1:W0uYgktzzwy38ynt542VoP0oR2s=
Bytes: 3406

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;