NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part1/3036

Text File | 1996-08-05 | 3.7 KB | 117 lines

Newsgroups: comp.lang.c
Path: in2.uu.net!tellab5!news
From: Joe Toth <[e-mail]>
Subject: Re: Segmentation fault at return statement. Can't figure out ....
X-Nntp-Posting-Host: sunh25
Content-Type: text/plain; charset=us-ascii
Message-ID: <[e-mail]>
Sender: [e-mail] (News)
Content-Transfer-Encoding: 7bit
Organization: Tellabs Operations, Inc.
References: <4dm2n1$[e-mail]> <4dop08$[e-mail]> <[e-mail]> <4e63p2$[e-mail]> <[e-mail]>
Mime-Version: 1.0
Date: Thu, 25 Jan 1996 14:09:06 GMT
X-Mailer: Mozilla 1.1 (X11; U; SunOS 4.1.3 sun4c)
X-Url: news:[e-mail]

Dan Pop <[e-mail]> wrote:
>[e-mail] (Ralph Silverman) writes:
>
>>Dan Pop ([e-mail]) wrote:
>>: [e-mail] (Ralph Silverman) writes:
>>
>>: >    'test'
>>: >    is likely to be
>>: >    a command of
>>: >    the
>>: >        operating system
>>
>>: So what?  There's no possible conflict/interaction between a function
>>: name in a C program and the name of an OS command.
>>
>>: >    1)    look it up.
>>: >    2)    rename yours.
>>:                 ^^^^^^^^^^^^^
>>: Why???
>>
>>        is it not possible
>>        that the link order of
>>        test.c etc.
>>    resulted in an executable named
>>    test.exe,
>>    test.com
>>    or such?
>
>On some systems, this may be possible, however, on the only system where
>"test" is a command (that I know of), this is not possible.
>
>No matter the name of the resulting executable, this cannot explain the
>behaviour observed by the original poster.
>
>Dan
>--
>Dan Pop
>CERN, CN Division
>Email: [e-mail]
>Mail:  CERN - PPE, Bat. 31 R-004, CH-1211 Geneve 23, Switzerland

Come on guys....

Lets give an answer that helps.


On some systems, 'test' is a builtin command of the appropriate
shell (csh, ksh, etc.).  On other systems, it is an executable.

In either case (which together cover all unix & DOS systems), the
compile/link process doesn't ever look at system/shell commands or
executables.

The compiler/linker resolve function calls on 'object' files.

If you want to invoke the 'test' executable (if there is one), you
use a 'C' call something like;

        system ( "test <test parameters>" );

This is the only way to execute an executable.

A 'C' function reference to 'test' requires a definition of
a function named 'test()'.  If it is not found in the local source,
the linker will look for it in the 'object' files (possibly in an
archive file) specified in the 'link' command.

So, as stated 'There's no possible conflict/interaction<...>'.

Back to the real problem....

The user did not show any of the code that was contained in 'test'
other than the 'return(999)'.  My bet is that the error is in the
missing code......

Possibly;
   writing outside the bounds of a locally defined array.
   misuse of a pointer.
   any of a number of coding errors.

Something in the code probably wrote over the return pointer.
IT DOES HAPPEN...................
It all depends on how & where the return pointers are stored, as
many systems don't protect the return pointer if it is in the
local variable heap (or does yours use stacks).
I've accidentally done it myself.

Without seeing the code, it is impossible to say exactly what
caused the problem.....

--
  _  _  ___  --------------------------+---------------------------------
  | / _  |    Joseph G. Toth Jr.       | Tellabs Operations, Inc.
\_| \_/  |                             | [e-mail]

> Every program has at least one bug and can be shortened by at least
> one instruction -- from which, by induction, it is evident that every
> program can be reduced to one instruction that does not work.
>                       -- Ken Arnold