Subject: Re: possible bug in YDL?
From: A Scottedward Hodel (a.s.hodel@Eng.Auburn.EDU)
Date: Mon Jun 19 2000 - 21:15:56 MDT
I'll look more into the standard to see what's supposed to happen. The
simulation package from which I drew the example has been used for a very
long time on many different platforms (alpha, Sun, Cray, SGI, Linux,
cywgin, etc.) using both Gnu and vendor supplied compilers, where applicable.
That doesn't mean that it correctly handles the va_list, but it gave me
enough concern to check with the mailing list.
Thanks for the feedback.
> From: Jim Cole <greyleaf@yggdrasill.net>
>A Scottedward Hodel's bits of Sun, 18 Jun 2000 translated to:
>
>>At first glance this looks like a gcc/gnu tools question, but the failure
>>only occurs under YDL, so I thought I'd send it in here:
>>
>>The code fragment below works fine using gcc-29.5.2 on a Compaq Alpha, under
>>cygwin on windows 95, and under machten Unix. (The last two using
>>earlier versions of gcc.)
>>
>>The problem appears to be that a va_list cannot be used twice in
>>the same subroutine. If someone else has the time to check this
>>behavior, I'd appreciate it. (Better yet, if there's an easy patch
>>to the va_start/va_end, va_args macros to make this code work properly,
>>that would be great!)
>
>I see the same thing as you when running your code with slightly newer
>compiler (2.95.2-1k) and binutils (2.9.5.0.34-0a) packages. The problem
>goes away if I run it under Linux with gcc 2.95.2 on an Intel machine.
>
>But is your code really correct? The little experience I have had with
>vararg stuff has always involved using va_arg to extract the arguments.
>The man pages say something about the state of va_list being undefined
>after a call to v*printf. But I am not sure if that only applies in the
>context of needing to call va_end or whether it applies to any call
>using a va_list. If you declare and start two va_list's, and hand one
>to each v*printf, the problem goes away. Same goes for extracting with
>va_arg and passing the arguments to printf/fprintf.
>
>Jim
>
>>
>>Here's the code.
>>
>> /* Problem in YDL that does not occur with SunOS, Dec Unix, or
>> * gnu tools under Machen Unix.
>> *
>> * Get a seg fault when executing line marked "MARKER B"
>> * Program received signal SIGSEGV, Segmentation fault.
>> * 0xff49454 in strlen () at soinit.c:59
>> * (gdb) up
>> * #1 0xff2d860 in _IO_vfprintf (s=0x10010a58,
>> * format=0x1b6 <Address 0x1b6 out of bounds>, ap=0x7ffffc30)
>> * at vfprintf.c:1259
>> * 1259 vfprintf.c: No such file or directory.
>> * (gdb) up
>> * #2 0x10000680 in infmsg (format=0x1000082c "starting simulation %s")
>> * at jnk.c:23
>> * 23 vfprintf (log_file, format, args);
>> * (gdb) print format
>> * $2 = 0x1000082c "starting simulation %s"
>> *
>> * If comment out line marked MARKER A, the seg fault does not occur
>> * YDL.
>> */
>>#include <stdio.h>
>>#include <stdarg.h>
>>
>>FILE *log_file;
>>char fname[] = "File name";
>>int lnum = 6;
>>
>>void
>>infmsg (char *format, ...)
>>{
>> va_list args; /* Arguments. */
>>
>> /* Write to standard out and to log file. */
>> va_start (args, format);
>> printf ("INF: ");
>> fprintf (log_file, "INF: ");
>> if (lnum != 0)
>> {
>> printf ("%s (%d) ", fname, lnum);
>> fprintf (log_file, "%s (%d) ", fname, lnum);
>> }
>> vprintf (format, args); /* MARKER A */
> > vfprintf (log_file, format, args); /* MARKER B */
> >
> > putchar ('\n');
> > putc ('\n', log_file);
> > va_end (args);
> >}
> >
> >int main()
>>{
>> log_file = fopen("outfile","w");
>> infmsg("starting simulation %s",fname);
>> fclose(log_file);
>> return 0;
>>}
>>
-- A S Hodel Assoc. Prof. Dept Elect and Computer Eng, Auburn Univ,AL 36849-5201 On leave at NASA Marshall Space Flight Center (256) 544-1426 Address until 31 July 2000:Mail Code TD-55, MSFC, Alabama, 35812 http://www.eng.auburn.edu/~scotte
This archive was generated by hypermail 2a24 : Mon Jun 19 2000 - 21:22:28 MDT