RE: FW: Interrupted system call problem

RE: FW: Interrupted system call problem

[deCarmo, Linden]

I pulled down the latest snapshot and everything works now.  Weird.

> -----Original Message-----
> From: Keith Seitz [ mailto:keiths@cygnus.com ]
> Sent: Friday, April 13, 2001 5:59 PM
> To: deCarmo, Linden
> Cc: 'insight@sourceware.cygnus.com'
> Subject: Re: FW: Interrupted system call problem
> 
> 
> On Thu, 12 Apr 2001, deCarmo, Linden wrote:
> 
> > I really would be grateful if someone had advice about how 
> to work around or
> > debug this.  It isn't a gdb issue and appears to be 
> specific to Insight and
> > select() on Solaris 8.  I even have tried multiple Solaris 8 boxes.
> 
> I've built a Solaris 8 Insight, and I don't see this problem.
> 
> Here's my testcase:
> 
> $ cat systemcall.c
> #include <unistd.h>
> #include <sys/time.h>
> #include <stdio.h>
> 
> int
> my_wait (long seconds, long microseconds)
> {
>   struct timeval tv;
> 
>   tv.tv_sec  = seconds;
>   tv.tv_usec = microseconds;
>   return (select (0, NULL, NULL, NULL, &tv));
> }
> 
> int
> main (int argc, char *argv[])
> {
>    int i;
> 
>    i = my_wait (15, 12345);
>    printf ("my_wait returned %d\n", i);
> 
>    exit (0);
> }
> $ gcc -g systemcall.c -o systemcall
> $ gdb -nw -q systemcall
> Reading symbols from systemcall...done.
> (gdb) r
> Starting program: /home/keiths/systemcall/systemcall 
> ^C
> Program received signal SIGINT, Interrupt.
> 0xff317ac8 in ?? ()
> (gdb) bt
> #0  0xff317ac8 in ?? ()
> #1  0x00010788 in my_wait (seconds=15, microseconds=12345) at 
> systemcall.c:12
> #2  0x000107bc in main (argc=1, argv=0xffbefb64) at systemcall.c:20
> (gdb) disassemble
> No function contains program counter for selected frame.
> (gdb)
> 
> If I do this with Insight, Insight stops but displays no 
> source and no 
> disassembly. It simply says "Select function to disassemble". Gdb is 
> lost, and, consequently, so is Insight.
> 
> This is with sources updated today from sources.redhat.com. Does this 
> testcase behave any differently for you?
> 
> Keith
> 

Re: FW: Interrupted system call problem

[Keith Seitz]

On Thu, 12 Apr 2001, deCarmo, Linden wrote:

> I really would be grateful if someone had advice about how to work around or
> debug this.  It isn't a gdb issue and appears to be specific to Insight and
> select() on Solaris 8.  I even have tried multiple Solaris 8 boxes.

I've built a Solaris 8 Insight, and I don't see this problem.

Here's my testcase:

$ cat systemcall.c
#include <unistd.h>
#include <sys/time.h>
#include <stdio.h>

int
my_wait (long seconds, long microseconds)
{
  struct timeval tv;

  tv.tv_sec  = seconds;
  tv.tv_usec = microseconds;
  return (select (0, NULL, NULL, NULL, &tv));
}

int
main (int argc, char *argv[])
{
   int i;

   i = my_wait (15, 12345);
   printf ("my_wait returned %d\n", i);

   exit (0);
}
$ gcc -g systemcall.c -o systemcall
$ gdb -nw -q systemcall
Reading symbols from systemcall...done.
(gdb) r
Starting program: /home/keiths/systemcall/systemcall 
^C
Program received signal SIGINT, Interrupt.
0xff317ac8 in ?? ()
(gdb) bt
#0  0xff317ac8 in ?? ()
#1  0x00010788 in my_wait (seconds=15, microseconds=12345) at systemcall.c:12
#2  0x000107bc in main (argc=1, argv=0xffbefb64) at systemcall.c:20
(gdb) disassemble
No function contains program counter for selected frame.
(gdb)

If I do this with Insight, Insight stops but displays no source and no 
disassembly. It simply says "Select function to disassemble". Gdb is 
lost, and, consequently, so is Insight.

This is with sources updated today from sources.redhat.com. Does this 
testcase behave any differently for you?

Keith

FW: Interrupted system call problem

[deCarmo, Linden]

I really would be grateful if someone had advice about how to work around or
debug this.  It isn't a gdb issue and appears to be specific to Insight and
select() on Solaris 8.  I even have tried multiple Solaris 8 boxes.

> -----Original Message-----
> From: deCarmo, Linden 
> Sent: Tuesday, April 03, 2001 8:24 AM
> To: 'insight@sourceware.cygnus.com'
> Subject: FW: Interrupted system call problem
> 
> 
> Just noticed that I forgot to copy the list on my response.
> 
> Any ideas to work around this?
> 
> > On Tue, 20 Mar 2001, deCarmo, Linden wrote:
> > 
> > > Ok, the latest snapshot hangs rather than reports 
> > "Interrupted system call"
> > > whenever the the code below is executed.  To be specific, 
> the source
> > > disappears and it reports:
> > > 
> > > Select function name to disassemble
> > > 
> > > ====
> > > 
> > > The stack is listed as ???? 
> > 
> > You're probably in a dynamic loader stub or similar, sitting 
> > at the first 
> > instruction. Gdb is obviously a little confused about where 
> it is... 
> > (This should not happen.)
> > 
> > > struct timeval tv;
> > >  	336	
> > > -	337		tv.tv_sec = seconds;
> > >  	338		tv.tv_usec = microseconds;
> > > -	340		return(select(0,NULL,NULL,NULL,&tv));
> > 
> > Let me guess, it does this on line 340? Then I would guess 
> > that shared 
> > libraries are a little broken on your system. Hmmm.... 
> > Solaris 2.8...  
> >
> 
> Works fine with ddd/gdb, and gdb -nw.
>  
> > > Any idea what I can try next?  Its weird that ddd doesn't 
> > have the same
> > > issue using gdb.
> > 
> > Does command line gdb exhibit the same behavior (run the 
> insight gdb 
> > with the "-nw" flag)?
> > 
> 
> No, works ok with -nw.  Only Insight exhibits this behavior.
>  
> 

FW: Interrupted system call problem

[deCarmo, Linden]

Just noticed that I forgot to copy the list on my response.

Any ideas to work around this?

> On Tue, 20 Mar 2001, deCarmo, Linden wrote:
> 
> > Ok, the latest snapshot hangs rather than reports 
> "Interrupted system call"
> > whenever the the code below is executed.  To be specific, the source
> > disappears and it reports:
> > 
> > Select function name to disassemble
> > 
> > ====
> > 
> > The stack is listed as ???? 
> 
> You're probably in a dynamic loader stub or similar, sitting 
> at the first 
> instruction. Gdb is obviously a little confused about where it is... 
> (This should not happen.)
> 
> > struct timeval tv;
> >  	336	
> > -	337		tv.tv_sec = seconds;
> >  	338		tv.tv_usec = microseconds;
> > -	340		return(select(0,NULL,NULL,NULL,&tv));
> 
> Let me guess, it does this on line 340? Then I would guess 
> that shared 
> libraries are a little broken on your system. Hmmm.... 
> Solaris 2.8...  
>

Works fine with ddd/gdb, and gdb -nw.
 
> > Any idea what I can try next?  Its weird that ddd doesn't 
> have the same
> > issue using gdb.
> 
> Does command line gdb exhibit the same behavior (run the insight gdb 
> with the "-nw" flag)?
> 

No, works ok with -nw.  Only Insight exhibits this behavior.