This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: PRecord sets memory even when it says it did not


 

> -----Original Message-----
> From: Hui Zhu [mailto:teawater@gmail.com] 
> Sent: Monday, September 14, 2009 12:15 PM
> To: Daniel Jacobowitz; Marc Khouzam; gdb@sourceware.org; 
> Michael Snyder; gdb-patches ml
> Subject: Re: PRecord sets memory even when it says it did not
> 
> On Mon, Sep 14, 2009 at 23:53, Daniel Jacobowitz 
> <drow@false.org> wrote:
> > On Mon, Sep 14, 2009 at 12:39:35PM +0800, Hui Zhu wrote:
> >> > (gdb) set var a = 8
> >> > Because GDB is in replay mode, writing to memory will 
> make the execution log unusable from this point onward. ?
> Write memory at address 0xbffff6a0?(y or [n]) n
> >> > Process record canceled the operation.
> >> > (gdb) p a
> >> > $2 = 8
> >
> > This should refetch the value from the target. ?What value 
> is cached?
> 
> This value didn't really changed.  It just be changed in cache, but
> not memory.  If you s or rs and check "a" again, you will found that
> "a"'s value is right.

I think Hui is right.  I ran another test with printf and stopping precord
and the memory seems ok, please see session below.

> BTW, this is really a strange thing.  If answer is enter, everything
> is OK.  If answer is no, it was changed.

For me the problem happens both times, when pressing 'n' or <enter>.
(you can see that in the session below too)

(gdb) l
1       #include <stdio.h>
2       int main()
3       {
4         int a = 0;
5         printf("a is %d\n",a);
6         return 0;
7       }
(gdb) start
Temporary breakpoint 1 at 0x8048485: file a.cc, line 4.
Starting program: /local/lmckhou/testing/a.out 
re
Temporary breakpoint 1, main () at a.cc:4
4         int a = 0;
(gdb) rec
(gdb) n
5         printf("a is %d\n",a);
(gdb) n
a is 0
6         return 0;
(gdb) rn
5         printf("a is %d\n",a);
(gdb) p a
$1 = 0
(gdb) set var a=8
Because GDB is in replay mode, writing to memory will make the execution log unusable from this point onward.  Write memory at address 0xbfffadf0?(y or [n]) 
Process record canceled the operation.
(gdb) p a
$2 = 8
(gdb) rec stop
Delete recorded log and stop recording?(y or n) y
(gdb) p a
$3 = 8
(gdb) n
a is 0        <------------------------------------------ a is actually set correctly
6         return 0;
(gdb) p a
$4 = 0

> 
> 
> >
> >> 2009-09-14 ?Hui Zhu ?<teawater@gmail.com>
> >>
> >> ? ? ? * record.c (record_xfer_partial): Call free_all_values when
> >> ? ? ? cancel the operation.
> >
> > I don't think this is a good idea; this is a memory management
> > function. ?It's not supposed to change the apparent values.
> >
> 
> Agree.  Looks we have a strange thing need to be handled.
> 
> 
> Thanks,
> Hui
> 


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]