Re: gEDA-user: pcb-20091103 for Gentoo, working on ebuild

[Peter Clifton <pcjc2@xxxxxxxxx>]

On Sat, 2009-11-07 at 17:47 +0000, Ineiev wrote:

> > hid/gtk/gui-dialog.c:172: warning: format not a string literal and no format arguments
> 
> Should be like it is: the function semantics differ from the printf() ones.

The correct fix for any bug sprewing this message is usually to pass the
format string as "%s", then the message as a format argument. This
aviods the possibility of exploits from a user-derived message string
getting into the printf type routine:

In this case, CHANGE:

  dialog = gtk_message_dialog_new (GTK_WINDOW (out->top_window),
				   GTK_DIALOG_MODAL |
				   GTK_DIALOG_DESTROY_WITH_PARENT,
				   GTK_MESSAGE_QUESTION,
				   GTK_BUTTONS_NONE, message);

TO:

  dialog = gtk_message_dialog_new (GTK_WINDOW (out->top_window),
   GTK_DIALOG_MODAL |
   GTK_DIALOG_DESTROY_WITH_PARENT,
   GTK_MESSAGE_QUESTION,
   GTK_BUTTONS_NONE, "%s", message);


> > flags.c:226: warning: cast from pointer to integer of different size
> 
> OffsetOf() shall not overflow even 16 bits when MAX_LAYERS < 2048.

We could probably cast the warning away. Our "OffsetOf" macro should use
a large enough integer type to match the pointer on the system, and by
in specific cases, we might choose to cast that into a smaller integer
type.


> > parse_l.l:198: warning: ignoring return value of ‘fwrite’, declared with attribute warn_unused_result
> > parse_l.c:1508: warning: ‘input’ defined but not used
> > res_lex.l:64: warning: ignoring return value of ‘fwrite’, declared with attribute warn_unused_result
> > res_lex.c:1139: warning: ‘input’ defined but not used
> 
> Parser generator artifacts IMVHO.

Looks like. You might be able to kludge around the "input" one, possibly
by taking the address of that function into some void* pointer in the
source .l file, then casting it away, but it is probably not worth it.

I don't get the "fwrite" one, as presumably later flex has fixed the
generated code for that case.

> > /usr/include/bits/string3.h:153: warning: call to __builtin___strncat_chk might overflow destination buffer
> > /usr/include/bits/string3.h:153: warning: call to __builtin___strncat_chk might overflow destination buffer
>
> No idea.


This definition looks like the required call:


extern __inline __attribute__ ((__always_inline__)) __attribute__ ((__gnu_inline__, __artificial__)) char *
__attribute__ ((__nothrow__)) strncat (char *__restrict __dest, __const char *__restrict __src, size_t __len)

{
  return __builtin___strncat_chk (__dest, __src, __len, __builtin_object_size (__dest, 2 > 1));
}

That file calls "strncat" in several places, on a string "accel" which
is 64 char long. Each call is passed "sizeof (accel)" as its "n"
argument, so it shouldn't overflow as far as I can see.

AHHA...

>From man 3 strncat:

       If src contains n or more characters, strncat() writes  n+1  characters
       to  dest  (n  from src plus the terminating null byte).  Therefore, the
       size of dest must be at least strlen(dest)+n+1.

So we need to pass "sizeof (accel) - 1".



> > hid/common/actions.c:214: warning: ignoring return value of ‘fwrite’, declared with attribute warn_unused_result
> > hid/common/actions.c:215: warning: ignoring return value of ‘fwrite’, declared with attribute warn_unused_result
> 
> If an error occurs while writing into stderr, probably the most useful
> reaction is to ignore it.

True.. I'm not sure if we can cast it away or not. I remember reading
some notes about GCC not liking you to cast such warnings away.

if (fwrite...) { /* NOP */ }

Or some portable, non-compiler whining alternative?


Thanks for looking into all of these Ineiev. (BTW, Please bug me if you
have any outstanding patches I promised to review / apply. I recall
layer colours.. )

Best wishes,

Peter C.





_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user