[systemd-devel] [PATCH] crash on suspend/resume ...

[Michael Meeks]

On Tue, 2014-02-11 at 16:11 +0100, Zbigniew Jędrzejewski-Szmek wrote:
> Printing of the three arguments is supposed to be conditional on the
> same condtion (object != NULL). Anyway, in your backtrace:

	Fair enough =)

> #9  log_do_header (header=header at entry=0xbfeef2ec
> "PRIORITY=6\nSYSLOG_FACILITY=3\nCODE_FILE=", level=level at entry=30, 
>     file=file at entry=0x80e5c6e "src/core/manager.c", line=line at entry=1736,
> func=func at entry=0x80e7556 <__func__.13183> "process_event", 
>     object_name=object_name at entry=0x0, object=object at entry=0x0, size=2048) at
> src/shared/log.c:445
> 
> ...they are both null, so the patch shouldn't change anything.

	True; the header contains:

$23 = "PRIORITY=6\nSYSLOG_FACILITY=3\nCODE_FILE=", '\000' <repeats 1113
times>...

	by the time it crashes; so - presumably:

	   file ? "CODE_FILE=" : "",
	   file ? LINE_MAX : 0, file, /* %.0s means no output */
	   file ? "\n" : "",

	is near the badness, but ...

	No idea; dies in:

#5  0xb74fb7c6 in _IO_vfprintf_internal (s=s at entry=0xbfeeee20,
format=<optimized out>, 
    format at entry=0x8117dec "PRIORITY=%i\nSYSLOG_FACILITY=%i\n%s%.*s%s%s
%.*i%s%s%.*s%s%s%.*s%sSYSLOG_IDENTIFIER=%s\n", 
    ap=0xbfeeef78 "V\216\021\b\326|\021\b\001", ap at entry=0xbfeeef64
"\006") at vfprintf.c:1629
1629		  process_string_arg (((struct printf_spec *) NULL));
(gdb) l
1624	
1625	      /* Process current format.  */
1626	      while (1)
1627		{
1628		  process_arg (((struct printf_spec *) NULL));
1629		  process_string_arg (((struct printf_spec *) NULL));
1630	
1631		LABEL (form_unknown):
1632		  if (spec == L_('\0'))
1633		    {

inside:

(gdb) down
#4  __strnlen_sse2 ()
at ../sysdeps/i386/i686/multiarch/strlen-sse2.S:125
125		pcmpeqb	(%eax), %xmm0

	an inlined strlen.

	All of the interesting process_string_arg variables are 'optimized out'
which is a shame.

	Hey ho,

		Michael.

-- 
 michael.meeks at collabora.com  <><, Pseudo Engineer, itinerant idiot