darkkirb/old-cross-binutils
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
old-cross-binutils/gdb/nat/gdb_ptrace.h
Pedro Alves 5401971915 C++: handle glibc's ptrace(enum __ptrace_request, ...) 
Building in C++ mode issues ~40 warnings like this:

 ../../src/gdb/linux-nat.c: In function ‘int linux_handle_extended_wait(lwp_info*, int, int)’:
 ../../src/gdb/linux-nat.c:2016:51: warning: invalid conversion from ‘int’ to ‘__ptrace_request’ [-fpermissive]
	ptrace (PTRACE_GETEVENTMSG, pid, 0, &new_pid);

The issue is that in glibc, ptrace's first parameter is an enum.
That's not a problem if we pick the PTRACE_XXX requests from
sys/ptrace.h, as those will be values of the corresponding enum.
However, we have fallback definitions for PTRACE_XXX symbols when the
system headers miss them (such as PTRACE_GETEVENTMSG above), and those
are plain integer constants.  E.g., nat/linux-ptrace.h:

 #define PTRACE_GETEVENTMSG	0x4201

One idea would be to fix this by defining those fallbacks like:

 -#define PTRACE_GETEVENTMSG	0x4201
 +#define PTRACE_GETEVENTMSG	((enum __ptrace_request) 0x4201)

However, while glibc's ptrace uses enum __ptrace_request for first
parameter:

  extern long int ptrace (enum __ptrace_request __request, ...) __THROW;

other libc's, like e.g., Android's bionic do not -- in that case, the
first parameter is int:

  long ptrace(int request, pid_t pid, void * addr, void * data);

So the fix I came up is to make configure/ptrace.m4 also detect the
type of the ptrace's first parameter and defin PTRACE_TYPE_ARG1, as
already does the for parameters 3-4, and then simply wrap ptrace with
a macro that casts the first argument to the detected type.  (I'm
leaving adding a nicer wrapper for when we drop building in C).

While this adds the wrapper, GNU/Linux files won't use it until the
next patch, which makes all native GNU/Linux files include
gdb_ptrace.h.

gdb/ChangeLog:
2015-07-24  Pedro Alves  <palves@redhat.com>

	* ptrace.m4 (ptrace tests): Test in C++ mode.  Try with 'enum
	__ptrace_request as first parameter type instead of int.
	(PTRACE_TYPE_ARG1): Define.
	* nat/gdb_ptrace.h [!PTRACE_TYPE_ARG5] (ptrace): Define as wrapper
	that casts first argument to PTRACE_TYPE_ARG1.
	* config.in: Regenerate.
	* configure: Regenerate.

gdb/gdbserver/ChangeLog:
2015-07-24  Pedro Alves  <palves@redhat.com>

	* config.in: Regenerate.
	* configure: Regenerate.
2015-07-24 15:12:15 +01:00

153 lines
4.5 KiB
C
Raw Blame History

/* Portable <sys/ptrace.h>
Copyright (C) 2004-2015 Free Software Foundation, Inc.
This file is part of GDB.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
#ifndef GDB_PTRACE_H
#define GDB_PTRACE_H
/* The <sys/ptrace.h> header was introduced with 4.4BSD, and provided
the PT_* symbolic constants for the ptrace(2) request numbers. The
ptrace(2) prototype was added later to the same header on BSD.
SunOS and GNU/Linux have slightly different symbolic names for the
constants that start with PTRACE_*. System V still doesn't have
(and probably never will have) a <sys/ptrace.h> with symbolic
constants; the ptrace(2) prototype can be found in <unistd.h>.
Fortunately all systems use the same numerical constants for the
common ptrace requests. */
#ifdef HAVE_PTRACE_H
# include <ptrace.h>
#elif defined(HAVE_SYS_PTRACE_H)
# include <sys/ptrace.h>
#endif
/* No need to include <unistd.h> since it's already included by
"defs.h". */
#ifndef PT_TRACE_ME
# define PT_TRACE_ME 0
#endif
#ifndef PT_READ_I
# define PT_READ_I 1 /* Read word in child's I space. */
#endif
#ifndef PT_READ_D
# define PT_READ_D 2 /* Read word in child's D space. */
#endif
#ifndef PT_READ_U
# define PT_READ_U 3 /* Read word in child's U space. */
#endif
#ifndef PT_WRITE_I
# define PT_WRITE_I 4 /* Write word in child's I space. */
#endif
#ifndef PT_WRITE_D
# define PT_WRITE_D 5 /* Write word in child's D space. */
#endif
#ifndef PT_WRITE_U
# define PT_WRITE_U 6 /* Write word in child's U space. */
#endif
/* HP-UX doesn't define PT_CONTINUE and PT_STEP. Instead of those two
ptrace requests, it has PT_CONTIN, PT_CONTIN1, PT_SINGLE and
PT_SINGLE1. PT_CONTIN1 and PT_SINGLE1 preserve pending signals,
which apparently is what is wanted by the HP-UX native code. */
#ifndef PT_CONTINUE
# ifdef PT_CONTIN1
# define PT_CONTINUE PT_CONTIN1
# else
# define PT_CONTINUE 7 /* Continue the child. */
# endif
#endif
#ifndef PT_KILL
# define PT_KILL 8 /* Kill the child process. */
#endif
#ifndef PT_STEP
# ifdef PT_SINGLE1
# define PT_STEP PT_SINGLE1
# else
# define PT_STEP 9 /* Single step the child. */
# endif
#endif
/* Not all systems support attaching and detaching. */
#ifndef PT_ATTACH
# ifdef PTRACE_ATTACH
# define PT_ATTACH PTRACE_ATTACH
# endif
#endif
#ifndef PT_DETACH
# ifdef PTRACE_DETACH
# define PT_DETACH PTRACE_DETACH
# endif
#endif
/* For systems such as HP/UX that do not provide PT_SYSCALL, define it
here as an alias for PT_CONTINUE. This is what the PT_SYSCALL
request is expected to do, in addition to stopping when entering/
exiting a system call. Chances are, if the system supports system
call tracing, enabling this feature is probably done separately;
and there is probably no special request that we would be required
to use when resuming the execution of our program. */
#ifndef PT_SYSCALL
# ifdef PTRACE_SYSCALL
# define PT_SYSCALL PTRACE_SYSCALL
#else
# define PT_SYSCALL PT_CONTINUE
# endif
#endif
/* Some systems, in particular DEC OSF/1, Digital Unix, Compaq Tru64
or whatever it's called these days, don't provide a prototype for
ptrace. Provide one to silence compiler warnings. */
#ifndef HAVE_DECL_PTRACE
extern PTRACE_TYPE_RET ptrace();
#endif
/* Some systems, at least AIX and HP-UX have a ptrace with five
arguments. Since we never use the fifth argument, define a ptrace
macro that calls the real ptrace with the last argument set to
zero. */
#ifdef PTRACE_TYPE_ARG5
# ifdef HAVE_PTRACE64
# define ptrace(request, pid, addr, data) \
ptrace64 (request, pid, addr, data, 0)
# undef PTRACE_TYPE_ARG3
# define PTRACE_TYPE_ARG3 long long
# else
# define ptrace(request, pid, addr, data) \
ptrace (request, pid, addr, data, 0)
# endif
#else
/* Wrapper that avoids adding a pointless cast to all callers. */
# define ptrace(request, pid, addr, data) \
ptrace ((PTRACE_TYPE_ARG1) request, pid, addr, data)
#endif
#endif /* gdb_ptrace.h */
Reference in a new issue View git blame Copy permalink