#include "gdbcore.h"
+#if defined(TARGET_SPARCLET) || defined(TARGET_SPARCLITE)
+#define SPARC_HAS_FPU 0
+#else
+#define SPARC_HAS_FPU 1
+#endif
+
#ifdef GDB_TARGET_IS_SPARC64
#define FP_REGISTER_BYTES (64 * 4)
#else
int deferred_stores = 0; /* Cumulates stores we want to do eventually. */
+
+/* Fetch a single instruction. Even on bi-endian machines
+ such as sparc86x, instructions are always big-endian. */
+
+static unsigned long
+fetch_instruction (pc)
+ CORE_ADDR pc;
+{
+ unsigned long retval;
+ int i;
+ unsigned char buf[4];
+
+ read_memory (pc, buf, sizeof (buf));
+
+ /* Start at the most significant end of the integer, and work towards
+ the least significant. */
+ retval = 0;
+ for (i = 0; i < sizeof (buf); ++i)
+ retval = (retval << 8) | buf[i];
+ return retval;
+}
+
+
/* Branches with prediction are treated like their non-predicting cousins. */
/* FIXME: What about floating point branches? */
/* printf_unfiltered ("set break at %x\n",next_pc); */
pc = read_register (PC_REGNUM);
- pc_instruction = read_memory_integer (pc, 4);
+ pc_instruction = fetch_instruction (pc);
br = isbranch (pc_instruction, pc, &target);
brknpc4 = brktrg = 0;
fi->bottom =
(fi->next ?
(fi->frame == fi->next->frame ? fi->next->bottom : fi->next->frame) :
- read_register (SP_REGNUM));
+ read_sp ());
/* If fi->next is NULL, then we already set ->frame by passing read_fp()
to create_new_frame. */
}
else
{
+ /* Should we adjust for stack bias here? */
get_saved_register (buf, 0, 0, fi, FP_REGNUM, 0);
fi->frame = extract_address (buf, REGISTER_RAW_SIZE (FP_REGNUM));
+#ifdef GDB_TARGET_IS_SPARC64
+ if (fi->frame & 1)
+ fi->frame += 2047;
+#endif
+
}
}
negative number if a flat frame) to the sp. FIXME: Does not
handle large frames which will need more than one instruction
to adjust the sp. */
- insn = read_memory_integer (addr, 4);
+ insn = fetch_instruction (addr);
if (X_OP (insn) == 2 && X_RD (insn) == 14 && X_OP3 (insn) == 0
&& X_I (insn) && X_SIMM13 (insn) < 0)
{
int offset = X_SIMM13 (insn);
/* Then look for a save of %i7 into the frame. */
- insn = read_memory_integer (addr + 4, 4);
+ insn = fetch_instruction (addr + 4);
if (X_OP (insn) == 3
&& X_RD (insn) == 31
&& X_OP3 (insn) == 4
/* Overwrite the frame's address with the value in %i7. */
get_saved_register (buf, 0, 0, fi, I7_REGNUM, 0);
fi->frame = extract_address (buf, REGISTER_RAW_SIZE (I7_REGNUM));
-
+#ifdef GDB_TARGET_IS_SPARC64
+ if (fi->frame & 1)
+ fi->frame += 2047;
+#endif
/* Record where the fp got saved. */
fi->fp_addr = fi->frame + fi->sp_offset + X_SIMM13 (insn);
/* Also try to collect where the pc got saved to. */
fi->pc_addr = 0;
- insn = read_memory_integer (addr + 12, 4);
+ insn = fetch_instruction (addr + 12);
if (X_OP (insn) == 3
&& X_RD (insn) == 15
&& X_OP3 (insn) == 4
CORE_ADDR pc = start_pc;
int is_flat = 0;
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
/* Recognize the `sethi' insn and record its destination. */
if (X_OP (insn) == 0 && X_OP2 (insn) == 4)
{
dest = X_RD (insn);
pc += 4;
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
}
/* Recognize an add immediate value to register to either %g1 or
&& (X_RD (insn) == 1 || X_RD (insn) == dest))
{
pc += 4;
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
}
/* Recognize any SAVE insn. */
pc += 4;
if (frameless_p) /* If the save is all we care about, */
return pc; /* return before doing more work */
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
}
/* Recognize add to %sp. */
else if (X_OP (insn) == 2 && X_RD (insn) == 14 && X_OP3 (insn) == 0)
if (frameless_p) /* If the add is all we care about, */
return pc; /* return before doing more work */
is_flat = 1;
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
/* Recognize store of frame pointer (i7). */
if (X_OP (insn) == 3
&& X_RD (insn) == 31
&& X_RS1 (insn) == 14)
{
pc += 4;
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
/* Recognize sub %sp, <anything>, %i7. */
if (X_OP (insn) == 2
&& X_RD (insn) == 31)
{
pc += 4;
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
}
else
return pc;
else
break;
pc += 4;
- insn = read_memory_integer (pc, 4);
+ insn = fetch_instruction (pc);
}
return pc;
while (frame1 != NULL)
{
if (frame1->pc >= (frame1->bottom ? frame1->bottom :
- read_register (SP_REGNUM))
+ read_sp ())
&& frame1->pc <= FRAME_FP (frame1))
{
/* Dummy frame. All but the window regs are in there somewhere.
/* See tm-sparc.h for how this is calculated. */
#ifdef FP0_REGNUM
#define DUMMY_STACK_REG_BUF_SIZE \
-(((8+8+8) * SPARC_INTREG_SIZE) + (32 * REGISTER_RAW_SIZE (FP0_REGNUM)))
+(((8+8+8) * SPARC_INTREG_SIZE) + FP_REGISTER_BYTES)
#else
#define DUMMY_STACK_REG_BUF_SIZE \
(((8+8+8) * SPARC_INTREG_SIZE) )
CORE_ADDR sp, old_sp;
char register_temp[DUMMY_STACK_SIZE];
- old_sp = sp = read_register (SP_REGNUM);
+ old_sp = sp = read_sp ();
#ifdef GDB_TARGET_IS_SPARC64
+ /* PC, NPC, CCR, FSR, FPRS, Y, ASI */
+ read_register_bytes (REGISTER_BYTE (PC_REGNUM), ®ister_temp[0],
+ REGISTER_RAW_SIZE (PC_REGNUM) * 7);
+ read_register_bytes (REGISTER_BYTE (PSTATE_REGNUM), ®ister_temp[8],
+ REGISTER_RAW_SIZE (PSTATE_REGNUM));
/* FIXME: not sure what needs to be saved here. */
#else
/* Y, PS, WIM, TBR, PC, NPC, FPS, CPS regs */
sp -= DUMMY_STACK_SIZE;
- write_register (SP_REGNUM, sp);
+ write_sp (sp);
write_memory (sp + DUMMY_REG_SAVE_OFFSET, ®ister_temp[0],
DUMMY_STACK_REG_BUF_SIZE);
- write_register (FP_REGNUM, old_sp);
+ if (strcmp (target_shortname, "sim") != 0)
+ {
+ write_fp (old_sp);
- /* Set return address register for the call dummy to the current PC. */
- write_register (I7_REGNUM, read_pc() - 8);
+ /* Set return address register for the call dummy to the current PC. */
+ write_register (I7_REGNUM, read_pc() - 8);
+ }
+ else
+ {
+ /* The call dummy will write this value to FP before executing
+ the 'save'. This ensures that register window flushes work
+ correctly in the simulator. */
+ write_register (G0_REGNUM+1, read_register (FP_REGNUM));
+
+ /* The call dummy will write this value to FP after executing
+ the 'save'. */
+ write_register (G0_REGNUM+2, old_sp);
+
+ /* The call dummy will write this value to the return address (%i7) after
+ executing the 'save'. */
+ write_register (G0_REGNUM+3, read_pc() - 8);
+
+ /* Set the FP that the call dummy will be using after the 'save'.
+ This makes backtraces from an inferior function call work properly. */
+ write_register (FP_REGNUM, old_sp);
+ }
}
/* sparc_frame_find_saved_regs (). This function is here only because
is a return address minus 8.) sparc_pop_frame knows how to
deal with that. Other routines might or might not.
- See tm-sparc.h (PUSH_FRAME and friends) for CRITICAL information
+ See tm-sparc.h (PUSH_DUMMY_FRAME and friends) for CRITICAL information
about how this works. */
static void sparc_frame_find_saved_regs PARAMS ((struct frame_info *,
memset (saved_regs_addr, 0, sizeof (*saved_regs_addr));
if (fi->pc >= (fi->bottom ? fi->bottom :
- read_register (SP_REGNUM))
+ read_sp ())
&& fi->pc <= FRAME_FP(fi))
{
/* Dummy frame. All but the window regs are in there somewhere. */
#ifdef GDB_TARGET_IS_SPARC64
for (regnum = FP0_REGNUM + 32; regnum < FP_MAX_REGNUM; regnum++)
saved_regs_addr->regs[regnum] =
- frame_addr + 32 * 4 + (regnum - FP0_REGNUM - 32) * 8
+ frame_addr + 32 * 4 + (regnum - FP0_REGNUM - 32) * 4
- DUMMY_STACK_REG_BUF_SIZE + 24 * SPARC_INTREG_SIZE;
#endif
#endif /* FP0_REGNUM */
+#ifdef GDB_TARGET_IS_SPARC64
+ for (regnum = PC_REGNUM; regnum < PC_REGNUM + 7; regnum++)
+ {
+ saved_regs_addr->regs[regnum] =
+ frame_addr + (regnum - PC_REGNUM) * SPARC_INTREG_SIZE
+ - DUMMY_STACK_REG_BUF_SIZE;
+ }
+ saved_regs_addr->regs[PSTATE_REGNUM] =
+ frame_addr + 8 * SPARC_INTREG_SIZE - DUMMY_STACK_REG_BUF_SIZE;
+#else
for (regnum = Y_REGNUM; regnum < NUM_REGS; regnum++)
saved_regs_addr->regs[regnum] =
frame_addr + (regnum - Y_REGNUM) * SPARC_INTREG_SIZE
- DUMMY_STACK_REG_BUF_SIZE;
+#endif
frame_addr = fi->bottom ?
- fi->bottom : read_register (SP_REGNUM);
+ fi->bottom : read_sp ();
}
else if (fi->flat)
{
{
/* Normal frame. Just Local and In registers */
frame_addr = fi->bottom ?
- fi->bottom : read_register (SP_REGNUM);
+ fi->bottom : read_sp ();
for (regnum = L0_REGNUM; regnum < L0_REGNUM+8; regnum++)
saved_regs_addr->regs[regnum] =
(frame_addr + (regnum - L0_REGNUM) * SPARC_INTREG_SIZE
CORE_ADDR next_next_frame_addr =
(fi->next->bottom ?
fi->next->bottom :
- read_register (SP_REGNUM));
+ read_sp ());
for (regnum = O0_REGNUM; regnum < O0_REGNUM+8; regnum++)
saved_regs_addr->regs[regnum] =
(next_next_frame_addr
}
}
/* Otherwise, whatever we would get from ptrace(GETREGS) is accurate */
+ /* FIXME -- should this adjust for the sparc64 offset? */
saved_regs_addr->regs[SP_REGNUM] = FRAME_FP (fi);
}
read_memory_integer (fsr.regs[O0_REGNUM + 7],
SPARC_INTREG_SIZE));
- write_register (SP_REGNUM, frame->frame);
+ write_sp (frame->frame);
}
else if (fsr.regs[I0_REGNUM])
{
locals from the registers array, so we need to muck with the
registers array. */
sp = fsr.regs[SP_REGNUM];
+#ifdef GDB_TARGET_IS_SPARC64
+ if (sp & 1)
+ sp += 2047;
+#endif
read_memory (sp, reg_temp, SPARC_INTREG_SIZE * 16);
/* Restore the out registers.
}
#endif /* STATIC_TRANSFORM_NAME */
\f
-#ifdef GDB_TARGET_IS_SPARC64
/* Utilities for printing registers.
Page numbers refer to the SPARC Architecture Manual. */
/* pages 40 - 60 */
switch (regno)
{
+#ifdef GDB_TARGET_IS_SPARC64
case CCR_REGNUM :
printf_unfiltered("\t");
dump_ccreg ("xcc", val >> 4);
case OTHERWIN_REGNUM :
printf ("\t%d", BITS (0, 31));
break;
+#else
+ case PS_REGNUM:
+ printf ("\ticc:%c%c%c%c, pil:%d, s:%d, ps:%d, et:%d, cwp:%d",
+ BITS (23, 1) ? 'N' : '-', BITS (22, 1) ? 'Z' : '-',
+ BITS (21, 1) ? 'V' : '-', BITS (20, 1) ? 'C' : '-',
+ BITS (8, 15), BITS (7, 1), BITS (6, 1), BITS (5, 1),
+ BITS (0, 31));
+ break;
+ case FPS_REGNUM:
+ {
+ static char *fcc[4] = { "=", "<", ">", "?" };
+ static char *rd[4] = { "N", "0", "+", "-" };
+ /* Long, yes, but I'd rather leave it as is and use a wide screen. */
+ printf ("\trd:%s, tem:%d, ns:%d, ver:%d, ftt:%d, qne:%d, "
+ "fcc:%s, aexc:%d, cexc:%d",
+ rd[BITS (30, 3)], BITS (23, 31), BITS (22, 1), BITS (17, 7),
+ BITS (14, 7), BITS (13, 1), fcc[BITS (10, 3)], BITS (5, 31),
+ BITS (0, 31));
+ break;
+ }
+
+#endif /* GDB_TARGET_IS_SPARC64 */
}
#undef BITS
}
-
-#endif
\f
int
gdb_print_insn_sparc (memaddr, info)
return sp;
}
+
+
+/* Extract from an array REGBUF containing the (raw) register state
+ a function return value of type TYPE, and copy that, in virtual format,
+ into VALBUF. */
+
+void
+sparc_extract_return_value (type, regbuf, valbuf)
+ struct type *type;
+ char *regbuf;
+ char *valbuf;
+{
+ int typelen = TYPE_LENGTH (type);
+ int regsize = REGISTER_RAW_SIZE (O0_REGNUM);
+
+ if (TYPE_CODE (type) == TYPE_CODE_FLT && SPARC_HAS_FPU)
+ memcpy (valbuf, ®buf [REGISTER_BYTE (FP0_REGNUM)], typelen);
+ else
+ memcpy (valbuf,
+ ®buf [O0_REGNUM * regsize +
+ (typelen >= regsize ? 0 : regsize - typelen)],
+ typelen);
+}
+
+
+/* Write into appropriate registers a function return value
+ of type TYPE, given in virtual format. On SPARCs with FPUs,
+ float values are returned in %f0 (and %f1). In all other cases,
+ values are returned in register %o0. */
+
+void
+sparc_store_return_value (type, valbuf)
+ struct type *type;
+ char *valbuf;
+{
+ int regno;
+ char buffer[MAX_REGISTER_RAW_SIZE];
+
+ if (TYPE_CODE (type) == TYPE_CODE_FLT && SPARC_HAS_FPU)
+ /* Floating-point values are returned in the register pair */
+ /* formed by %f0 and %f1 (doubles are, anyway). */
+ regno = FP0_REGNUM;
+ else
+ /* Other values are returned in register %o0. */
+ regno = O0_REGNUM;
+
+ /* Add leading zeros to the value. */
+ if (TYPE_LENGTH (type) < REGISTER_RAW_SIZE(regno))
+ {
+ bzero (buffer, REGISTER_RAW_SIZE(regno));
+ memcpy (buffer + REGISTER_RAW_SIZE(regno) - TYPE_LENGTH (type), valbuf,
+ TYPE_LENGTH (type));
+ write_register_bytes (REGISTER_BYTE (regno), buffer,
+ REGISTER_RAW_SIZE(regno));
+ }
+ else
+ write_register_bytes (REGISTER_BYTE (regno), valbuf, TYPE_LENGTH (type));
+}
+
+
+/* Insert the function address into a call dummy instsruction sequence
+ stored at DUMMY.
+
+ For structs and unions, if the function was compiled with Sun cc,
+ it expects 'unimp' after the call. But gcc doesn't use that
+ (twisted) convention. So leave a nop there for gcc (FIX_CALL_DUMMY
+ can assume it is operating on a pristine CALL_DUMMY, not one that
+ has already been customized for a different function). */
+
+void
+sparc_fix_call_dummy (dummy, pc, fun, value_type, using_gcc)
+ char *dummy;
+ CORE_ADDR pc;
+ CORE_ADDR fun;
+ struct type *value_type;
+ int using_gcc;
+{
+ int i;
+
+ /* Store the relative adddress of the target function into the
+ 'call' instruction. */
+ store_unsigned_integer (dummy + CALL_DUMMY_CALL_OFFSET, 4,
+ (0x40000000
+ | (((fun - (pc + CALL_DUMMY_CALL_OFFSET)) >> 2)
+ & 0x3fffffff)));
+
+ /* Comply with strange Sun cc calling convention for struct-returning
+ functions. */
+ if (!using_gcc
+ && (TYPE_CODE (value_type) == TYPE_CODE_STRUCT
+ || TYPE_CODE (value_type) == TYPE_CODE_UNION))
+ store_unsigned_integer (dummy + CALL_DUMMY_CALL_OFFSET + 8, 4,
+ TYPE_LENGTH (value_type) & 0x1fff);
+
+#ifndef GDB_TARGET_IS_SPARC64
+ /* If this is not a simulator target, change the first four instructions
+ of the call dummy to NOPs. Those instructions include a 'save'
+ instruction and are designed to work around problems with register
+ window flushing in the simulator. */
+ if (strcmp (target_shortname, "sim") != 0)
+ {
+ for (i = 0; i < 4; i++)
+ store_unsigned_integer (dummy + (i * 4), 4, 0x01000000);
+ }
+#endif
+}
+
+
+/* Set target byte order based on machine type. */
+
+static int
+sparc_target_architecture_hook (ap)
+ const bfd_arch_info_type *ap;
+{
+ int i, j;
+
+ if (ap->mach == bfd_mach_sparc_sparclite_le)
+ target_byte_order = LITTLE_ENDIAN;
+ return 1;
+}
+
\f
void
_initialize_sparc_tdep ()
{
tm_print_insn = gdb_print_insn_sparc;
tm_print_insn_info.mach = TM_PRINT_INSN_MACH; /* Selects sparc/sparclite */
+ target_architecture_hook = sparc_target_architecture_hook;
+}
+
+
+#ifdef GDB_TARGET_IS_SPARC64
+
+/* Compensate for stack bias. Note that we currently don't handle mixed
+ 32/64 bit code. */
+CORE_ADDR
+sparc64_read_sp ()
+{
+ CORE_ADDR sp = read_register (SP_REGNUM);
+
+ if (sp & 1)
+ sp += 2047;
+ return sp;
+}
+
+CORE_ADDR
+sparc64_read_fp ()
+{
+ CORE_ADDR fp = read_register (FP_REGNUM);
+
+ if (fp & 1)
+ fp += 2047;
+ return fp;
+}
+
+void
+sparc64_write_sp (val)
+ CORE_ADDR val;
+{
+ CORE_ADDR oldsp = read_register (SP_REGNUM);
+ if (oldsp & 1)
+ write_register (SP_REGNUM, val - 2047);
+ else
+ write_register (SP_REGNUM, val);
+}
+
+void
+sparc64_write_fp (val)
+ CORE_ADDR val;
+{
+ CORE_ADDR oldfp = read_register (FP_REGNUM);
+ if (oldfp & 1)
+ write_register (FP_REGNUM, val - 2047);
+ else
+ write_register (FP_REGNUM, val);
+}
+
+/* The SPARC 64 ABI passes floating-point arguments in FP0-31. They are
+ also copied onto the stack in the correct places. */
+
+CORE_ADDR
+sp64_push_arguments (nargs, args, sp, struct_return, struct_retaddr)
+ int nargs;
+ value_ptr *args;
+ CORE_ADDR sp;
+ unsigned char struct_return;
+ CORE_ADDR struct_retaddr;
+{
+ int x;
+ int regnum = 0;
+ CORE_ADDR tempsp;
+
+ sp = (sp & ~(((unsigned long)TYPE_LENGTH (builtin_type_long)) - 1UL));
+
+ /* Figure out how much space we'll need. */
+ for (x = nargs - 1; x >= 0; x--)
+ {
+ int len = TYPE_LENGTH (check_typedef (VALUE_TYPE (args[x])));
+ value_ptr copyarg = args[x];
+ int copylen = len;
+
+ /* This code is, of course, no longer correct. */
+ if (copylen < TYPE_LENGTH (builtin_type_long))
+ {
+ copyarg = value_cast(builtin_type_long, copyarg);
+ copylen = TYPE_LENGTH (builtin_type_long);
+ }
+ sp -= copylen;
+ }
+
+ /* Round down. */
+ sp = sp & ~7;
+ tempsp = sp;
+
+ /* Now write the arguments onto the stack, while writing FP arguments
+ into the FP registers. */
+ for (x = 0; x < nargs; x++)
+ {
+ int len = TYPE_LENGTH (check_typedef (VALUE_TYPE (args[x])));
+ value_ptr copyarg = args[x];
+ int copylen = len;
+
+ /* This code is, of course, no longer correct. */
+ if (copylen < TYPE_LENGTH (builtin_type_long))
+ {
+ copyarg = value_cast(builtin_type_long, copyarg);
+ copylen = TYPE_LENGTH (builtin_type_long);
+ }
+ write_memory (tempsp, VALUE_CONTENTS (copyarg), copylen);
+ tempsp += copylen;
+ if (TYPE_CODE (VALUE_TYPE (args[x])) == TYPE_CODE_FLT && regnum < 32)
+ {
+ /* This gets copied into a FP register. */
+ int nextreg = regnum + 2;
+ char *data = VALUE_CONTENTS (args[x]);
+ /* Floats go into the lower half of a FP register pair; quads
+ use 2 pairs. */
+
+ if (len == 16)
+ nextreg += 2;
+ else if (len == 4)
+ regnum++;
+
+ write_register_bytes (REGISTER_BYTE (FP0_REGNUM + regnum),
+ data,
+ len);
+ regnum = nextreg;
+ }
+ }
+ return sp;
+}
+
+/* Values <= 32 bytes are returned in o0-o3 (floating-point values are
+ returned in f0-f3). */
+void
+sparc64_extract_return_value (type, regbuf, valbuf, bitoffset)
+ struct type *type;
+ char *regbuf;
+ char *valbuf;
+ int bitoffset;
+{
+ int typelen = TYPE_LENGTH (type);
+ int regsize = REGISTER_RAW_SIZE (O0_REGNUM);
+
+ if (TYPE_CODE (type) == TYPE_CODE_FLT && SPARC_HAS_FPU)
+ {
+ memcpy (valbuf, ®buf [REGISTER_BYTE (FP0_REGNUM)], typelen);
+ return;
+ }
+
+ if (TYPE_CODE (type) != TYPE_CODE_STRUCT
+ || (TYPE_LENGTH (type) > 32))
+ {
+ memcpy (valbuf,
+ ®buf [O0_REGNUM * regsize +
+ (typelen >= regsize ? 0 : regsize - typelen)],
+ typelen);
+ return;
+ }
+ else
+ {
+ char *o0 = ®buf[O0_REGNUM * regsize];
+ char *f0 = ®buf[FP0_REGNUM * regsize];
+ int x;
+
+ for (x = 0; x < TYPE_NFIELDS (type); x++)
+ {
+ struct field *f = &TYPE_FIELDS(type)[x];
+ /* FIXME: We may need to handle static fields here. */
+ int whichreg = (f->loc.bitpos + bitoffset) / 32;
+ int remainder = ((f->loc.bitpos + bitoffset) % 32) / 8;
+ int where = (f->loc.bitpos + bitoffset) / 8;
+ int size = TYPE_LENGTH (f->type);
+ int typecode = TYPE_CODE (f->type);
+
+ if (typecode == TYPE_CODE_STRUCT)
+ {
+ sparc64_extract_return_value (f->type,
+ regbuf,
+ valbuf,
+ bitoffset + f->loc.bitpos);
+ }
+ else if (typecode == TYPE_CODE_FLT)
+ {
+ memcpy (valbuf + where, &f0[whichreg * 4] + remainder, size);
+ }
+ else
+ {
+ memcpy (valbuf + where, &o0[whichreg * 4] + remainder, size);
+ }
+ }
+ }
}
+#endif
projects / binutils-gdb.git / commitdiff