| Message ID | 527D12A4.6070109@linaro.org |
|---|---|
| State | Accepted |
| Headers | show |
On 08/11/13 16:34, Will Newton wrote: > > Since as far back as the beginning of the sourceware repository > the ARM port has printed an error "Infinite loop detected" when > the next_pc calculated is the same as the current one, for example > when encountering a branch to the current PC address. > > This causes the test gdb.base/random-signal.exp as the error message > is not expected. I have not been able to find a good reason for the > message to be here so remove it and let the test pass. > > gdb/ChangeLog: > > 2013-11-08 Will Newton <will.newton@linaro.org> > > * arm-tdep.c (arm_get_next_pc): Remove "Infinite loop detected" > error message. I guess the reasoning for this is that you can't single-step a 'branch-to-self' instruction, since the next instruction (where the new breakpoint has to go) is the same as the current instruction. However, it ought to be the higher levels of GDB that handle this, not some random low-level function like get-next-pc. Ok. R.
On 11/19/2013 03:06 PM, Richard Earnshaw wrote: > On 08/11/13 16:34, Will Newton wrote: >> >> Since as far back as the beginning of the sourceware repository >> the ARM port has printed an error "Infinite loop detected" when >> the next_pc calculated is the same as the current one, for example >> when encountering a branch to the current PC address. >> >> This causes the test gdb.base/random-signal.exp as the error message >> is not expected. I have not been able to find a good reason for the >> message to be here so remove it and let the test pass. >> >> gdb/ChangeLog: >> >> 2013-11-08 Will Newton <will.newton@linaro.org> >> >> * arm-tdep.c (arm_get_next_pc): Remove "Infinite loop detected" >> error message. > > I guess the reasoning for this is that you can't single-step a > 'branch-to-self' instruction, since the next instruction (where the new > breakpoint has to go) is the same as the current instruction. You should. It should stop again in the same place. > However, it ought to be the higher levels of GDB that handle this, not > some random low-level function like get-next-pc. > I just tried it on x86-64 (on a gdb branch that makes x86 use software single-step), and it works as expected: (gdb) disassemble Dump of assembler code for function main: 0x000000000040049c <+0>: push %rbp 0x000000000040049d <+1>: mov %rsp,%rbp => 0x00000000004004a0 <+4>: jmp 0x4004a0 <main+4> 0x00000000004004a2 <+6>: pop %rbp 0x00000000004004a3 <+7>: retq End of assembler dump. (gdb) si 6 asm ("here: jmp here"); (gdb) set debug infrun 1 (gdb) si infrun: clear_proceed_status_thread (process 29432) infrun: proceed (addr=0xffffffffffffffff, signal=GDB_SIGNAL_DEFAULT, step=1) infrun: resume (step=1, signal=GDB_SIGNAL_0), trap_expected=0, current thread [process 29432] at 0x4004a0 next pc: 0x4004a0 ==> 0x4004a0 infrun: inserting single-step breakpoint at 0x4004a0 infrun: wait_for_inferior () infrun: target_wait (-1, status) = infrun: 29432 [process 29432], infrun: status->kind = stopped, signal = GDB_SIGNAL_TRAP infrun: infwait_normal_state infrun: TARGET_WAITKIND_STOPPED infrun: stop_pc = 0x4004a0 infrun: stepi/nexti infrun: stop_stepping 6 asm ("here: jmp here"); (gdb)
On 19 November 2013 15:06, Richard Earnshaw <rearnsha@arm.com> wrote: > On 08/11/13 16:34, Will Newton wrote: >> >> Since as far back as the beginning of the sourceware repository >> the ARM port has printed an error "Infinite loop detected" when >> the next_pc calculated is the same as the current one, for example >> when encountering a branch to the current PC address. >> >> This causes the test gdb.base/random-signal.exp as the error message >> is not expected. I have not been able to find a good reason for the >> message to be here so remove it and let the test pass. >> >> gdb/ChangeLog: >> >> 2013-11-08 Will Newton <will.newton@linaro.org> >> >> * arm-tdep.c (arm_get_next_pc): Remove "Infinite loop detected" >> error message. > > I guess the reasoning for this is that you can't single-step a > 'branch-to-self' instruction, since the next instruction (where the new > breakpoint has to go) is the same as the current instruction. > > However, it ought to be the higher levels of GDB that handle this, not > some random low-level function like get-next-pc. Thanks, applied. Single stepping an infinite loop works before and after this patch, the patch just prevents an error message getting printed on every step.
diff --git a/gdb/arm-tdep.c b/gdb/arm-tdep.c index 1ed21ea..035894c 100644 --- a/gdb/arm-tdep.c +++ b/gdb/arm-tdep.c @@ -4970,17 +4970,9 @@ arm_get_next_pc (struct frame_info *frame, CORE_ADDR pc) CORE_ADDR nextpc; if (arm_frame_is_thumb (frame)) - { - nextpc = thumb_get_next_pc_raw (frame, pc); - if (nextpc == MAKE_THUMB_ADDR (pc)) - error (_("Infinite loop detected")); - } + nextpc = thumb_get_next_pc_raw (frame, pc); else - { - nextpc = arm_get_next_pc_raw (frame, pc); - if (nextpc == pc) - error (_("Infinite loop detected")); - } + nextpc = arm_get_next_pc_raw (frame, pc); return nextpc; }