Section 7.2. (mis)Optimization : Can not play the debug session as described in the tutorial

[the-cat-that-codes]

Hi everyone,

I have some troubles at running this section correctly.

First, here are my version of rustup and rustc :

$ rustup --version
rustup 1.23.1 (3df2264a9 2020-11-30)
info: This is the version for the rustup toolchain manager, not the rustc compiler.
info: The currently active `rustc` version is `rustc 1.53.0-nightly (07e0e2ec2 2021-03-24)`

The problem is that I can not execute step by step (using the gdb next instruction) the binary launched with cargo run --release. Here after is my session :

$ cargo run --release
    Finished release [optimized + debuginfo] target(s) in 0.02s
     Running `arm-none-eabi-gdb -q -x openocd.gdb '/home/XXX/Dev/Rust/Embedded_Discovery/discovery-master/target/thumbv7em-none-eabihf/release/registers'`
Reading symbols from /home/XXX/Dev/Rust/Embedded_Discovery/discovery-master/target/thumbv7em-none-eabihf/release/registers...
0x08000252 in registers::__cortex_m_rt_main () at /rustc/07e0e2ec268c140e607e1ac7f49f145612d0f597/library/core/src/ptr/mod.rs:1150
1150	/rustc/07e0e2ec268c140e607e1ac7f49f145612d0f597/library/core/src/ptr/mod.rs: Aucun fichier ou dossier de ce type.
Loading section .vector_table, size 0x194 lma 0x8000000
Loading section .text, size 0x1220 lma 0x8000194
Loading section .rodata, size 0x1b4 lma 0x80013b4
Start address 0x8000194, load size 5480
Transfer rate: 13 KB/sec, 1826 bytes/write.
Breakpoint 1 at 0x8000224: file src/07-registers/src/main.rs, line 9.
Breakpoint 2 at 0x8001078
Breakpoint 3 at 0x80013b2
Note: automatically using hardware breakpoints for read-only addresses.

Breakpoint 1, main () at src/07-registers/src/main.rs:9
9	#[entry]
(gdb) s
halted: PC: 0x0800022a
registers::__cortex_m_rt_main () at src/07-registers/src/main.rs:11
11	    aux7::init();
(gdb) n
halted: PC: 0x08000254
halted: PC: 0x08000236
halted: PC: 0x0800023a
halted: PC: 0x0800023e
18	        ptr::write_volatile(GPIOE_BSRR as *mut u32, 1 << 9);
(gdb) n
halted: PC: 0x08000240
halted: PC: 0x08000244
halted: PC: 0x08000246
halted: PC: 0x0800024a
halted: PC: 0x0800024c
halted: PC: 0x08000250
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
halted: PC: 0x08000252
^Chalted: PC: 0x08000252
halted: PC: 0x08000252

Program received signal SIGINT, Interrupt.
0x08000252 in registers::__cortex_m_rt_main () at /rustc/07e0e2ec268c140e607e1ac7f49f145612d0f597/library/core/src/ptr/mod.rs:1150
1150	/rustc/07e0e2ec268c140e607e1ac7f49f145612d0f597/library/core/src/ptr/mod.rs: Aucun fichier ou dossier de ce type.

after the second nextinstruction, I enter in the loop rather to go to the next write_volatile ...

If I disassemble, I get :

Loading section .vector_table, size 0x194 lma 0x8000000
Loading section .text, size 0x1220 lma 0x8000194
Loading section .rodata, size 0x1b4 lma 0x80013b4
Start address 0x8000194, load size 5480
Transfer rate: 13 KB/sec, 1826 bytes/write.
Breakpoint 1 at 0x8000224: file src/07-registers/src/main.rs, line 9.
Breakpoint 2 at 0x8001078
Breakpoint 3 at 0x80013b2
Note: automatically using hardware breakpoints for read-only addresses.

Breakpoint 1, main () at src/07-registers/src/main.rs:9
9	#[entry]
(gdb) s
halted: PC: 0x0800022a
registers::__cortex_m_rt_main () at src/07-registers/src/main.rs:11
11	    aux7::init();
(gdb) n
halted: PC: 0x08000254
halted: PC: 0x08000236
halted: PC: 0x0800023a
halted: PC: 0x0800023e
18	        ptr::write_volatile(GPIOE_BSRR as *mut u32, 1 << 9);
(gdb) disassemble /m
Dump of assembler code for function registers::__cortex_m_rt_main:
10	fn main() -> ! {
   0x0800022a <+0>:	push	{r7, lr}
   0x0800022c <+2>:	mov	r7, sp

11	    aux7::init();
   0x0800022e <+4>:	bl	0x8000254 <aux7::init>
   0x08000232 <+8>:	movw	r0, #4120	; 0x1018
   0x08000236 <+12>:	mov.w	r1, #512	; 0x200
   0x0800023a <+16>:	movt	r0, #18432	; 0x4800
=> 0x0800023e <+20>:	str	r1, [r0, #0]
   0x08000240 <+22>:	mov.w	r1, #2048	; 0x800
   0x08000244 <+26>:	str	r1, [r0, #0]
   0x08000246 <+28>:	mov.w	r1, #33554432	; 0x2000000
   0x0800024a <+32>:	str	r1, [r0, #0]
   0x0800024c <+34>:	mov.w	r1, #134217728	; 0x8000000
   0x08000250 <+38>:	str	r1, [r0, #0]
   0x08000252 <+40>:	b.n	0x8000252 <registers::__cortex_m_rt_main+40>

End of assembler dump.

I have noticed that the file mod.rs is not found. Could it be the problem ?
I have searched also in the directory that we can see in the tutorial (~/.rustup/toolchains/stable-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/ptr/mod.rs) but it is not. May anyone explain me what I have missed ?

Many thanks for your responses :-)

[pcardune]

I've run into the same problem.

[benschneider]

I see the individual string movements are there, try using stepi (or si) to step by machine instruction rather than line. Lights should then individually switch on and off.

[NitinSaxenait]

I am also working on the same chapter and facing the same issue in mis(optimization). Nothing comes after this
Breakpoint 1, main () at src/07-registers/src/main.rs:7 7 #[entry] (gdb) s registers::__cortex_m_rt_main () at src/07-registers/src/main.rs:9 9 aux7::init(); (gdb) n 16 ptr::write_volatile((GPIOE_BSRR as *mut u32), (1<<9));

[dodsonmg]

I think this is an artefact of optimisation, which can create a bit of a discontinuity between your source and the assembly. The debugger attempts to step through your source, but the source no longer sufficiently corresponds with the assembly and it gets a bit lost (this is the extent to which I understand how debugging and debugging symbols work... it's not a very deep understanding!). It still executes the program.

The simple solution, as @benschneider points out, is to just step through the assembly one instruction at a time with nexti instead of stepping through the source with next. You can view the assembly in the debugger as it's executed with layout asm.

@japaric: I'm happy to submit a PR, but there appear to be several stale issues and PRs associated with this section already (#238, #245, and #237 are the obvious ones). Also, it looks like the most recent commit to discovery/src/07-registers/optimization.md (cdf6be) actually changed from stepping through the assembly to stepping through the source, so maybe we in this thread are simply doing something wrong...