First program linking issues

[yuri91]

Hi,

I am following the Copper book, and I am currently having problems with the very first program (http://japaric.github.io/copper/first/build.html).

The program compiles, but the objdump output is the following:

$arm-none-eabi-objdump -CD target/thumbv7m-none-eabi/debug/app

target/thumbv7m-none-eabi/debug/app:     file format elf32-littlearm


Disassembly of section .text:

00000000 <_reset-0x8>:
   0:	20010000 	andcs	r0, r1, r0
   4:	00000009 	andeq	r0, r0, r9

For some reason the main function is optimized away at some point.
If I make rustc emit the assembly file, I can see the function is there, with the right symbol (_reset), so I guess that the problem comes while linking.

[japaric]

@yuri91 are you using a mac? Does the gcc you are using come from the gcc-arm-embedded tap/cask? Someone reported (rust-embedded/discovery#34) that that gcc has some bug like the one you reported here. Could you try the gcc from the px4/px4 tap? (Sorry, I don't have the exact instructions to do that as I don't access to a mac)

[yuri91]

I am not using a mac, I am using Arch Linux. The version of arm-none-eabi-gcc that I had installed from the community repo was 6.3.
I looked the version currently shipped with Ubuntu, and it is 4.8 or 4.9.
So I tried the 4.9 myself, and It works.
Thank you very much for your help, I didn't think it could be an actual bug in the linker.

Also, I looked into the other two related issues (rust-embedded/discovery#34 and japaric/f3/issues/75), and I think that they have a 6.x version of GCC too.

[japaric]

I'm using Arch as well with gcc 6.3 but I can't repro your problem. I get:

$ arm-none-eabi-objdump -Cd target/thumbv7m-none-eabi/debug/bar

target/thumbv7m-none-eabi/debug/bar:     file format elf32-littlearm


Disassembly of section .text:

00000000 <_reset-0x8>:
   0:   20010000        .word   0x20010000
   4:   00000009        .word   0x00000009

00000008 <_reset>:
   8:   b083            sub     sp, #12
   a:   e7ff            b.n     c <_reset+0x4>
   c:   202a            movs    r0, #42 ; 0x2a
   e:   9001            str     r0, [sp, #4]
  10:   9000            str     r0, [sp, #0]
  12:   e7ff            b.n     14 <_reset+0xc>
  14:   e7fe            b.n     14 <_reset+0xc>

Does changing the loop in main to

    loop {
        unsafe {
            ::core::ptr::read_volatile(0x0 as *const u32);
        }
    }

makes the linker preserve the _reset symbol? With gcc 6.3 I mean. There's a LLVM bug when you have loops without side effects but that's supposed to only occur when optimizations (+LTO) are enabled. Which is not the case here but just to discard the possibility.

[yuri91]

Ok, my bad:

I was actually using yet another version of GCC (6.2.1) from the AUR.
Back when I installed that package (probably a year ago or so) there wasn't the one in [community] yet.
So the actual package in [community] with GCC 6.3 works, like the 4.9 one.

[japaric]

Good to know it's solved. I'll add a note to the book suggesting trying a different linker if you get a binary that's too different from what's in the book.

[RageYL]

Hi,

I had the same issue with gcc 4.9. It seems to come from the fact that the entry point has the same name as the variable _reset you define in the ld script. here is 2 ways to fix it:

• use KEEP(*(.text._reset))
  or a better alternative
• rename your entry point to _start or something else (both in the ld script and the main.rs) or rename the _reset variable in the ld script

Hope it helps

[tahnok]

I am also experiencing this issue on arch. I'm using

• arm-none-eabi-gcc from community ( 6.3.1 20170330)
• xargo 0.3.6
• rustc 1.18.0-nightly (7627e3d31 2017-04-16)

and compiling in 02-hardware with xargo build --target thumbv7em-none-eabihf --verbose which produces this binary: https://oolong.tahnok.me/cdn/app

I cloned this repo, so the code should be correct. Tried changing the entry point and changing the loop to that unsafe read_volatile, but same output.

[japaric]

The copper book is pretty outdated at this point and won't have time to update in the short term. You are better off starting with this blog post.