make DEFAULT_HANDLER an undefined symbol

[japaric]

fixes #58

I found out that this fixes #58 while writing the embedonomicon

[japaric]

@crawford could you give this a spin? I tested with the example code you posted in #58 but perhaps you should try this out with some more complex program.

cc @pftbest you may want to include this in msp430-rt

[pftbest]

@japaric Sorry, but looks like this fix doesn't work for me. Are you sure it fixed the issue on your side? I still see that EXCEPTIONS array is wrong both in debug and non-lto release.

[crawford]

@japaric Gah! I didn't see your note since GitHub doesn't always send me email alerts... I tried this out on my application (rustc 1.27.0-nightly-7360d6dd6 with LLD) and it seems to work. It looks like @pftbest's patch might be more complete though.

[pftbest]

lld may not have this issue, it is probably GNU ld specific.

I've created a bug report for binutils, will see what they say.

[crawford]

@pftbest Uh oh. I'm not seeing the original issue anymore (even with GNU ld).

[pftbest]

@crawford In your bug report you have posted an example project called "simple". Just now I took it's source code, modified it to work with cargo on a recent nightly (updated cortex-m-rt to version 4 and removed Xargo.toml) and then changed the linker script to be the same as in this PR.

Here is what i got: updated.tar.gz

Then I build it and look at the output

cargo build
arm-none-eabi-objdump -d -j .text -j .vector_table target/thumbv7m-none-eabi/debug/default-test

And I see that the output is incorrect:

00000008 <EXCEPTIONS>:
   8:	000004df 000004df 000004df 000004df     ................
  18:	000004df 00000000 00000000 00000000     ................
  28:	00000000 000004df 000004df 00000000     ................
  38:	000004df 000004df                       ........

...

000004de <BUS_FAULT>:
 4de:	f3ef 8008 	mrs	r0, MSP
 4e2:	f7ff bff5 	b.w	4d0 <_ZN11cortex_m_rt15default_handler17h73065f5c8934a2c2E>
 4e6:	defe      	udf	#254	; 0xfe

Can you please build this project on your side to compare?

[crawford]

@pftbest I'm seeing the same behavior as you for the example code, but not for my application. My application seems to link properly. I wonder if it is because I don't have a direct dependency on cortex-m-rt.

[japaric]

Subsumed by #69