Use pyelftools for Realtek post-build script

[theotherjimmy]

Description

The online compiler can be very picky about what is and is not allowed in a post-build scirpt. For a while now, the Realtek post-build script has been preventing anyone from successfully compiling for the RTL8195AM in the online compiler. This refactor changes the code that was failing to be much simpler, and it works in the website. I have tested this on a staging site.

For reference, the traceback:

  File "/usr/src/mbed-sdk/tools/toolchains/__init__.py", line 1152, in link_program
    self.binary(r, elf, bin)
  File "/usr/src/mbed-sdk/tools/hooks.py", line 54, in wrapper
    post_res = tooldesc["post"](t_self, *args, **kwargs)
  File "/usr/src/mbed-sdk/tools/targets/__init__.py", line 538, in binary_hook
    rtl8195a_elf2bin(t_self, elf, binf)
  File "/usr/src/mbed-sdk/tools/targets/REALTEK_RTL8195AM.py", line 283, in rtl8195a_elf2bin
    segment = parse_load_segment(t_self.name, image_elf)
  File "/usr/src/mbed-sdk/tools/targets/REALTEK_RTL8195AM.py", line 212, in parse_load_segment
    return parse_load_segment_armcc(image_elf)
  File "/usr/src/mbed-sdk/tools/targets/REALTEK_RTL8195AM.py", line 143, in parse_load_segment_armcc
    for line in subprocess.check_output(cmd, shell=True, universal_newlines=True).split("\n"):
  File "/usr/lib/python2.7/subprocess.py", line 574, in check_output
    raise CalledProcessError(retcode, cmd, output=output)
CalledProcessError: Command '"/opt/armcc5_06_u3/bin/fromelf" --text -v --only=none /tmp/chroots/ch-e3a8e63e-cb37-4195-ae50-1fe69442b3e4/build/mbed-os-example-fat-filesystem.REALTEK_RTL8195AM.elf' returned non-zero exit status 127

Error 127 is "command not found" https://stackoverflow.com/questions/1763156/127-return-code-from#1763178. I fixed this by refactoring so that we don't use any commands.

Pull request type

[X] Fix
[ ] Refactor
[ ] New target
[ ] Feature
[ ] Breaking change

[theotherjimmy]

@samchuarm @ARMmbed/team-realtek

[samchuarm]

Hi @ARMmbed/team-realtek , please review the changes and provide your feedback if any.

[tung7970]

Thanks for the effort. It looks much nicer.

One small thing, shouldn't the 'attr' variable be better named 'addr' in function _parse_load_segment_inner ?

[0xc0170]

+1 simplification, elftools 💯

[theotherjimmy]

@tung7970 Yes, that's a typo. I'll fix that now.

[theotherjimmy]

@0xc0170 @tung7970 I revised the commit to correct the typo. Could you both take another look?

[tung7970]

LGTM

[cmonr]

Interesting. Why does the code now no longer need to check compilers?

/morph build

[tung7970]

The new script checks the elf file directly rather than output from each compiler’s elf dumper.

[mbed-ci]

Build : SUCCESS

Build number : 1766
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/6621/

Triggering tests

/morph test
/morph uvisor-test
/morph export-build
/morph mbed2-build

[mbed-ci]

Exporter Build : SUCCESS

Build number : 1403
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/6621/

[mbed-ci]

Test : SUCCESS

Build number : 1575
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/6621/1575

[adbridge]

@theotherjimmy Is this fixing #5976 ??

[theotherjimmy]

Yes, but I want to verify that once this goes live on the website, so I omitted a "Resolves" directive.