Add support for ARM exception handler ABI

[LeadroyaL]

Dwarf and ehabi are two different unwind model. ARM ELF file use .ARM.exidx and .ARM.extab secions to unwind stack trace. The description of data can be shown with command binutils readelf -u filename. Pyelftools doesn't parse these two sections, which is reason for this pull request.

Like DWARFInfo, EHABIInfo is also independent with ELFFile. It just need initialized by a SHT_ARM_EXIDX section.

Also I add argument -au, --arm_unwind in readelf.py. I cannot use -u because this pull request only compatible for ARM.

Referense:
https://github.com/llvm/llvm-project/blob/master/llvm/tools/llvm-readobj/ARMEHABIPrinter.h
https://developer.arm.com/documentation/ihi0038/b/
binutils: readelf.c

[eliben]

Thank you for the contribution. I'm not much familiar with ARM and its EHABI. CCing some folks who contributed ARM stuff before who could help review this

@frederiksdun @nickdesaulniers

[nickdesaulniers]

I found the inline references to the source code to LLVM's decoder helpful.

Should the tests have the source file that produced them, as well as comments on how to rebuild the test object files?

Looking at some of the bitwise operations, I'm curious about the perennial question for bi-endian architectures; does this work for big endian (compile with -mbig-endian). If yes; please add little and big endian binaries. If not, well I guess little endian is more standard, and something is better than nothing.

@smithp35 is the expert here.

[LeadroyaL]

Should the tests have the source file that produced them, as well as comments on how to rebuild the test object files?

You're right. I'll try to add the testcase from LLVM-readelf and produce them later.

Looking at some of the bitwise operations, I'm curious about the perennial question for bi-endian architectures; does this work for big endian (compile with -mbig-endian). If yes; please add little and big endian binaries. If not, well I guess little endian is more standard, and something is better than nothing.

Sorry I forget to handle endian, I'll fix it later also.

[nickdesaulniers]

The implementation LGTM; just style nits that elftools/ehabi/decoder.py has way too many UpperCamelCased local variable identifiers. I think those should only be used for classes. (Github highlights them in orange). Some methods also use _UpperCamelCase, and I think those should be _snake_case. PEP8 says:

Function names should be lowercase, with words separated by underscores as necessary to improve readability.

[nickdesaulniers]

cool LGTM; thanks for all of the work that went into this.

@eliben I think this is good to merge.

[eliben]

Just a note that I'll be offline for a week or so, so this gives you plenty of time to address the (mostly minor) comments. I'll re-review and merge when I'm back

[eliben]

Thanks!