We have no such testing. This makes impossible
to add support for new ELFOSABI_* tags.
Details
Diff Detail
Event Timeline
This is fine, but I wonder what you think about testing using od, to avoid things being circular here? Presumably you will use yaml2obj to test llvm-readobj prints the right value for the field, which means that theoretically, both could be broken in the same way and we wouldn't notice.
So I dug around, and there was once-upon-a-time testing for this, but you deleted it in rL367988... It wasn't as comprehensive as this though.
Well. But we know the correct enum values from start, don't we? For example we check OS/ABI: CloudABI (0x11), we know that enum value of ELFOSABI_CLOUDABI is 0x11,
and that llvm-readelf dumps the same. So it can't be broken I think.
Ok, I see a possbile scenario now. Like yaml2obj writes X instead of Y and then readelf reads X and shows Y. But that way we should use od everywhere and verify od with other tools too, because od might be broken in the same way too :) It is too much probably?