App loading with more features

[Ericson2314]

1. Relocations are performed, obviating the need for manual pic sprinkling.

2. Globals with initial values (vs all-0 globals in BSS) are now initialized to those values.

Contains #68, though if that PR is not wanted this is easy to undo.

[yhql]

Thanks for this landmark contribution!
I did merge #68.
Wondering about the use of link_wrap.sh, which would not work on Windows/Mac (might be a detail but it is nice to not have to worry about this anymore). This could be integrated into cargo-ledger for platform indepence, what do you think?
(I suspect you don't use cargo-ledger though 😄)

[Ericson2314]

We actually do use cargo-ledger! Yeah I suppose it could be made part of that, good idea.

[yhql]

Ok, then this is a 'nice' option for cross-platform building, but would be really nice if that wasn't even necessary. I recall you mentioned trying to have a section containing relocations directly in the linker script does not work @Ericson2314 , because then it's also considered something to relocate and trouble ensues. Do you have a link to anything that mentions this problem? Even if it points into LLVM's source code, I'm ready 😄

[dfordivam]

The problem with doing the "link-wrap" phase in the cargo ledger is that the cargo run and cargo test commands will not be able to work. And I am not sure invoking cargo-ledger from cargo test would be a good idea.

[yhql]

Yes agreed! So I just need to find a way to execute those commands in Windows properly, but having a linker field that can be different depending on the host does not appear to be possible.

Your PR is great, I'm just unclear on why we can't have the .rel.* sections in their own sections. I didn't find any information that would confirm that this is the intended behaviour, or a blindspot of lld.

[yogh333]

See #128