[RFC] Removal of 'deplibs'?

Hi all,

Did you know that there is a 'deplibs' keyword? Neither did I. It looks like it's not really used for much. We don't document it in the Language Reference. The only support I see if for reading/writing bitcode and manipulating modules. There's a bunch of code that could be removed if we could get rid of this.

Does anyone use this thing? Does anyone mind if I nuke it?

-bw

Apparently Pygments knows about it
<https://bitbucket.org/birkenfeld/pygments-main/src/24c16194941a7464a0cbe27896cc103b9bbab6a9/pygments/lexers/asm.py?at=default#cl-243>.
If you nuke this please send patches (or pull-requests, or whatever
they use in that world) to Pygments. Alternatively, if you really
can't be bothered, file a PR and assign it to me...

-- Sean Silva

Please nuke it. It is really old and was never used. If it comes back it should be done with module metadata.

-Chris

Nuking. Thanks!

Sean, I'll send a ping to the Pygments wonks.

-bw

Dears,

If you remove this keyword, will you keep the backward compatibility in LLVM 3.x?

We use this keyword in our project for checking dependent libraries, and we rely on the promise that bitcode format will keep compatible in LLVM 3.x. In our project, we compile and save bitcodes for later use, but we always try to keep our runtime upstream. Hence, we need readers recognize previous version bitcodes as well, at least in LLVM 3.x. However, the incompatibility happened once in LLVM 3.1, and we deeply hope it won’t happen again in the future 3.x version.

Many thanks.
Jason

I think it is perfectly reasonable to accept and ignore deplibs in old .bc files. Bill, can you take care of that?

-Chris

I added back support for reading the .bc files in r168779. Please let me know if it works for you.

-bw