[PATCH] D30812: AsmPrinter: Don't treat symbols with prefix data as code

[Moritz Angermann via Phabricator via llvm-commits]

angerman added a comment.

I will try to see if I run into this again. The case I had yesterday ended up being a botched relocation that did an invalid sign extend.

However the description is, what I believe to be the underlying reason for the function/object mangling in ghc. I am uncertain if it's
still valid though. Your description certainly makes sense, and I'm not certain that the described scenario could happen only if the
PLT is used via linking. What you said makes me believe though that if the symbol's address is used for a relative offset that this
would have to go through the GOT in any case.

I'll revise this issue once I'll have a proper test case at hand.


https://reviews.llvm.org/D30812