[PATCH] D30699: [ELF] - Stop producing broken output for R_386_GOT32X relocation.

Mon Mar 20 00:53:48 PDT 2017

grimar added a comment.

In https://reviews.llvm.org/D30699#704135, @ruiu wrote:

> You are talking based on https://github.com/hjl-tools/x86-psABI/wiki/intel386-psABI-draft.pdf, right? There is a table of relocation types on p.36. Please open that page and see R_386_GOT32 and R_386_GOT32X. The calculation expressions for the two relocations are the same, including the footnote saying that "[a]pplied on memory operand without base register when position-independent code is disabled." So, they are the same, unless you do optional optimization for GOT32X.

Hmm no, I was talking basing on latest released ABI (The Intel386 psABI version 1.1, https://github.com/hjl-tools/x86-psABI/wiki/intel386-psABI-1.1.pdf). On p.36 calculation expression is different there. 
I did not notice that draft before, it clarifies the situation.

But then question is: do you think we should implement draft version of ABI ? It is most probably contains change that will be released I guess, but it still did not yet.

That I think means we still can go with R_386_GOT32X first, because it is already released.

https://reviews.llvm.org/D30699