[llvm] r237327 - Don't omit the constant when computing a cross-section relative relocation.

[Andy Ayers]

Ok, will do.

From: Rafael Espíndola [mailto:rafael.espindola at gmail.com]
Sent: Tuesday, May 26, 2015 10:33 AM
To: Andy Ayers
Cc: llvm-commits
Subject: RE: [llvm] r237327 - Don't omit the constant when computing a cross-section relative relocation.


You should then split the test in 2. One uses llc and checks the generated assembly for the desired expression.

The other uses llvm-mc to test that a given expression produces the desired relocation.

Please *never* use llc with binary output in a test unless you are testing a code path that really can't be reached otherwise.
On May 26, 2015 12:31 PM, "Andy Ayers" <andya at microsoft.com<mailto:andya at microsoft.com>> wrote:
We want to ensure that the bitcode expressions are properly simplified so that when assembled they produce the expected relocations.

-----Original Message-----
From: Rafael Espíndola [mailto:rafael.espindola at gmail.com<mailto:rafael.espindola at gmail.com>]
Sent: Monday, May 25, 2015 5:26 AM
To: Andy Ayers
Cc: llvm-commits
Subject: Re: [llvm] r237327 - Don't omit the constant when computing a cross-section relative relocation.

> +; Verify the assembler produces the expected expressions ; RUN: llc
> +-mtriple=x86_64-pc-win32 %s -o - | FileCheck %s ; Verify the .fix
> +data section conveys the right offsets and the right relocations ;
> +RUN: llc -mtriple=x86_64-pc-win32 -filetype=obj %s -o - |
> +llvm-readobj -relocations -expand-relocs -sections -section-data |
> +FileCheck %s --check-prefix=READOBJ

Can't this be tested with llvm-mc instead of llc?

Cheers,
Rafael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150526/b7f94f37/attachment.html>