[PATCH] D104644: [AIX][XCOFF] Support 64-bit relocation writing and related tests

[Maryam via Phabricator via llvm-commits]

MaryamBen marked 2 inline comments as done.
MaryamBen added inline comments.


================
Comment at: llvm/test/CodeGen/PowerPC/aix-xcoff-toc.ll:14
+; RUN:     --xcoff-traceback-table=false -filetype=obj -o %t.o < %s
+; RUN: llvm-readobj --syms %t.o | FileCheck --check-prefixes SYM,SYM64 %s
 
----------------
jhenderson wrote:
> MaryamBen wrote:
> > jhenderson wrote:
> > > You may wish to add --match-full-lines to here and the equvialent 32-bit FileCheck command. There's the potential for a false pass in lines like the following:
> > > 
> > > ```
> > > ; SYM64:         SymbolAlignmentLog2: 3
> > > ```
> > > That line could match any of the following output:
> > > ```
> > > SymbolAlignmentLog2: 3
> > > SymbolAlignmentLog2: 32
> > > SymbolAlignmentLog2: 32131241251
> > > ```
> > > 
> > > etc.
> > In some cases the abstraction is used to combine the 32/64 bit.
> > 
> > If you think it's better, I will add it whenever it's possible. 
> > In some cases the abstraction is used to combine the 32/64 bit.
> 
> I'm not sure I follow how that impacts the comment I made, could you clarify? There's nothing wrong with the use of the prefix.
> 
> I'm suggesting your FileCheck command becomes: `FileCheck --check-prefixes SYM,SYM64 %s --match-full-lines` and `FileCheck --check-prefixes SYM,SYM32 %s` similarly.
For symbols there is no problem.

I thought you wanted me to add --match-full-line to every FileCheck which is not possible in the following cases:

1-When we check just the beginning: 
; CHECK: .comm globa[RW],4,2
the real value is .comm globa[RW],4,2 # @globa

2- Avoid writing address in 64 bit :
; DIS: 00000000 <.text>:
instead of 0000000000000000 <.text>:





CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D104644/new/

https://reviews.llvm.org/D104644