[PATCH] D72973: using symbol index+symbol name + storage mapping class as label for llvm-objdump -D

[Hubert Tong via Phabricator via llvm-commits]

hubert.reinterpretcast added inline comments.


================
Comment at: llvm/tools/llvm-objdump/llvm-objdump.cpp:1383
                          SectionAddr + Start + VMAAdjustment);
-
-      outs() << SymbolName << ":\n";
+      if (Obj->isXCOFF()) {
+        printXCOFFSymbolDescription(Symbols[SI], SymbolName);
----------------
DiggerLin wrote:
> daltenty wrote:
> > DiggerLin wrote:
> > > daltenty wrote:
> > > > Shouldn't this behaviour be option gated under an XCOFF specific option? As is this is going to result in output incompatible with binutils objdump on AIX, which could break tools.
> > > when dissemble the xcoff object file, we change the label to symbol index+ symbol name + storage mapping class, it is not a new option. I think it is different with the  with binutils objdump on AIX.
> > The fact that it is different with the with binutils objdump on AIX is the issue. As the top of this file indicates: "The flags and output of this program should be near identical to those of binutils objdump."
> > 
> > We need to add an XCOFF specific option, something like "--expanded-symbol-description", to turn on this behavior change.
> @hubert.reinterpretcast , @jasonliu , what do you think about david's opinion ? I am prefer to keep current implement.
I think there may be practical issues if we don't follow the binutils behaviour. At the least, it means that having the richer output format by default creates yet another situation where assumptions that people can make on other platforms would not hold for AIX.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D72973/new/

https://reviews.llvm.org/D72973