[PATCH] D70720: [llvm-objdump] Display locations of variables alongside disassembly

[Oliver Stannard (Linaro) via Phabricator via llvm-commits]

ostannard marked 6 inline comments as done.
ostannard added inline comments.


================
Comment at: llvm/lib/Support/FormattedStream.cpp:45
+
     ++Column;
+
----------------
MaskRay wrote:
> This is incorrect due to East Asian Width.... http://www.unicode.org/reports/tr11/tr11-36.html
> 
> ```lang=cpp
> % cat a.cc
> int main() {
>   int 喵🙂 = 0;
>   int 喵喵🙂 = 15;
>   int 喵🙂喵🙂 = 24;
>   喵🙂 = 28;
>   喵喵🙂 = 32;
> }
> % clang -c a.cc
> % llvm-objdump -S --no-show-raw-insn --debug-vars a.o
> 
> a.o:    file format elf64-x86-64
> 
> 
> Disassembly of section .text:
> 
> 0000000000000000 main:
> ; int main() {                                          ┠─ 喵🙂 = <unknown op DW_OP_fbreg (145)>
>                                                         ┃ ┠─ 喵喵🙂 = <unknown op DW_OP_fbreg (145)> 
>                                                         ┃ ┃ ┠─ 喵🙂喵🙂 = <unknown op DW_OP_fbreg (145)>
>        0:       pushq   %rbp                            ┃ ┃ ┃     
>        1:       movq    %rsp, %rbp                      ┃ ┃ ┃     
>        4:       xorl    %eax, %eax                      ┃ ┃ ┃     
> ;   int 喵🙂 = 0;                                         ┃ ┃ ┃   
>        6:       movl    $0, -4(%rbp)                    ┃ ┃ ┃     
> ;   int 喵喵🙂 = 15;                                       ┃ ┃ ┃   
>        d:       movl    $15, -8(%rbp)                   ┃ ┃ ┃     
> ;   int 喵🙂喵🙂 = 24;                                      ┃ ┃ ┃   
>       14:       movl    $24, -12(%rbp)                  ┃ ┃ ┃     
> ;   喵🙂 = 28;                                            ┃ ┃ ┃   
>       1b:       movl    $28, -4(%rbp)                   ┃ ┃ ┃     
> ;   喵喵🙂 = 32;                                           ┃ ┃ ┃   
>       22:       movl    $32, -8(%rbp)                   ┃ ┃ ┃     
> ; }                                                     ┃ ┃ ┃   
>       29:       popq    %rbp                            ┃ ┃ ┃     
>       2a:       retq                                    ┻ ┻ ┻
> ```
> 
There's some code in the Support library which does this, which also simplifies this function a bit. It handles east asian characters correctly (at least the ones in your example), but doesn't get the correct widths for emojis. Clang seems to handle emojis by printing a hex value which it knows the length of:

  $ /work/llvm/build/bin/clang -fsyntax-only -c width.c 
  width.c:1:15: error: use of undeclared identifier 'a'
  int 喵<U+1F642> = a;
                    ^
  1 error generated.

Fixing that is going to be a lot of work, and isn't related to this patch, so I'll just make it work for characters supported by the existing unicode support library for now.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70720/new/

https://reviews.llvm.org/D70720