[RemoveDIs] Update syntax highlighting to include debug records

[SLTozer]

This patch updates the vim and vscode syntax highlighters to recognize debug records. I wasn't able to build the vscode extensions locally, so I've only tested this pattern by checking that the pattern works in vscode search.

[llvmbot]

@llvm/pr-subscribers-debuginfo

Author: Stephen Tozer (SLTozer)

Changes

This patch updates the vim and vscode syntax highlighters to recognize debug records. I wasn't able to build the vscode extensions locally, so I've only tested this pattern by checking that the pattern works in vscode search.

---

Full diff: https://github.com/llvm/llvm-project/pull/93660.diff

2 Files Affected:

• (modified) llvm/utils/vim/syntax/llvm.vim (+3)
• (modified) llvm/utils/vscode/llvm/syntaxes/ll.tmLanguage.yaml (+2-1)

diff --git a/llvm/utils/vim/syntax/llvm.vim b/llvm/utils/vim/syntax/llvm.vim
index d86e3d1ddbc27..c45391477f2c2 100644
--- a/llvm/utils/vim/syntax/llvm.vim
+++ b/llvm/utils/vim/syntax/llvm.vim
@@ -36,6 +36,9 @@ syn keyword llvmStatement sle slt srem store sub switch trunc udiv ueq uge ugt
 syn keyword llvmStatement uitofp ule ult umax umin une uno unreachable unwind
 syn keyword llvmStatement urem va_arg xchg xor zext
 
+" Debug records.
+syn match llvmStatement /\v#dbg_(assign|declare|label|value)/
+
 " Keywords.
 syn keyword llvmKeyword
       \ acq_rel
diff --git a/llvm/utils/vscode/llvm/syntaxes/ll.tmLanguage.yaml b/llvm/utils/vscode/llvm/syntaxes/ll.tmLanguage.yaml
index d80c3778bbe78..da07491c9d88d 100644
--- a/llvm/utils/vscode/llvm/syntaxes/ll.tmLanguage.yaml
+++ b/llvm/utils/vscode/llvm/syntaxes/ll.tmLanguage.yaml
@@ -34,7 +34,8 @@ patterns:
     captures:
       1:
         name: storage.type.llvm
-  - match: "\\badd\\b|\
+  - match: "(?<=\\s|^)#dbg_(assign|declare|label|value)\\b\
+            \\badd\\b|\
             \\baddrspacecast\\b|\
             \\balloca\\b|\
             \\band\\b|\

[OCHyams]

Is there someone else you can add to the review that has edited the file or reviewed a recent edit?

Rather than using "fancy" regex I would rather just have #dbg_assign #dbg_label etc checked separately. Less error prone, easier to read (for me!), but I won't push strongly for it.

Any chance we can get emacs too (https://github.com/llvm/llvm-project/blob/main/llvm/utils/emacs/llvm-mode.el) - I can look at adding later if you're not sure about emacs.

[OCHyams]

The other lines end |\ - I assume this one should too?

Can you explain what (?<=\\s|^) is doing please for the uninitiated?

[SLTozer]

The other lines end |\ - I assume this one should too?

Yes, it should - I'll fix that.

Can you explain what (?<=\s|^) is doing please for the uninitiated?

Look-behind - it lets us make sure that this is preceded by the start of the line or whitespace without including that whitespace in the match (so if you had a custom background for instructions, it wouldn't highlight the space behind the "#").

[OCHyams]

What's \v, and why do we need to wrap in /? (Not familiar with this stuff)

[felipepiovezan]

/ is just what you type in vim when you're about to start a search
\v is to control the "very magic" behaviour of regex spelling:

Vim's default 'magic' setting makes characters have the same meaning as in grep, and \v (very magic) makes them the same as the extended regular expressions used by egrep.

In practice, it means fewer escape sequences for the regex operators

[felipepiovezan]

vim changes LGTM

[jryans]

VS Code changes look good to me.

[OCHyams]

Tested the emacs change locally, that LGTM (not super familiar, but it works and looks reasonable).