Doc does not describe what's currently implemented in the wild

[haberman]

I was writing a .wasm parser today, and I noticed that the "Hello, World" .wasm file I got from https://webassembly.studio/ includes debug info sections. Here's a list of all sections I found in the file:

     VM SIZE                       FILE SIZE
 --------------                 --------------
   NAN%       0 .debug_info      13.6Ki  30.0%
   NAN%       0 .debug_str       9.38Ki  20.8%
   NAN%       0 .debug_line      7.60Ki  16.8%
   NAN%       0 .debug_abbrev    4.14Ki   9.2%
   NAN%       0 Code             3.86Ki   8.6%
   NAN%       0 .debug_pubtypes  3.10Ki   6.9%
   NAN%       0 Data             1.19Ki   2.6%
   NAN%       0 .debug_pubnames  1.05Ki   2.3%
   NAN%       0 .debug_ranges       353   0.8%
   NAN%       0 name                310   0.7%
   NAN%       0 .debug_loc          276   0.6%
   NAN%       0 Import              102   0.2%
   NAN%       0 Type                 73   0.2%
   NAN%       0 Export               57   0.1%
   NAN%       0 .debug_macinfo       38   0.1%
   NAN%       0 Function             23   0.0%
   NAN%       0 Global               23   0.0%
   NAN%       0 Element              12   0.0%
   NAN%       0 [WASM Header]         8   0.0%
   NAN%       0 Table                 7   0.0%
   NAN%       0 Memory                5   0.0%
 100.0%       0 TOTAL            45.1Ki 100.0%

I found this repo and the doc that describes wasm debugging capabilities. But it doesn't say anything about what is currently implemented. The .debug_* sections above seem to be named like DWARF, but are they actually DWARF? What semantics are implemented?

[fitzgen]

Thanks for bringing this up.

Yes, they are DWARF. Their inclusion is due to a pretty recent change in LLVM trunk. I don't know the exact semantics and there is no standard for how DWARF applies to wasm yet. The wasm WG is setting up a sub-charter for debugging, and I expect that group to provide some guidance once it is up and running.