What I'm trying to do is to simplify processing lld output by automated parsing tool.
Ideally I'd like to have something like this for every warning or error message:
<path_to_ld.lld> : error|warning : <location> : <message>
This makes it easy to extract semantic values from error message and change its format if needed.
For instance MS Visual Studio expects error message in this format:
<location> : error : <message>
And one just needs to swap tokens to get the desired result.
I haven't updated unit tests yet - will do if this patch is a right thing.
I am not sure it makes sense to always print the output name. In here the issue has nothing to do with the files, it is just a broken -m option.
If you want a common format, maybe printing something like" <command line>:" in here would be better.