Currently in draft at the moment, reasons:

1. Need to migrate various altered symbol names into single xref usage. (will be done in separate commit for readability)
   • This currently affect unnamed cross reference symbols from revisions.
2. At least ability to determine non-LTCG and LTCG xrefs need to remain separate for X reason.
3. Need to purge global cross reference strings list.

With full xreference cache system, we can reduce unnecessary rescan same symbol. Except, we do still have other optimization to do. I believe this pull request partially resolve #87 ticket.

With cross reference's symbols separated into their own library file. It will be easier to maintenance and organizing.

Pros:

1. Allow context's xref_database to allocate and cache for every linked symbols.
2. Reduce partial of redundant scan process system and stop re-report for same address.

Cons:

1. Add new symbol, aka group or derived usage, will require manual add into specific library's def file.
   • Compiler will stop building if cross reference symbol doesn't exist. This helps enforce contributor to verify their work. Then add new symbol if it's actually a new symbol.
2. Any symbol rename will also need to rename inside .def file. Using IDE's rename process may will detect other relative same name. Such as in the comment section.

Hmm, I thought I had mentioned before. You can use "extend oovpa test to verify symbol's reference against revisions" commit for verification of internal test is working for compare revisions' xref vs symbol group's xref difference.