Luke Street
Luke Street
It turns out this specific failure is because of a tricky linker flag: `-code_merging all,aggressive` It causes issues with analysis and splitting, because `.ctors` entries were merged together, and I'll...
RELs with the same ID aren't _supposed_ to be loaded simultaneously. Mario Party similarly links RELs individually (which causes the duplicate IDs) so that they _can't_ reference each other. But...
This is a bug / unintentional behavior with decomp.me and should be submitted there. Relevant discussion: https://discord.com/channels/897066363951128586/897066363951128590/1249864215993389076
Resolved by https://github.com/decompme/decomp.me/pull/1403
You can adjust the alignment by specifying `align:x` in symbols.txt. Or is this a different issue?
Is this still an issue?
Implemented in #268
> Do you mind if i pick up parts of it (e.g. "remove snapshotting") separately? Go for it! I'll rebase accordingly once I get time to work on it again.
If handle_shim_call was always async, one Boxed future could be avoided. It’s only used to conditionally return a Future. Maybe it would work to make that function return `Option` instead?...
In 5c92c7d93f1320b08405fff7a92adbd94f440b1b I was able to simplify the async shim handling. It pushes the responsibility for storing and polling the future up into the top-level event loop (cli or web,...