I noticed when testing 1brc that we have a problem reading arrays, sometimes, because of interference with the support for dynamically changing the type of a variable; there's a difference between compatible type changes, when we want to keep the field, and incompatible changes, when we want to throw it out.
I've put in a fix, but we should also add more tests and maybe something overall more robust.
I noticed when testing 1brc that we have a problem reading arrays, sometimes, because of interference with the support for dynamically changing the type of a variable; there's a difference between compatible type changes, when we want to keep the field, and incompatible changes, when we want to throw it out.
I've put in a fix, but we should also add more tests and maybe something overall more robust.