Josh Stone

A static safe variant sounds fine to me, or a scope-like variant with a closure to continue in the current thread. (Like your suggested workaround, but the closure doesn't even...

@oleid, the resulting file descriptor will probably be 3 indeed, following 0=stdin, 1=stdout, 2=stderr. But as a fuse author you don't get to choose the file descriptor -- this is...

OK, I think it's dd's `O_TRUNC` causing the issue, which you can try avoiding with `dd conv=notrunc`. When opening with truncation, fuse will also call your `setattr` with size 0,...

It seems reasonable, but beware that this makes `crossbeam` a **public** dependency due to that exposed `Scope`. (i.e. updating to a new semver of crossbeam would be a breaking change.)...

One way to minimize the dependency is to re-export those necessary items (`scope` and `Scope`), so `fuse` users don't have to directly depend on `crossbeam` themselves. Or they can use...

I think re-exports are a good idea, but I'll leave it to @zargony to judge the public dependency. This will also need a semver bump for the change to existing...

> I'm a bit unsure what happens with the public BackgroundSession struct, which has a crossbeam ScopedJoinHandle member. Will that be completely encapsulated from clients of fuse? That will be...

Oh wait, I didn't realize that the fields of `BackgroundSession` are `pub`. In that case it does become another part of the public dependency. But I question whether the field...

All of those failures have test data of manually written bytes, in little-endian. They'll either need to be manually duplicated for big-endian, or figure out the appropriate byte-swaps in each...

> or maybe even just `rayon`. cc @cuviper thoughts? How could I say no? 😉 But yes, indeed, those look like prime candidates for parallel iterators. If you wanted to...