pveber

These examples are not cyclic but certainly wrong since they imply the existence of two different objects with the same path. However I think this verification is more on of...

See the commit above, saying: we adopt the simple (and more common) design where paths which are cyclic ocaml values are _not_ supported and _not_ checked. In particular, some functions...

Comment from 1d80c25: "By merging this branch, we have implementations for [reify] and [fold] that can cope with cycles. The first one does so by counting on the file system...

Oddly, I don't. I tried to change the initialization of the RNG several times but I don't see any failing test. Some OSX-specific problem?

Maybe you could try to create it manually in utop and see what happens? As for the second test, could you ls the contents of the temp directory after `create_test_directory`...

I just noticed that there was two important commits I had left behind (mistakenly left on the branch for the fold function since they only deal with exists and mkdir)....

This is really twisted... But it gets worse. Look at this slight modification of the commands you describe: ``` $ make clean && make native (* change a.ml *) $...

I made a recursive diff on the whole directory between the state just after editing `a.ml` and after the second invocation, and the only relevant difference is `a.o`, which is...

Just as a memo: in `biocaml.unix`, there was a module using C stubs. The corresponding unit test program would not run saying it couldn't find `dllbiocaml_unix.so`. Sorry the description might...

Ok, got it. Your example is wrong, here's what you should be doing: ``` $ make clean && make native (* change a.ml *) $ cp lib/a.ml _build/lib/a.ml $ cd...