Mutable variables read shouldn't be substituted #7683
Labels
Comments
|
Comment author: @gasche Given that the order of evaluation of arguments is unspecified, I have the impression that both reported behaviors (printing 1 or 2) are correct with respect to the language specification. Do you have an example where the observed behavior is actually a bug? |
|
Comment author: @gasche Are you replying this because there is no example that is actually a bug (I would find it surprising, and an interesting information), or because you would rather avoid doing the work of thinking of an example that is actually a bug? |
|
Comment author: @nojb I tried to find an example that is a bug but so far without success. |
|
Comment author: @mshinwell I have posted an example of a bug which I think is due to this issue on #1499. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Original bug ID: 7683
Reporter: @chambart
Status: confirmed (set by @mshinwell on 2017-12-06T17:08:41Z)
Resolution: open
Priority: normal
Severity: minor
Version: 4.07.0+dev/beta2/rc1/rc2
Category: middle end (typedtree to clambda)
Monitored by: @nojb @gasche @alainfrisch
Bug description
In closure.ml the Uvar constructor is supposed to be pure by 'is_simple_argument', but when the variable is mutable this
can be wrong.
Steps to reproduce
This program exhibit the problem, it returns 1 in bytecode and flambda and 2 with closure.
let f () n () =
n
let g () =
let r = ref 0 in
f (incr r) !r (incr r)
let () = print_int (g ())
The text was updated successfully, but these errors were encountered: