Clojure

thread-bound? returns true (implying set! should succeed) even for non-binding thread

Details

  • Type: Defect Defect
  • Status: Open Open
  • Priority: Major Major
  • Resolution: Unresolved
  • Affects Version/s: None
  • Fix Version/s: None
  • Component/s: None
  • Labels:
    None
  • Patch:
    Code
  • Approval:
    Prescreened

Description

thread-bound? returns true for thread-local bindings that were not bound by the current thread. This is at odds with the docstring, which states that if true is returned, set! should succeed.

ere is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread  clojure.lang.Var.set (Var.java:230)
    user=> 

Approach: thread-bound? should return false in the case where there is a binding and that binding was not established by the current thread. The patch adds a new function to Var as core does not have visibility into the non-public Var$TBox.

Patch: thread-bound.diff

Screened by: Alex Miller

Activity

Hide
Paul Stadig added a comment - - edited

I have attached a patch that changes clojure.lang.Var and clojure.core/thread-bound? to only return true if a Var is set!-able.

Show
Paul Stadig added a comment - - edited I have attached a patch that changes clojure.lang.Var and clojure.core/thread-bound? to only return true if a Var is set!-able.
Paul Stadig made changes -
Field Original Value New Value
Attachment thread-bound.diff [ 11535 ]
Andy Fingerhut made changes -
Patch Code [ 10001 ]
Hide
Alex Miller added a comment -

REPL example?

Show
Alex Miller added a comment - REPL example?
Paul Stadig made changes -
Description thread-bound? returns true for a non-binding thread, this result (according to the docstring) implies that set! should succeed. However, thread-bound? does not check that any binding that might exist was created by the current thread, and calling set! fails with an exception when it is called from a non-binding thread, even though thread-bound? returns true.

thread-bound? should return false if there is a binding, and that binding was not established by the current thread.
thread-bound? returns true for a non-binding thread, this result (according to the docstring) implies that set! should succeed. However, thread-bound? does not check that any binding that might exist was created by the current thread, and calling set! fails with an exception when it is called from a non-binding thread, even though thread-bound? returns true.

thread-bound? should return false if there is a binding, and that binding was not established by the current thread.

Here is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread clojure.lang.Var.set (Var.java:230)
    user=>
Paul Stadig made changes -
Description thread-bound? returns true for a non-binding thread, this result (according to the docstring) implies that set! should succeed. However, thread-bound? does not check that any binding that might exist was created by the current thread, and calling set! fails with an exception when it is called from a non-binding thread, even though thread-bound? returns true.

thread-bound? should return false if there is a binding, and that binding was not established by the current thread.

Here is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread clojure.lang.Var.set (Var.java:230)
    user=>
thread-bound? returns true for a non-binding thread, this result (according to the docstring) implies that set! should succeed. However, thread-bound? does not check that any binding that might exist was created by the current thread, and calling set! fails with an exception when it is called from a non-binding thread, even though thread-bound? returns true.

thread-bound? should return false if there is a binding, and that binding was not established by the current thread.

Here is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

{code}
    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread clojure.lang.Var.set (Var.java:230)
    user=>
{/code}
Paul Stadig made changes -
Description thread-bound? returns true for a non-binding thread, this result (according to the docstring) implies that set! should succeed. However, thread-bound? does not check that any binding that might exist was created by the current thread, and calling set! fails with an exception when it is called from a non-binding thread, even though thread-bound? returns true.

thread-bound? should return false if there is a binding, and that binding was not established by the current thread.

Here is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

{code}
    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread clojure.lang.Var.set (Var.java:230)
    user=>
{/code}
thread-bound? returns true for a non-binding thread, this result (according to the docstring) implies that set! should succeed. However, thread-bound? does not check that any binding that might exist was created by the current thread, and calling set! fails with an exception when it is called from a non-binding thread, even though thread-bound? returns true.

thread-bound? should return false if there is a binding, and that binding was not established by the current thread.

Here is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

{noformat}
    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread clojure.lang.Var.set (Var.java:230)
    user=>
{noformat}
Hide
Joe Gallo added a comment -

Sure thing, Alex – here's a repl example I just ran this morning.

; nREPL 0.1.7
user> (def ^:dynamic *set-me* nil)
#'user/*set-me*
user> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
#'user/try-to-set
user> (try-to-set)
IllegalStateException Can't set!: *set-me* from non-binding thread  clojure.lang.Var.set (Var.java:230)
user>
Show
Joe Gallo added a comment - Sure thing, Alex – here's a repl example I just ran this morning.
; nREPL 0.1.7
user> (def ^:dynamic *set-me* nil)
#'user/*set-me*
user> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
#'user/try-to-set
user> (try-to-set)
IllegalStateException Can't set!: *set-me* from non-binding thread  clojure.lang.Var.set (Var.java:230)
user>
Stuart Halloway made changes -
Approval Triaged [ 10120 ]
Rich Hickey made changes -
Issue Type Defect [ 1 ] Enhancement [ 4 ]
Alex Miller made changes -
Issue Type Enhancement [ 4 ] Defect [ 1 ]
Alex Miller made changes -
Approval Triaged [ 10120 ] Prescreened [ 10220 ]
Description thread-bound? returns true for a non-binding thread, this result (according to the docstring) implies that set! should succeed. However, thread-bound? does not check that any binding that might exist was created by the current thread, and calling set! fails with an exception when it is called from a non-binding thread, even though thread-bound? returns true.

thread-bound? should return false if there is a binding, and that binding was not established by the current thread.

Here is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

{noformat}
    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread clojure.lang.Var.set (Var.java:230)
    user=>
{noformat}
{{thread-bound?}} returns true for thread-local bindings that were not bound by the current thread. This is at odds with the docstring, which states that if true is returned, {{set!}} should succeed.

ere is an example REPL session where a thread establishes a binding, those bindings are conveyed to a second thread, the second thread checks thread-bound? to see if it can set the binding, thread-bound? returns true indicating that the binding can be set, the second thread tries to set the binding, and the second thread gets an IllegalStateException:

{noformat}
    Clojure 1.5.1
    user=> (def ^:dynamic *set-me* nil)
    #'user/*set-me*
    user=> (defn try-to-set [] (binding [*set-me* 1] (doall (pcalls #(if (thread-bound? #'*set-me*) (set! *set-me* (inc *set-me*)))))))
    #'user/try-to-set
    user=> (try-to-set)
    IllegalStateException Can't set!: *set-me* from non-binding thread clojure.lang.Var.set (Var.java:230)
    user=>
{noformat}

*Approach:* thread-bound? should return false in the case where there is a binding *and* that binding was not established by the current thread. The patch adds a new function to Var as core does not have visibility into the non-public Var$TBox.

*Patch:* thread-bound.diff

*Screened by:* Alex Miller

People

Vote (4)
Watch (1)

Dates

  • Created:
    Updated: