Segmentation fault from subscriber defined in dotimes

[pazeshun]

I want to define multiple subscribers by dotimes, but that trial failed with segmentation fault.
Here is the simple sample code:

(ros::roseus "test" :anonymous nil)
(ros::roseus-add-msgs "std_msgs")

(setq test-list (list 1 2))
(setq topics (list "~test1" "~test2"))
(dotimes (i (length test-list))
  (ros::subscribe (elt topics i) std_msgs::Bool
                  #'(lambda (a msg) (print (elt test-list a))) i 1))
(ros::spin)

When I published to /test/test1 or /test/test2, segmentation fault occurs.
Debugging with @Affonso-Gui found the following code also occurs segmentation fault:

(ros::roseus "test" :anonymous nil)
(ros::roseus-add-msgs "std_msgs")

(setq var 0)
(let (b)
  (ros::subscribe "~test" std_msgs::Bool
                  #'(lambda (msg) (print var))))
(ros::spin)

When we erase local variable b, add one more argument to lambda fuction, or don't use global variable var, segmentation fault doesn't occur.

[pazeshun]

We found dotimes problem comes from let by expanding macro:

4.irteusgl$ macroexpand '(dotimes (i 1) (print 1))
(let ((i 0) (#:dotimes355 1)) (declare (integer i #:dotimes355)) (while (< i #:dotimes355) (print 1) (setq i (1+ i))) nil)
5.irteusgl$ pprint *
(let
   ((i 0) (#:dotimes355 1))
   (declare (integer i #:dotimes355))
   (while (< i #:dotimes355) (print 1) (setq i (1+ i)))
   nil)
nil

[furushchev]

It looks like this is related to some of the existing issues:
#251
#324

This problem is originated from the limitation of euslisp: a closure is unsupported.

As seen in the link below, lambda function in let scope is defined differently from it in global scope.
#324 (comment)

In the example you mentioned, this problem can be avoided by changing codes like below for example:

(ros::roseus "test" :anonymous nil)

(setq test-list (list 1 2))
(setq topics (list "~test1" "~test2"))
(dotimes (i (length test-list))
  (ros::subscribe (elt topics i) std_msgs::Bool
    `(lambda-closure () 0 0 (a msg) (print (elt test-list a))) i 1))
(ros::spin)

[k-okada]

related to euslisp/EusLisp#147 (comment) ?

[pazeshun]

related to euslisp/EusLisp#147 (comment) ?

I think so. This issue comes from the known issue.
The main point of this issue becomes an alert of implicit let inclusion like dotimes.