cell.entries assignation may delete part of the stored content

Reported by Samuel GOUGEON (@sgougeon)

-- Bug description --


cell.entries assignation may delete part of the stored content


-- Scilab error message --


-->a=makecell([2,2], -3, [ 'a' 'b' ; 'f' 'gh'], %z, list( -1, "ok" ))
 a  =
!-3  ["a","b";"f","gh"]  !
!                        !
!z   list(-1,"ok")       !
 
-->a(2,2)
 ans  =

!-1  "ok"  !
 
-->a(2,2)(2)
 ans  =
  
 "ok"   
 
-->a(2,2)(2)="nok"
 !--error 10000 
Invalid assignement: for insertion in cell, use e.g. x(i,j).entries=y
at line       3 of function generic_i_ce called by :  
at line       3 of function %c_i_ce called by :  
a(2,2)(2)="nok"
// OK. Then, 
 
-->a(2,2).entries
 ans  =
       ans(1)
 
  - 1.  
 
       ans(2)
 ok   
 
-->a(2,2).entries(2)
 ans  =
 
 ok   
 
-->a(2,2).entries(2)="nok"
 a  =
 
!-3  ["a","b";"f","gh"]  !
!                        !
!z   -1                  !

-->a(2,2).entries(2)
                  !--error 21 
Index invalide.
// The list(-1,"ok") is updated into -1 (constant) instead of into list(-1,"nok") (list)


-- How to reproduce the bug --


a=makecell([2,2], -3, [ 'a' 'b' ; 'f' 'gh'], %z, list( -1, "ok" ))
a(2,2)
a(2,2)(2)
a(2,2)(2)="nok"
a(2,2).entries
a(2,2).entries(2)
a(2,2).entries(2)="nok"
a(2,2).entries(2)
typeof(a(2,2).entries)

assigned to @vincent.couvert

added 5.0 through 5.3.x Bugzilla Lists cells structs Major PC labels

By adrien vogt-schilb

I would say this problem should be noted has "major severity" I am experimenting a lot of troubles when using tlists and hypermatrices, they often get corrupted, fields (or dimensions) use to disapear during execution, randomly (therefore, I cannot post any steps to reproduce the bugs). People in my "team" (at work) just warn new guys like me not to use tlist and hypermatrices because they find them not reliable. Usually, trying to use tlist leads to switching to matlab.

They lead to non reproductible bugs, so we avoid using them and none of use is able to post here in bugzilla. I believe that resolution of this kind of bugs (and similar bugs like those listed at the end of this post ) could help finding out if there is something wrong in the tlists implementations. (and thanks a lot the scilab team for the great job) see also: #6607 (closed) #5602 (closed)

By Samuel GOUGEON (@sgougeon)

(In reply to #5602 (comment 1126030439))

I would say this problem should be noted has "major severity" Done. Moreover, version reset from GIT 5.2 to 5.2.1

By Francois Granade

(orginally entered in version "5.2.1 final version")

By Samuel GOUGEON (@sgougeon)

By the way,

-->a(2,2)(2) ans =

"ok"

is erroneous: It should yield an error, since the addressed cell is scalar whereas a(2,2).entries(2) // is ok

By Samuel GOUGEON (@sgougeon)

Non regression test added: https://gitlab.com/scilab/legacy_codereview/-/wikis/changes/19457

By Antoine ELIAS (@antoine.elias)

merged @​​ 843f3cea

closed

mentioned in issue #5588 (closed)