143634 - dependency-tree.html.tmpl has infinite loop if there is a dependency loop

Status: RESOLVED DUPLICATE of bug 143560

(Bugzilla :: Bugzilla-General, defect, P1)

Description

[Gervase Markham [:gerv]]

We have code which checks for dependency loops. It obviously isn't working - on
Landfill, bug 4 depends on bug 5, and bug 5 depends on bug 4, and if you go to
the dependency tree of either bug,

May  8 04:41:01 landfill kernel: Out of Memory: Killed process 9995
(showdependencyt).
May  8 04:47:58 landfill kernel: Out of Memory: Killed process 10007
(showdependencyt).
May  8 05:04:48 landfill kernel: Out of Memory: Killed process 10272
(showdependencyt).
May  9 16:35:07 landfill kernel: Out of Memory: Killed process 1104
(showdependencyt).

So, we need to fix the loop-checking code :-)

Gerv

Comment 1

[Gervase Markham [:gerv]]

Attachment: Patch v.1

The problem is not in the CGI, but in the template. The CGI's recursion control
works correctly. However, in the template, we do the following in the code for
displaying a single entry:

      [% INCLUDE display_tree bug_id=dep_id 
	   IF dep.dependencies.size > 0 && !dep.seen %]
      [% dep.seen = 1 %]

Obviously, if there's a recursive loop, dep.seen will never get set to 1 on any
of the bugs in the loop.

The fix is to move the mark to the beginning of the block, as in the attached
patch. I have a local copy of the landfill DB, and this fixes the problem for
me.

Gerv

Comment 2

[Gervase Markham [:gerv]]

Tobias - this is your code. Could you review? :-)

Gerv

Comment 3

[Tobias Burnus]

+  [% tree.$bug_id.seen = 1 %]
This looks ok since the bug of which the dependencies are shown is marked as
seen. (Fixes A->B->[C->D->]A)

-    [% dep.seen = 1 %]
This is equivalent to tree.$dep_id.seen.
(For the case A->B->C->B)

While I can see the point for adding the $bug_id line, I don't think it makes
sense to remove the other line.

if you keep both lines, you get a r=burnus, but note that I didn't do this part
but Myk Melez did.

Comment 4

[Bradley Baetz (:bbaetz)]

Dupe of bug 143560, which has exactly the same patch on it...

See my comments there.

*** This bug has been marked as a duplicate of 143560 ***

Comment 5

[Dave Miller [:justdave]]

clearing target in DUPLICATE/WORKSFORME/INVALID/WONTFIX bugs so they'll show up
as untriaged if they get reopened.