Rudder

After a short diving session in the CFengine SVN, we found a commit that might match this bug : r3007

Patch :

--- src/verify_packages.c    (révision 3006)
+++ src/verify_packages.c    (révision 3007)
@@ -1688,7 +1688,7 @@
           }
        else
           {
-          cfPS(cf_error,CF_FAIL,"",pp,a,"!! Package \"%s\" cannot be verified -- no match\n",pp->promiser);
+          cfPS(cf_inform,CF_FAIL,"",pp,a,"!! Package \"%s\" cannot be verified -- no match\n",pp->promiser);
           }

        break;
@@ -1960,7 +1960,7 @@
     ReplaceStr(line,lineSafe,sizeof(lineSafe),"%","%%");
     CfOut(cf_inform,"","Q:%20.20s ...:%s",cmd,lineSafe);

-    if (line[0] != '\0' && verify)
+    if (verify && line[0] != '\0')
        {
        if (a.packages.package_noverify_regex)
           {
@@ -1976,21 +1976,25 @@

  packmanRetval = cf_pclose(pfp);

- if(verify)  // return code check for verify policy
+ if(verify && a.packages.package_noverify_returncode != CF_NOINT)
     {
-    if(a.packages.package_noverify_returncode != CF_NOINT)
+    if(a.packages.package_noverify_returncode == packmanRetval)
        {
-       if(a.packages.package_noverify_returncode == packmanRetval)
-      {
-          cfPS(cf_inform,CF_FAIL,"",pp,a,"!! Package verification error (returned %d)",packmanRetval);
-          retval = false;
-      }
-       else
-      {
-          CfOut(cf_verbose, "", " Package successfully verified from return code");
-      }
+       cfPS(cf_inform,CF_FAIL,"",pp,a,"!! Package verification error (returned %d)",packmanRetval);
+       retval = false;
        }
+    else
+       {
+       cfPS(cf_inform,CF_NOP,"",pp,a,"-> Package verification succeeded (returned %d)",packmanRetval);
+       }
     }
+ else if(verify && a.packages.package_noverify_regex)
+    {
+    if(retval)  // set status if we succeeded above
+       {
+       cfPS(cf_inform,CF_NOP,"",pp,a,"-> Package verification succeeded (no match with package_noverify_regex)");
+       }
+    }
  else if(setCmdClasses)  // generic return code check
     {
     retval = VerifyCommandRetcode(packmanRetval,true,a,pp);

I will confirm the bug on our version, and try to recompile it with this patch for testing purpose.

This is a bug. It has been reported to CFengine => https://cfengine.com/bugtracker/view.php?id=831 , now we have to wait for a solution.

It seems that this bug has not been fixed yet.

This should not be in a Discussion state since a patch must be applied during packaging.

Nicolas, you set this ticket in Progress without any assignee. A ticket in progress must have an assigne, so I assign it to you.

I'll catch this, since NPE is unavailable at the moment.

I took a look at the CFEngine 3.2.0 and 3.2.4 code base (= Rudder 2.3 and 2.4), and the patch is totally inapplicable.

As discussed with NCH, the patch is originally made for the CFEngine 3.3.X code base, that has been drastically simplified and modified by Mikhail Gusarov, and the impacted function is completely different, see the example below:

static int ExecuteSchedule(struct CfPackageManager *schedule,enum package_actions action)

{ struct CfPackageItem *pi;
  struct CfPackageManager *pm;
  int size,estimated_size,retval = true,verify = false;
  char *command_string = NULL;
  struct Attributes a = {{0}};
  struct Promise *pp;
  int ok;       

for (pm = schedule; pm != NULL; pm = pm->next)
   {
   if (pm->action != action)
      {
      continue;
      }

   if (pm->pack_list == NULL)
      {
      continue;
      }

   estimated_size = 0;

   for (pi = pm->pack_list; pi != NULL; pi=pi->next)
      {   
      size = strlen(pi->name) + strlen("  ");

      switch (pm->policy)
         {
         case cfa_individual:             

             if (size > estimated_size)
                {
                estimated_size = size + CF_MAXVARSIZE;
                }             
             break;

         case cfa_bulk:

             estimated_size += size + CF_MAXVARSIZE;
             break;

         default:
             break;
         }
      }

   pp = pm->pack_list->pp;
   a = GetPackageAttributes(pp);

And the patch:

@@ -856,6 +856,9 @@ static int ExecuteSchedule(PackageManager *schedule, enum package_actions action

                 for (pi = pm->pack_list; pi != NULL; pi = pi->next)
                 {
+                    pp = pi->pp;
+                    a = GetPackageAttributes(pp);
+
                     char *sp, *offset = command_string + strlen(command_string);

                     if (a.packages.package_file_repositories && (action == cfa_addpack || action == cfa_update))

I can not reasonably apply this patch blindly, without beeing able to balance the consequences, I'm sorry.

Jon, what should we do here ?

I have just checked and I confirm that this issue is no longer present in CFEngine 3.4 (and therefore in Rudder 2.6+).

It does however still affect CFEngine 3.2 and therefore Rudder 2.4 (and 2.3 and 2.5 but they are no longer supported). Fixing this in CFEngine 3.2 is non trivial, and upstream has been clear they do not intend to provide patches or new versions on older releases.

As a result, I can't see us putting a high priority on this bug, for now.

Actually, I take that back, it's not that hard. I attach a patch that should fix this. I haven't tested it, but it is issued from https://cfengine.com/bugtracker/view.php?id=829 and seems logical to me.

Since version 2.4 is not maintained anymore, retageting this issue to 2.6

This ticket seems to be out of date, since we only support CFEngine >= 3.5 now.

Could it be closed as won't fix - 2.6 not supported anymore ?

Nicolas ?
Isn't this the bug you are working on ?

Bug no longer present in 3.4 and later.
Closing