Re: (E)LTS report for April 2023

[Dameon Wagner <dameon.wagner@it.ox.ac.uk>]

On Wed, May 03 2023 at 17:51:20 +0100, Dameon Wagner scribbled
 in "Re: (E)LTS report for April 2023":
> On Mon, May 01 2023 at 12:33:51 +0200, Tobias Frost scribbled
>  in "(E)LTS report for April 2023":
> > I've worked during April 2023 on the below listed packages, for Freexian
> > LTS/ELTS [1]
> <SNIP>
> > ELTS:
> > =====
> > 
> > syslog-ng: ELA-832-1 for stretch and jessie (CVE-2022-38725)
> 
> Hi Tobi,
> 
> We've just upgraded a few systems with syslog-ng=3.5.6-2+deb8u1, but
> have seen some issues with the daemon continuously restarting, and
> have reverted to 3.5.6-2 for the time being, which solved the issue.
> 
> Looking at the backported patches I _think_ the issue might be with
> "patches/0027-syslogformat-sd_step-_process_any_char.patch", in the
> hunk around line 169:
> 
> #---8<-----------------------------------------------------------------
> @@ -502,16 +502,16 @@
>          }
>        if (left > 0 && *src == ']')
>          {
> -          ssd_step_and_store(self, &src, &left);;
> +          s_process_any_char(self, &src, &left);;
>          }
>      }
> #---8<-----------------------------------------------------------------
> 
> which differs from the other changes in the patch that are just
> `s/sd_step_and_store/_process_any_char/` and don't have the doubled-up
> semicolons?
> 
> I'm planning to build a test package in the morning and will report
> back if it seems to fix things -- unless you get there first that is!

I can confirm that fixing the above solves the continuous restart
issue we were seeing.  Adding the following patch to the end of the
series in debian/patches and rebuilding was all that was required:

#---8<-----------------------------------------------------------------
--- a/modules/syslogformat/syslog-format.c
+++ b/modules/syslogformat/syslog-format.c
@@ -507,7 +507,7 @@
         }
       if (left > 0 && *src == ']')
         {
-          s_process_any_char(self, &src, &left);;
+          _process_any_char(self, &src, &left);
         }
     }
   if (left > 0 && *src == ':')
#---8<-----------------------------------------------------------------

Though it may be neater to fix the earlier patches where the initial
issue was introduced (I'm not sure where there are preferences or
policies regarding a fix like this), namely:

*   0025-syslogformat-utilize-sd_step.patch
*   0027-syslogformat-sd_step-_process_any_char.patch

Thanks again.

Dameon.

-- 
><> ><> ><> ><> ><> ><> ooOoo <>< <>< <>< <>< <>< <><
Dr. Dameon Wagner, Unix Platform Services
IT Services, University of Oxford
><> ><> ><> ><> ><> ><> ooOoo <>< <>< <>< <>< <>< <><