The last update to dnsmasq has run fine on all of our IPFire systems with the Google and Level 3 DNS servers for an extended period of time. Could all of the issues be a conflict with the dnsmasq code and *certain* DNS providers? If so this seems like a much harder problem to fix.
Fred
-----Original Message----- From: Matthias Fischer [mailto:matthias.fischer@ipfire.org] Sent: Tuesday, January 19, 2016 2:59 PM To: development@lists.ipfire.org Subject: Re: [PATCH] dnsmasq 2.75: next patch... (No.50)
Hi,
On 19.01.2016 08:39, R. W. Rodolico wrote:
I installed this package yesterday and dnsmasq broke a few minutes
ago.
About 24 hours.
Worked for 24 hours and then crashed? Weird. But thinking of it, this could be one of the reasons why its not crashing *here*. Every 24 hours my ISP cuts the connection, and our router gets a new IP and 'dnsmasq' restarts!
Strange thing: when I was using the servers you recommended 84.200.69.80, 84.200.70.40 I did not have any problems even though you
have been updating very frequently. However, I reverted to the old DNS
servers 209.244.0.3 8.8.4.4 and in less than a few days (I think I did it with the 15th or 14th update), it broke again. Those servers are, respectively, resolver1.level3.net and one of the google ones.
So as I understand you, its related with the DNS servers you use!? With your ISP servers, it crashes and with the german servers, it doesn't?
Let me know if you want me to use the 84 DNS servers. Hell, I may just
decide to build my own caching DNS servers!!!
Ok, solution found - that was easy... ;-))
But as long as I'm working on this, I'm more and more sure that these intermediate crashes are really a combination between the DNS servers and 'dnsmasq' itself. Sad to say, I haven't got the (C-)skills to debug this.
Right now, I'm trying to compile 'dnsmasq' with patches 051-053: it now crashes in 'forward.c' during compilation if I activate the DNSSEC-option (-e 's|/* #define HAVE_DNSSEC */|#define HAVE_DNSSEC|g' ). Reason: patch No.053, I think.
In contrast, the '2.76test6'-version compiles without *any* error, if I delete and disable *all* of our adjustments.
As Simon wrote: "Conditional combination has a nasty combinatorial explosion. I should hack up a regression test to build all possible variants."
But obviously this wasn't done yet. If I leave everything as it is, its building...
Perhaps I'll try another request on the dnsmasq-list during the next days. The last one had no effect, no answers, nothing.
Best, Matthias