I think so.
Like I said in #76 of this thread, I usually only preserve a couple of config files:
/etc/config/dhcp, firewall, luci_statistics, qos, wireless and radvd, /etc/dropbear/authorized_keys for having SSH key-based authentication and /etc/vsftpd.conf for FTP server.

All other config files get always updated from the firmware defaults in the flash image. I think that those two config files you mentioned, have disappeared, but I haven't really checked.

K thanks. Probably time to wipe my configs and reset from defaults anyway. It's been a while and a lot of updates have touched the settings.

Noticed one odd thing with the new firmware, Firefox was stuck on some "waiting" displays for the LAN/WAN settings, current time, didn't display the new "Auto refresh" at the top, and didn't show a few other subfields in the gui. IE9 was working though. I did a shift-refresh of the admin page  from FF and everything was fine. Maybe the html timestamps were stale - or maybe related to the missing ntp sync. But the issue didn't clear up until a manual refresh.

(Last edited by tt on 3 Nov 2011, 15:55)

I have noticed something similar when switching between Backfire Luci 0.10 and Trunk Luci trunk via firmware flash. They have different icon sets etc. and sometimes the displays do not refresh properly in Firefox (9.0a2) after the flash unless doing the manual refresh. I haven't noticed any problems once Firefox has got rid of all the stuff related to the other Luci version.

Regular FF 7.0.1 here. Well, easy enough to work around.

Is there some simple way to see the default config files, without resetting the router to defaults?  Are they sitting on one of the mtd partitions somewhere? The "firstboot" script details are a little hard to crack.

Not to my knowledge. I have checked my config files against the originals in package sources.
E.g. firewall in https://dev.openwrt.org/browser/trunk/p … all.config , etc.

If I remember correctly, the only real changes in the past few months have been in the firewall default rules and then this rdate/ntpd change.  (network, qos, wireless have not changed.)

EDIT: My bad. Of course there is...
compare the files in /rom directory against the actual ones. E.g. "diff /rom/etc/config/firewall /etc/config/firewall" etc.
But as LuCI changes the format by adding single quotes, you get a bit too many differencies.

(Last edited by hnyman on 3 Nov 2011, 17:10)

A-ha! That helps a lot, thanks. Mine has a lot of old/obsolete settings hanging around, it seems.

It would be nice if there was a way to save only a set of "overrides". Hmm. Maybe a file with a list of "uci" commands, that could be executed manually. Kind of like your "hnsettings.sh" approach, but with uci commands instead of a tarball.

(Last edited by tt on 5 Nov 2011, 02:53)

Here's a quick script to show the config file differences a little more clearly by fixing up the single quotes first. The settings aren't always displayed in the same order, so sometimes it shows a difference which isn't significant, but it's close enough to eyeball!

Also it seems that some configurations don't have a default, such as /etc/config/wireless which is always generated. So it's worth checking for missing files by hand.

#!/bin/sh
cd /etc/config
for F in *
do
        echo
        echo "===== $F ====="
        uci -c /rom/etc/config export $F >/tmp/$F.default 2>/dev/null
        case $? in 0) ;; *) echo '>> no default <<' ;; esac
        uci export $F >/tmp/$F.current
        diff /tmp/$F.default /tmp/$F.current
        rm -f /tmp/$F.default /tmp/$F.current
done

Yep, wireless is always generated, and similarly LED definitions get inserted into system from uci-defaults.

That might be an ok approach. Your diff script might help finding the differences, and then it is only question about writing the proper uci commands to modify settings.

That approach might even be better than mine (replacing whole config files), as the smaller changes in definitions (like added default firewall rules) will get implemented automatically by using the default config files and then you personal settings are just modify a single item in the config file.

EDIT: I will include that script of yours as /etc/checksettings.sh in my builds.

(Last edited by hnyman on 5 Nov 2011, 10:51)

Sounds great! I plan to poke around the uci stuff in more detail, the "-m import" and "add" options look interesting. I'm thinking the diff script can generate something that can be mangled into a uci file that can capture changed settings. That should in theory be enough to perform a basic migration.

Does your build have Jumbo frame passthrough support like arohk's build?

I believe jumbo passthrough support has been added in trunk some time ago.

Yes, for the trunk as it is now a standard part of trunk like Arokh said.
https://dev.openwrt.org/changeset/28267

But I think that the support has not been backported to Backfire, and I have not done it separately. Haven't really checked the matter.

(Last edited by hnyman on 17 Nov 2011, 21:00)

OK, thanks for the answer. I am running your Backfire edition, so I'll propably wait for awhile. If it gets into Backfire or if the lack of it starts to bugg me then I'll switch to trunk.

Not many changes to my code during the month, as there haven't been so many significant changes in trunk and even less in Backfire.

Adding timestamp to kernel logs was a nice addition (https://dev.openwrt.org/changeset/29353) for trunk, so I added that also for Backfire in version 29399. I also made a ticket suggesting adding that officially to Backfire (https://dev.openwrt.org/ticket/10540).

Something has changed, because I can not install igmpproxy on this to built backfire-r29195 and backfire-r29399 for WNDR3700 v1 ,
and usually no problem with download and install opkg packages on your built.
I switch to trunk and there were no problems.

Interesting, I first thought that it would be some minor change in kernel between rc6 and today, but I noticed that Backfire r29399 crashes even when I try to install a self-compiled additional package. Router reboots immediately.

Did I understand you correctly that you also tested r29195? And it also crashed?

It seems to crash with both LuCI and commandline opkg, so it is probably nothing to do with LucI.

EDIT: after some testing I have notice that installing packages still works with the original r28681 (which is identical to rc6), but fails with 2011-11-13-r28997 both with rc6 repository modules and with self-compiled modules. I will search for the regression point, but this looks like a real regression either with Backfire sources or one of the add-on modules (like the changed Busybox version etc.).

EDIT: looks rather strange, as a newly recompiled r28681 also fails. Makes me think that the switch from Ubuntu 11.04 to 11.10 as the build environment may have caused something.

(Last edited by hnyman on 4 Dec 2011, 22:23)

yes you have understood me correctly I tested also r29195 , and with both LuCI and commandline and it simply crash and then restart.

You might test the new Backfire r29422:  http://koti.welho.com/hnyman1/Openwrt/b … 011-12-04/

I compiled it using the old Ubuntu 11.04 build environment, and it installs packages flawlessy. I tested with several packages including your igmpproxy.

The only change in code compared to the failing r29399 (which I have already removed from the FTP), is that I have disabled the kernel log timestamp (printk). Otherwise the same config.

So, either the culprit would be printk (but it was not included in r29195), or more likely the Ubuntu 11.10 environment somehow produces a different binary than my old 11.04 environment. Sounds strange, but so far the #1 suspect.'

I just tested the new Backfire r29422, and it works.

Thanks.

Seems to have a package issue. It works fine but on my WNDR3700v1, the Network->Wifi Luci page now prints:

Installing it gives no joy though.

(Last edited by tt on 6 Dec 2011, 03:50)

Probably due to jow moving iwinfo from LuCI to main Backfire just at the same time I built it. I will probably need to rebuild .config to get all the new dependencies correct.

Well, I figured out the missing dependency on libiwinfo-lua just at the same time as Jow fixed it in source repository.
http://luci.subsignal.org/trac/changeset/8055

the Backfire version r29461-2011-12-06 should have both a working Wifi config and package installation capability.
http://koti.welho.com/hnyman1/Openwrt/b … 011-12-06/

(Kernel log timestamps are enabled, so the only material difference between this and the "crash when installing new packages" version is that this is compiled in Ubuntu 11.04, not 11.10. I have to try to debug and find out which part of the build environment causes the problem.)

All better! Also the log timestamps are a very nice addition. Thanks.

I wrote a longer story to the Openwrt-devel mailing list, but based on my debug work so far, it looks like the kernel version of the underlying buildsystem for some reason has impact on the kernel configuration (?) of the firmware to be compiled. And as Ubuntu 11.10 has kernel 3.0, that breaks things compared to the 2.6 kernel in Ubuntu 11.04 (and in Backfire).

https://lists.openwrt.org/pipermail/ope … 13100.html

There might be an actual problem with kernel config with the 11.10 compiled version. Around line 29300-30000 of the log there is the part with scripts/kconfig. And there the 11.04 version log talks about /boot/config-2.6.38-12-generic, while the 11.10 log contains /boot/config-3.0.0-13-generic. That difference then causes several differences for the various kernel build option selections made by the build process.

Neither of those versions mentioned in the build log is actually the 2.6.32.27, which the Backfire kernel actually is.

Apparently the /boot/config... file mentioned in the build log below is from the buildhost system, not from the Openwrt source. Strange.

@@ -29338,9 +30082,10 @@
    HOSTCC  scripts/kconfig/zconf.tab.o
    HOSTLD  scripts/kconfig/conf
  scripts/kconfig/conf -o arch/mips/Kconfig
-/boot/config-2.6.38-12-generic:3812:warning: symbol value 'm' invalid for 
FB_VESA
+/boot/config-3.0.0-13-generic:2182:warning: symbol value 'm' invalid for IWL4965
+/boot/config-3.0.0-13-generic:3956:warning: symbol value 'm' invalid for FB_VESA
  #
-# using defaults found in /boot/config-2.6.38-12-generic
+# using defaults found in /boot/config-3.0.0-13-generic
  #
  *

(Last edited by hnyman on 6 Dec 2011, 20:56)

I have updated the Backfire version to r29594. That matches the code of the final Backfire 10.03.1 release, as r29594 is the tagging of the 10.03.1 release at the svn repository, made today. (The official 10.03.1 version can be said to be 29592 instead, but it is the same Backfire code.) My build should also match the LuCI version of the final 10.03.1, as LuCI r8131 is the tag of "0.10.0" in LuCI svn.

Backfire 10.03.SVN (r29594) / LuCI 0.10 Branch (0.10+svn8131)
http://koti.welho.com/hnyman1/Openwrt/b … e-10.03.1/

Trunk:
with the trunk build I have wondered about the stability of the new iwinfo based wireless-stats tab in luci-statistics. The recent builds have been occasionally crashing if that graph is enabled. But when the iwinfo section is disabled in luci-statistics, the last build r29571-2011-12-19 has been quite stable. The reason of the crashes is not quite clear yet, but I am keeping the graph off for the time being, just in case. Related discussion: https://dev.openwrt.org/ticket/10630 , http://luci.subsignal.org/trac/ticket/363

(Last edited by hnyman on 22 Dec 2011, 19:29)