HW offloading helps WAN-LAN throughput/NAT or is there more to it?
What kind of WAN-LAN throughput can we expect on our R7800s just with software NAT?
In my experience R7800 can do near-gigabit up and down on the test without offloading of any kind. At least, the 17.x builds did. Not sure if it can handle full-duplex gigabit going at the same time.
@hnyman one question did you recompile package between version upgrade? can you try to clean openssl and nginx and recompile them ? This could be the cause of the broken nginx package without engine support.
I need help.
Anyone knows about this?
openwrt-18.06 and master
what different
how can i choose
I have not compiled nginx by myself.
I believe that the problem arises when somebody uses opkg to install and use the buildbot-compiled nginx to a private system, as buildbot currently builds openssl with non-default engine options (as they are forced by some other packages) so nginx possibly picks that support in the compile phase in buildbot, while a private firmware built with the default openssl settings has no engine support.
18.06 is the current "stable" branch
master is the branch fior ongoing development.
If you are unsure what you want, use the stable 18.06
OK THX your reply
Hello,
i set up a virtualbox environment with debian just to build the firmware.
So, the first thing i want to say @hnyman is THANK YOU because i never realized it takes so long to build a firmware file.. i could assign just one core of my i5 notebook to the virtual machine and it took an unbelievable 6 hours to build. Thanks a lot for your work
That said, now i'm being amused by this possibility and i'd like to ask you:
- how can i, if i can, modify the list of packages installed? for example, if i do not want uhttpd and standard luci, adding nginx ssl and luci for nginx, how should i do? I'd like to start from your image because i find it.. well.. close to perfect
- if i add nginx to the packages to be installed, is it going to be built with correct options to work on the build or should i modify something?
Again thanks for your time
Well, i give really detailed advice in message #2 of this thread. As you already have a build system, simply copy my .config.init and start modifying it. Like I say in message 2, use make defconfig or menuconfig to expand it to full .config before build.
Unless you're giving the VM just one core that sounds really slow unless you're on a first gen Core i5 CPU. I know that Meltdown and Spectre adds performance penalties but that's slow.
It takes roughly 1h on my old ThinkPad T430 to compile the toolchain and stock firmware however on barebone Alpine 3.9 without ccache.
Intel(R) Core(TM) i5-3320M CPU @ 2.60GHz , 8Gb of RAM, Mechanical HDD
make -j4
well, yesterday night i had some more fun with this thing
i discovered the wonderful world of make menuconfig (sorry @hnyman, i really hadn't understood that part of the second message..) and i also gave more cores and more ram to the VM. if i find some time this evening i'll try to compile the image with nginx
thanks a lot both of you
(i have a i5 3210 so it should be slightly slower than yours, i dedicated 4 cores - even in virtualbox complained about that - and 4GB, i hope i'll get it compiled within an hour or so :))
thank you @hnyman for your build. it is very stable, fast and very much appreciated.
The devs have committed a few minutes ago a change in openssl, so that it will have the engine support on by default (matching the current de-facto status in buildbot forced by some other packages).
https://github.com/openwrt/openwrt/commit/450d44a8ead2217f8acf541a4eaa4ad560b3e5ac
So, my future builds (20190417 and later, in master) will have that config symbol on, and the engine related problems should then disappear.
Hello, sadly i have another question for you
after having created a build environment with your script, what happens when the master version changes?
can i "simply" go on using the updateaNmake.sh script to always have the image patched with your patch files or have i all the times to start back downloading patches from dropbox?
do your patches change version by version?
thanks a lot
updateNmake.sh script makes a git pull call and tries also to update the feeds. If there are no incompatible changes, the whole process is automatic and nothing else is needed.
But in those rare cases, where my old patches create a conflict with the new main sources, you need to use text editor and git commands to resolve the possible conflicts.
Ps.
always remember to do a "make clean" before a build.
Have the lockups been resolved? I've used Kongs build of Openwrt and it's stable. But the webgui is not (I think I have to reboot the router before I can access the webgui), and it comes with a horrible theme by default, and lastly I'd like to roll my own build with my preferred packages preinstalled with the builder.
I think Kong just used these patches: http://www.desipro.de/openwrt/sources/
I am having an issue in this version.
Headsup:
Revert "uhttpd: disable concurrent requests by default"
Have to edit /etc/config/uhttpd
by hand to set option max_requests 3
and then run /etc/init.d/uhttpd restart
Is vsftpd
gone from the latest builds? Not that I need it or miss it, just checking.