00:54 <schmars[m]> 23.05 imagebuilde has a hiccup: https://github.com/openwrt/openwrt/issues/13881 - i have the same issue. will post better formatted logs lol

01:28 <philipp64> ldir: who owns dnsmasq?

01:29 <philipp64> and if so, can you review PR #13765?

03:48 <owrt-images-builds> Build [#124](https://buildbot.staging.openwrt.org/images/#/builders/120/builds/124) of `openwrt-23.05_x86/generic` failed.

03:52 <dwfreed> schmars[m]: does it work if you revert https://github.com/openwrt/openwrt/commit/21e5db97c410f4008c8fe8515fb79a7cde368dbf

09:33 <kabel> Ansuel: so I backported to kernel 5.15 the qca8k changes that add support for accessing internal PHY registers via management frames

09:34 <kabel> Ansuel: and even with the fix of locking the mdio bus, the conflict with the external WAN PHY still occurs there

09:34 <kabel> Ansuel: I did not get to the root of the problem there, but I have a suspicion of what may be happening

09:41 <kabel> Ansuel: the qca8k_phy_eth_command() function locks the bus, sends mgmt packets, waits for reply packets, and unlocks the bus. But it only waits for a specified amout of time (2000 msecs?)

09:41 <kabel> Ansuel: so what if it sends the packet to access the register, but for some reason the packet is stalled in the egress queue of the network device for some reason?

09:52 <kabel> Ansuel: the mgmt packet reaches the qca8k chip too late, the driver already times out and unlocks the mdio bus

09:52 <kabel> Ansuel: but I am not sure, since I cannot reproduce this on master

17:54 <ynezz> open source and risc-v as a national security risk :P https://selectcommitteeontheccp.house.gov/sites/evo-subsites/selectcommitteeontheccp.house.gov/files/evo-media-document/11.1.23-riscv-letter.pdf

17:59 <f00b4r0> between that and the EU's CRA... 🤦‍♂️

18:02 <schmars[m]> dwfreed: it did :) i added the patch from master to our build tooling temporarily until new imagebuilder tarballs are built

18:11 <gladiac> So I invested more time in getting the binary size for vaultwarden down even more. I think I hit rock bottom with 12.8MB.

18:12 <gladiac> The final optimization size-wise was activating panic-on-abort. This option alone decreased the binary-size by over 2MB

18:13 <gladiac> This page was really helpful in optimizing rust binary-sizes: https://github.com/johnthagen/min-sized-rust#abort-on-panic

18:15 <gladiac> The standard vaultwarden release build for amd64 produces a 38MB binary. In contrast to that I consider 12.8MB a big win.

18:36 <gladiac> However, vaultwarden only includes a web-ui for it's own admin-interface. There is an OPTIONAL web-ui to actually use vaultwarden. It is optional because the webservices are enough for the ios/android apps. If all you want is app-support, you do not need it. However, if you do want the web-ui, that is another 42.3MB of angular "goodness".

18:37 <gladiac> I am thinking of making the web-ui a separate package so endusers can decide if they need it or want to save those 40+MB of css/js stuff.

18:39 <gladiac> Fun fact, the web-ui basically is just the original bitwarden web-ui, just with minor adaptations to reflect it's usage on a vaultwarden server instead of the official bitwarden server and a few other little things.

18:54 <Mangix> f00b4r0: ping

19:17 <f00b4r0> Mangix: pong?