13:11 <mwalle> is there any reason why the port (80) is hardcoded in the new wget command?

14:27 <rfs613> mwalle: i haven't looked, but can you use http://server:port/whatever syntax?

15:29 <mwalle> rfs613: I had a quick look at the source code and the SERVER_PORT seems to be #defined to 80

15:32 <mwalle> rfs613: oh, Sean had a remark that the syntax looks quite odd, ie. wget loadaddr ipOrHost:path/to/file, and there was a comment that URI schemes could be possible in the future

15:32 <mwalle> (wget reuses the parsing, ie. the same syntax as tftp or nfs)

16:29 <Tartarus1> Yeah, that's why it's a bit funky, it follows u-boot convention, not wget convention

19:34 <Flole> I just discovered some kind of bug/issue that appeared between 2022.04 and 22.07: On 2022.04 I can boot my odroid c2 over pxe perfectly fine, when I upgrade to 2022.07 I am seeing an "Synchronous Abort" handler, esr 0x96000004" after it wrote "Working FDT set to 7a950000". There have been almost 2000 commits between those 2 releases, so it's very hard to figure out why/how this broke, does

19:34 <Flole> anybody have a clue?

19:52 <cambrian_invader> flokli: git bisect?

19:52 <flokli> Wat

19:52 <cambrian_invader> man git bisect

19:52 <flokli> Ack

19:52 <flokli> Bit bisects gits

20:38 <Flole> Alright, I'll try to bisect and see if I can figure it out

20:43 <Flole> Another thing I noticed: When I tftp a file and there is a ping towards the device running u-boot there are timeouts in the tftp process. Is there anything I can do about that? Increase some buffer or something?

22:43 <Tartarus1> We do not have a high performance network stack, don't do that if you can just not do that ;)

22:58 <Flole> Unavoidable unfortunately, many devices try to reach this thing, I already increased the interval at what they are trying to connect but it's still too much. Would increasing the RX buffers help? Or not optimizing the u-boot build for size so it becomes faster?

22:59 <Flole> I currently have it set to 4, maybe 16 would be more appropriate?

23:09 <Flole> Just tried optimizing for speed, turning on LTO and raising that buffer setting to 16, no improvement

23:14 <tpw_rules> fwiw u-boot is very single threaded

23:16 <Flole> Uhm, maybe a different approach: Can I set $something in the pxelinux.cfg to use the new wget feature and the tcp stack to download the uImage and kernel? That should at least speed this entire stuff up once the config is transmitted through tftp

23:26 <Flole> Just checked that, not supported (yet) apparently

23:34 <Flole> I figured out which commit broke booting on my odroid c2 completely, it's a9bf024b2933bba0e23038892970a18b72dfaeb4

23:35 <Flole> Interestingly I am not even using EFI, but had it enabled. If I disable it then I can boot even after that commit

23:35 <Flole> with "enabled" I mean "selected in make menuconfig". I de-selected it there and now it works even after that commit