Nick_Lowe has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<philipp64>
hi all… the x.509 changes to strongswan are coming along and I have a connection up, according to swanctl. but… I’m not sure I’ve configured the default zone or the firewall rules appropriately to test this correct. can someone look at this with me? thanks
black_ant has joined #openwrt-devel
black_ant has joined #openwrt-devel
<stintel>
* pkg_hash_fetch_best_installation_candidate: Packages for libftdi found, but incompatible with the architectures configured
gch9812138 has quit [Read error: Connection reset by peer]
gch9812138 has joined #openwrt-devel
<owrt-snap-builds>
build #102 of rtl838x/generic is complete: Failure [failed checkarch] Build details are at http://buildbot.openwrt.org/master/images/builders/rtl838x%2Fgeneric/builds/102 blamelist: Rosen Penev <rosenp@gmail.com>, John Crispin <john@phrozen.org>, Rui Salvaterra <rsalvaterra@gmail.com>, John Audia <graysky@archlinux.us>, Kevin Darbyshire-Bryant <ldir
<owrt-snap-builds>
@darbyshire-bryant.me.uk>, Sander Vanheule <sander@svanheule.net>, Josef Schlehofer <pepe.schlehofer@gmail.com>
<stintel>
great. apu2 image doesn't boot anymore and apparently I don't have the required cable to use its serial console
<stintel>
just fucking great
<dangole>
i start to regret that this fast-forward switch to wolfssl happened: transmission-daemon segfaults weirdly with it's cyassl crypto (from 2005), while mbedtls and openssl still work great. and thanks to libcurl there is no easy way to switch ssl library.
bookworm has quit [Read error: Connection reset by peer]
<dangole>
imho libcurl should have goddamn variants for each crypto library. leave alone gnutls, for that we got gnurl alredy
<stintel>
almost 6AM here and still bisecting this shit
<stintel>
am supposed to work in ~4h but I took the day off as I've been bisecting this for hours
<stintel>
I am going to bet this is the problem:
<stintel>
- config restore -
<stintel>
/etc/preinit: line 11: can't open /rom/etc/shadow: no such file
<stintel>
/etc/preinit: line 11: can't open /rom/etc/group: no such file
<stintel>
/etc/preinit: line 11: can't open /rom/etc/passwd: no such file
<stintel>
[ 9.617724] urandom-seed: Seed file not found (/etc/urandom.seed)
<stintel>
aka the problem I warned people about
<stintel>
probably only tested on squashfs
<stintel>
5 hours of bisecting for nothing
<aparcar[m]>
stintel: which commit introduced the problem?
<stintel>
I was unable to find out with bisect
<stintel>
because I wrote image to USB stick and booted from that
<stintel>
which doesn't test sysupgrade
<stintel>
so 99% sure it's related to the "switch core services to another user than root"
<aparcar[m]>
oh I see
mangix has joined #openwrt-devel
mangix has quit [Read error: Connection reset by peer]
mangix has joined #openwrt-devel
<aparcar[m]>
stintel: what is the pro for ext4 vs squsahfs? Isn't squash better anyway because ext4 doesn't really do sysupgrades?
<stintel>
I'm using an APU2 with 16GB SSD, a real filesystem seems more suitable
<stintel>
nonetheless, we support ext4, so we should break it
<stintel>
ok so booting the device with OpenWrt on USB, then mounting the /boot of the SSD and runngin "tar xcf /boot/sysupgrade.tgz /etc/passwd" and then rebooting fixed it
<stintel>
definitely related to switching ubusd from root to ubusd
<stintel>
whoever the fuck caused it, please test it and fucking fix it
<stintel>
lost a day of pay because of your fucking untested commit
<mangix>
first time running btrfs scrub. I'm surprised nothing has blown up
<stintel>
btrfs problems are often caused by buggy drive
<stintel>
using btrfs as main fs on all but embedded for almost 10 years
<stintel>
it's much more stable than openwrt, even though that's a weird comparison
black_ant has quit [Ping timeout: 256 seconds]
<mangix>
stintel: I'm running it on a device that has had probably over 100 kernel freezes
<stintel>
mangix: probably unrelated to btrfs
<mangix>
because of someone insistance on carrying a buggy patchset
<mangix>
*insisting
<mangix>
no I mean I expect btrfs errors after all these kernel freezes
<mangix>
none so far
<stintel>
it's 2020, for anything !embedded, there is btrfs, bcachefs or zfs, in order of preference
<stintel>
anything else makes no sense
<stintel>
silent corruption is real. filesystems that don't detect it are obsolete
<mangix>
well, the device I'm talking about is an mvebu device. It's somewhat embeddewd.
<stintel>
different story. I'd love to see btrfs becoming a first class citizen in openwrt, though
<stintel>
imagine sysupgrade to a new subvolume, boot fails, revert to previous subvolume
<stintel>
"unbreakableWrt"
<mangix>
wasn't there a proposal to replace squashfs + overlay to btrfs on the mailing list?
<stintel>
I suggested it once, somewhere
<mangix>
there's one really really big problem with that though
<stintel>
not relevant for small flash devices though
<mangix>
btrfs testing on big endian is a joke
<stintel>
mangix: we could help with that
<mangix>
last I tested 4.19 on OpenWrt, I created a volume with mkfs.btrfs and it got corrupted within 8 seconds
<stintel>
I remember submitting a patch for bird, and upstream being amazed about me running it on mips64
root has joined #openwrt-devel
<stintel>
we just need to help eachother out
root has quit [Client Quit]
<mangix>
I know it's within 8 seconds as it immediately remounts ro
<mangix>
anyway, I tried reporting upstream. Didn't get too far. I needed to give more info. I wasn't familiar enough with btrfs at the time.
girth has joined #openwrt-devel
<mangix>
stintel: you have a mips64 device?
<stintel>
mangix: ERL
<mangix>
ah is that that ubiquiti device?
<stintel>
I have 2, one is in my "testing" vlan
<stintel>
yes, edgerouter ltie
<stintel>
lite*
<stintel>
apparently I should have gotten an extra apu2 too
<stintel>
could have saved all of us from my rant earlier
<mangix>
sounds like work is needed to get btrfs in here
<stintel>
but the fact that I warned about the hackiness of the patch to add new users/groups during sysupgrade, and the fact it still broke ...
<mangix>
that way you could just revert a subvolume
<mangix>
the turris people do this actually
<stintel>
yes, no boot within ~180s? revert to previous subvol/snapshot and be fucking done with it
<mangix>
the turris omnia is actually more flexible than that. the reset button actually has 6 modes
<mangix>
boot to last snapshot, boot to first snapshot, etc...
<stintel>
could be but I'm not a turris fan
<stintel>
they should aim to contribute more instead of maintaining their own fork
<mangix>
well, they don't anymore for newer releases
<mangix>
they still maintain their fork for their legacy branch
<mangix>
but stable is based on 18.06
<stintel>
it's just stupid
<stintel>
get your shit upstream (with us) and we'll maintain it for ya (more or less)
<mangix>
yeah, they've been doing that
<stintel>
that's not my experience
<mangix>
I think for the DSA patch they want it upstream at kernel.org first
<stintel>
but it's easier to maintain a "fork" with shitty code and/or history and don't contribute back, than to address comments and fix it so it can be accepted
<stintel>
oh great, all my IoT stuff is offline because of the fucking outage
<stintel>
don't know what I find worse, that someone reported my issue 14d ago, or that a "lead" dev decided to degrade ext4 images to "development tool" status
<mangix>
the latter
<stintel>
my comment agrees
<stintel>
we need to fix this somehow
<mangix>
Amusingly I've used ext4 on an Archer C7v2 before
<stintel>
can we somehow hardcode users that are read during boot, so every service is able to start\
<mangix>
I want to say it booted faster. I don't remember
<stintel>
and in case those users don't exist in the "real system" we keep running with the old UID
<stintel>
and inform the user
mangix has quit [Quit: leaving]
<stintel>
it's not ideal but 10000000x better than leaving system unbootable
mangix has joined #openwrt-devel
<mangix>
alright, kicked garbage cpufreq patches to the curb
<stintel>
anyone good at implementing automated tests? I will pay someone to prevent the issue I just had from happening
<stintel>
I rely on OpenWrt for my day to day tasks
<mangix>
stintel: git history shows the turris people having 15 commits
<mangix>
at least it's something :\
<stintel>
yeah right
<stintel>
sorry, if you're not trying to upstream, I don't like you
<mangix>
that was obtained with git log --pretty=%h»¦«%ae»¦«%s»¦«%aD | grep -c nic.cz
<mangix>
ah looks like that's not completely accurate. some of them use their gmail addresses
<stintel>
mangix: compare that to commits they have in their fork
<stintel>
it's just not acceptable
<ynezz>
why not? it's FOSS
* ynezz
hides
<stintel>
doesn't even matter. omnia support wasnt really pushed by them iirc, more a community effort
<mangix>
true
<mangix>
Turris MOX support is also non-existent
<stintel>
ynezz: maintaining a fork without PR to upstream causes extra effort for both parties
<stintel>
it should be discouraged
dorf has joined #openwrt-devel
<ynezz>
it's hard if your managers don't understand upstream first
<mangix>
found 17 more email addresses from the turris people
<stintel>
they
<mangix>
so 32 commits
<mangix>
*17 more commits
<stintel>
they'll get it eventually if everyone resigns
goliath has quit [Quit: SIGSEGV]
<ynezz>
for me it's a goverment like entity then a company
<mangix>
stintel: new patch on mailing list. looks related to your troubles
<mangix>
ping rmilecki
<mangix>
whoops, I have that backwards
Fishman has quit [Excess Flood]
Fishman has joined #openwrt-devel
feriman1 has joined #openwrt-devel
gch9812138 has quit [Quit: Ping timeout (120 seconds)]
gch9812138 has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
ivanich has joined #openwrt-devel
feriman1 has quit [Ping timeout: 240 seconds]
feriman1 has joined #openwrt-devel
<aparcar[m]>
wait so how do we get those turris people to contribute more?
<aparcar[m]>
isn't bkpepe part of them?
<mangix>
he is
<dorf>
while you're leveraging, send round a hit squad to deal with NXP :)
<ynezz>
s/NXP/any vendor/
<dorf>
valid point :)
<ynezz>
and in my eayes NXP is at least trying
<mangix>
GL.iNet might be good on that front
<ynezz>
like the number of contributions?
<dorf>
wel NXP need to try a lot harder with the 88W8864
<ynezz>
that's already abandonware
<mangix>
yeah that's totally avabdoned
<mangix>
*abandoned
<dorf>
if they've abandoned it, the least they could do is open source the firmware blob. costs them nothing.
<ynezz>
as always, there are probably some IP strings attached
<aparcar[m]>
mangix: plenty of trustworthy people tested this omnia thing, I'd merge it
<aparcar[m]>
ynezz: what's the scope here, involve $vendors more into helping out at openwrt?
<dorf>
as I understand it, Marvell originally intended to upstream the firmware to Linux, but it never got very far. :|
<rmilecki>
mangix: pong
<mangix>
rmilecki: what is this libroxml thing?>
<rmilecki>
? XML parser
<mangix>
is it used by anything?
<rmilecki>
it used to be used by ucwmp
<rmilecki>
by nbd
<ynezz>
aparcar[m]: no scope, just generic morning rants
<aparcar[m]>
:)
<mangix>
rmilecki: so it's safe to remove?
<ynezz>
move, pls
<ynezz>
removing is too harsh and not acceptable I would say
<rmilecki>
i'd rather move it to packages, I have some software using i
<mangix>
ok
rmilecki has quit [Quit: Konversation terminated!]
<rsalvaterra>
aparcar[m]: No, airtime policies must be manually configured, of course.
<rsalvaterra>
'morning!
<aparcar[m]>
rsalvaterra: evening
<rsalvaterra>
Almost night, I wager. ;)
<aparcar[m]>
rsalvaterra: I hoped there would be some automatic magic
<rsalvaterra>
Sorry, no magic here, only tohojo's heavy science. :)
<aparcar[m]>
sounds like we want some luci magic on setting up a second wifi
<rsalvaterra>
Possibly, but I don't use LuCI at all.
<mangix>
new ksmbd version released
<mangix>
wonder why they added kerberos
<rsalvaterra>
I'm more worried about the last two mvebu commits… I hope they didn't break sysupgrade on my CZ11NIC13. (I'm still playing with the Redmi AC2100, so I won't use the Omnia for the next few days/weeks.)
<rsalvaterra>
At the time we supported Linux 4.19, but we've since moved all targets to 5.4, I believe.
<mangix>
yep
<rsalvaterra>
In that case, it could be merged as-is.
<rsalvaterra>
Ah, no, we need to drop the @ge5.1.
feriman1 has joined #openwrt-devel
<ynezz>
aparcar[m]: well, if it works and doesn't break things it's fine with me
<ynezz>
I didn't reviewed the implementation
<aparcar[m]>
rsalvaterra: I'm getting mixed messages here. What is your course of action?
<mangix>
that first change is questionable
<mangix>
the ge@5.1 part can be removed
nitroshift has joined #openwrt-devel
<rsalvaterra>
What first change?
<mangix>
on your linked patch
<rsalvaterra>
You mean adding the lzo-rle.ko?
feriman2 has joined #openwrt-devel
<mangix>
yes. there's a ge@5.1 part
<mangix>
it's not necessary anymore
<rsalvaterra>
I can explain the questionable change.
<mangix>
I know what it is, but since there's no kernel below 5.4, it's not necessary
feriman1 has quit [Read error: Connection reset by peer]
<aparcar[m]>
please send a proper v$(n +1) patch explaining it
<rsalvaterra>
Ah, I thought you were talking about putting lzo and lzo-rle in the same bag (which is due to an upstream silliness).
ivanich has quit [Quit: Konversation terminated!]
Nick_Lowe has joined #openwrt-devel
<rsalvaterra>
aparcar[m]: Should I? It's not my patch. :/
<mangix>
no. all these @KERNEL_X.X lines tend to get removed. there's also no point matching various kernel versions
<mangix>
unless you're backporting to 19.07 or something
<mangix>
huh, no more 3 series kernels are supported upstream. BrainSlayer must be sad.
<rsalvaterra>
Linux 3.x? What year is this? O_o
* rsalvaterra
notices Linux 3.18 running on his phone…
<mangix>
hahahaha
<mangix>
I remember he was pissed when 3.10 got dropped
<mangix>
because broadcom devices had some pci issue that he couldn't fix
<rsalvaterra>
The biggest problem of Broadcom devices is Broadcom itself. :P
grift_ has joined #openwrt-devel
<mangix>
last supported kernel is 4.4. that was a really good kernel.
grift has quit [Ping timeout: 260 seconds]
<mangix>
ramips hackery worked just fine with it.
<rsalvaterra>
4.4 has my first patch ever! :D
<rsalvaterra>
Which, by the way, was to fix an issue I noticed on my WDR3600. The lz4 algorithm was broken on big-endian and I really wanted to fix it.
<rsalvaterra>
I even got an old Power Mac G5, so I could test more easily. :)
Ycarus has joined #openwrt-devel
* mangix
hates big endian
* rsalvaterra
<3 big-endian <3
<mangix>
mostly because nobody tests for it
<aparcar[m]>
mangix: nobody should be replaced by ci
<rsalvaterra>
I love it exactly because of that.
<mangix>
btrfs last I tested was totally broken on big endian. ksmbd took hundreds of commits to get it working on big endian. I had to generate lots of tcpdump output. f2fs-tools was broken on big endian for a while
<mangix>
what a bug that was
ivanich has joined #openwrt-devel
<rsalvaterra>
That's a problem of kernel developers who assume too much.
<mangix>
the hillarious part about that is that they claim to test big endian (btrfs people when I reported the issue)
<rsalvaterra>
The biggest problem with btrfs isn't big-endian support, but that's another story. :P
<mangix>
i disagree
* rsalvaterra
is looking forward to bcachefs.
<rsalvaterra>
Btrfs ate my data too many times (aka once).
<mangix>
I have the opposite experience
<rsalvaterra>
When did you start using btrfs?
<mangix>
btrfs has survived multiple kernel freezes due to a totally broken mvebu patchset
<mangix>
a year ago
<mangix>
I'm running btrfs scrub. No errors
<mangix>
simply amazing
<rsalvaterra>
Surviving kernel freezes is not exactly a feat for a filesystem worth using.
<rsalvaterra>
Any decent journaling filesystem does that.
<rsalvaterra>
Well, I *stopped* using btrfs around 2014… :P
<mangix>
add to that that i'm running my array in RAID5 mode (aka totally dangerous even for btrfs)
black_ant has joined #openwrt-devel
black_ant has quit [Changing host]
black_ant has joined #openwrt-devel
<rsalvaterra>
mangix: I see you like living on the edge. Try disabling barriers too. :P
<mangix>
no :)
<rsalvaterra>
Aww… XD
<mangix>
the kernel freeze was from a totally broken mvebu patchset that implements dynamic frequency scaling
<mangix>
the problem is that I've had over 100 freezes with various different kernels
<mangix>
from 4.19 to 5.9
<rsalvaterra>
Ah, I remember actually having DFS on the Omnia, with the original TurrisOS (Linux 4.4). I enabled it manually (changed the governor to ondemand) and had no issues at all…
<mangix>
you would have if you tried running btrfs scrub
<mangix>
it only freezes on high I/O load
<rsalvaterra>
All kinds of I/O, or only block device I/O?
<rsalvaterra>
(My connection at the time was 100 Mb/s.)
<mangix>
definitely the latter. I've experienced it when running btrfs scrub and qbittorrent
<mangix>
connected to a btrfs array of 4 drives
<rsalvaterra>
Wait… you have a RAID-5… of spinning rust…? O_o
<rsalvaterra>
Tell me they're SSDs.
<mangix>
on mvebu? you have to be kidding me
<mangix>
SSDs would be faster than the interface
<rsalvaterra>
Speed is not the point…
<mangix>
and what would that be?
<rsalvaterra>
… it's data safety.
<mangix>
I don't know that SSDs are safer than spinning rust
<rsalvaterra>
The rebuild of a RAID-5 is extremely stressful for the whole matrix, especially if you have the bare minimum amount of drives (four).
<aparcar[m]>
mangix: mhh I need to find a way for reasonable CI testing this
victhor has joined #openwrt-devel
<f00b4r0>
aparcar[m]: ping
feriman2 has quit [Ping timeout: 272 seconds]
<mrkiko>
noltari: ping
dangole has joined #openwrt-devel
adrianschmutzler has joined #openwrt-devel
* enyc
meows
<enyc>
mangix: hrrm curious now, are things happening so OpenWRT 20 may actually use gcc-10 ?
<aparcar[m]>
f00b4r0: pong
<adrianschmutzler>
mangix: if you just move libroxml to package, and not just delete it, I would merge the patch to master
<adrianschmutzler>
like we did yesterday, so please add it and I will push to master later
feriman2 has joined #openwrt-devel
<f00b4r0>
aparcar[m]: I just replied to you on GH, I had given the gist of that reply to adrian earlier (it more colorful terms), but now it's logged for future reference ;)
<f00b4r0>
tl;dr: kernel2minor is devil's work, and shouldn't be touched with a 10ft pole.
<aparcar[m]>
great thank you
<aparcar[m]>
f00b4r0: that what they told about the openwrt build system and look at me know, cruising
<f00b4r0>
hehe
<rsalvaterra>
enyc: Don't count on it… GCC 10.x enables a much more aggressive inliner, which increases code size significantly. We'd need to evaluate if it's actually beneficial in terms of performace.
<f00b4r0>
well, if you can parse code commented in cyrillic, be my guest ;)
<aparcar[m]>
f00b4r0: okay that's good (and sad) feedback, thanks for taking the time
<f00b4r0>
aparcar[m]: while I don't like to be ad-hominem, I've come to the conclusion that any piece of code submitted directly or indirectly by "adron-s" should be seriously scrutinized
<f00b4r0>
aparcar[m]: he was also responsible for packing a near full copy of u-boot, laden with his own copyright, as an "intermediary loader" for ipq Mikrotik devices
<f00b4r0>
aparcar[m]: thankfully that nonsense has been removed thanks to the hard work of John Thompson (john-tho)
Acinonyx_ has quit [Ping timeout: 260 seconds]
<enyc>
rsalvaterra: i see, can those inline-options simply be turned-off ?
<aparcar[m]>
enyc: gcc 10 is not on the roadmap for 20.xx, I think the work will timely begin once the new release branch is merged (ideally mid december)
<rsalvaterra>
enyc: Yes, they can be turned off, fortunately.
<mrkiko>
enyc: what does the acronym meows stands for ? :D
<aparcar[m]>
f00b4r0: I don't know anything about this but I understand your concerns
<f00b4r0>
aparcar[m]: np, that's the point of the explanation :)
<rsalvaterra>
mangix: Having a Russian ex-wife, the cringe is strong in that one. :P
black_ant has joined #openwrt-devel
black_ant has quit [Changing host]
black_ant has joined #openwrt-devel
<rsalvaterra>
Also, what variant? There's Russian cyrillic, Ukrainian cyrillic, Bulgarian cyrillic, etc.… :P
<mangix>
All of them
<mangix>
and yes, you have a stroke reading aome of them.
<rsalvaterra>
Oh, well… I just sent the link to my ex-wife and I'm bracing for impact. :P
Night-Shade has joined #openwrt-devel
Night-Shade has quit [Client Quit]
<rsalvaterra>
I'm having a strange issue on some systems (mainly the Omnia and this Redmi AC2100, which are totally unrelated)… Sometimes, on sysupgrade, the machine just reboots, without upgrading. I have to repeat the process several times, until eventually it effectively does the upgrade. Has anyone noticed this?
Acinonyx_ has joined #openwrt-devel
Acinonyx has quit [Ping timeout: 260 seconds]
<adrianschmutzler>
rsalvaterra: I only have stuff like that on 4/32 devices, where I attributed it to insufficient RAM (without ever verifying ...)
<rsalvaterra>
adrianschmutzler: I wouldn't say the Omnia is RAM limited… :P
<rsalvaterra>
I also saw the patch dropping the caches before sysupgrade… I can't say I like it, it shouldn't be needed…
<adrianschmutzler>
that's why I never submitted it here
<adrianschmutzler>
but I'm using it very extensively downstream
<adrianschmutzler>
but I still cannot believe that it really stopped 100 % of the soft-bricks, just this little change.
<adrianschmutzler>
actually, that would be even more reason to look for the actual cause ;-)
<rsalvaterra>
That must be fixed by people who have forgotten more about memory management than I'll ever know in my entire life. :P
<adrianschmutzler>
I also remember somebody telling me that the change would be completely useless and it's not expected to help at all ...
<rsalvaterra>
The mm subsystem is *extremely* conservative, since it impacts basically everything where Linux runs.
<rsalvaterra>
adrianschmutzler: I would have told you that at the time, if you asked me.
<rsalvaterra>
It doesn't make sense. But if it works, it works.
<rsalvaterra>
Anyway, I believe it's safe to assume the issue I'm seeing is unrelated.
<adrianschmutzler>
I'm glad I tried it anyway ... :-)
<adrianschmutzler>
it's not so fun if you have 3-5 % soft-brick on upgrades ...
<rsalvaterra>
I wonder if playing with swappiness and overcommit settings will have any effect…
<karlp>
I have lots of update failures if I don't reboot first.
<karlp>
normally it's "try upgrade, reboots, but reboots into same image as before, try again, works fine"
<rsalvaterra>
karlp: Exactly what I see!
<karlp>
I completely understand that it _shouldn't_ be useful, but sometthing's clearly not working properly somewhere
<adrianschmutzler>
karlp: I still have this kind with 4/32
<karlp>
I've never tried this dropping caches thing, but *sahrugs*
<rsalvaterra>
But sometimes it just reboots more than once. Yesterday it was like 3-4 times, on the Redmi(!).
<adrianschmutzler>
but it's just booting the old and not breaking anymore
<karlp>
well, I'm, on 16/64, and it still happens...
<adrianschmutzler>
before the cache "fix", part of these never came back up
<adrianschmutzler>
and actually what most people did was rebooting every device before any upgrade
<karlp>
I've never had that at least, but I do carry a watchdog patch on ath79.
<adrianschmutzler>
that was downstream, actually probably already at 18.06 (so ar71xx)
<adrianschmutzler>
where I originally dealt with the cache thing
<adrianschmutzler>
but still, I don't think this is really connected to rsalvaterra's current issues
rmilecki has quit [Quit: Konversation terminated!]
<ynezz>
I think, that this explicit dropping of caches works because there is some race somewhere on unhandled error return (retry one more time after error)
<rsalvaterra>
ynezz: I also suspect my issue is caused by some sort of race condition.
urjaman has quit [Read error: Connection reset by peer]
urjaman has joined #openwrt-devel
<rsalvaterra>
The problem is, I'm not exactly using the serial console all the time, to see what's happening at sysupgrade… :P
nitroshift has quit [Quit: Gone that way --->]
<karlp>
ther'e s nothing useful in my experience
<ynezz>
sometimes you see more with console
<ynezz>
but indeed, it would be nice to have all the output in the syslog
<ynezz>
so the serial console wouldn't be necessary
<rsalvaterra>
My configuration is not exactly vanilla, of course… but I'm sure all developers have their own patches/hacks haven't been/can't be upstreamed.
black_ant has quit [Quit: simplicity does not kill]
linzst has joined #openwrt-devel
black_ant has joined #openwrt-devel
black_ant has quit [Changing host]
black_ant has joined #openwrt-devel
dangole has quit [Remote host closed the connection]
black_ant has quit [Quit: simplicity does not kill]
black_ant has quit [Quit: simplicity does not kill]
<russell-->
i have some devices where i have to explicitly stop snmpd before a sysupgrade and maybe a few other processes to avoid OOM'ing
<russell-->
i wonder if procd is restarting them after sysupgrade tries to TERM and KILL them
dangole has joined #openwrt-devel
<rsalvaterra>
russell--: And maybe rmmod(ing) all non-vital modules isn't entirely unreasonable…
Net147 has quit [Read error: Connection reset by peer]
black_ant has joined #openwrt-devel
black_ant has joined #openwrt-devel
<jow>
russell--: the TERM/KILL loop runs in stage2 at which point procd should've exec'd itself into upgraded
<jow>
however I was talking with blogic the other day that we need to implement a proper "init 1" equivalent in procd (that is, gracefully shutting down all servies as if /etc/init.d/$service stop has been called, respecting settings such as term grace period etc.)
Net147 has joined #openwrt-devel
<jow>
afair he has took a look at a possible implementation and it is only a few dozen lines or so
<rsalvaterra>
Ooh, runlevels for procd, that's so sysvinit! :D
<jow>
once this is done we can get rid of the TERM/KILL shell loop
<jow>
next todo item would be invoking /sbin/block to handle umounts
<jow>
swapoff etc.
<rsalvaterra>
jow: I'm trying to remove swapon/swapoff from block… :|
<rsalvaterra>
(Only the applets, not the hotplug functions.)
<rsalvaterra>
The BusyBox implementations are more featured.
<dangole>
talking about hotplug-call: it's kinda be nice to have hotplug calls trigger though ubus, so we can have acl for them (as ntpd and probably other services are using them and making that work non-root is tricky otherwise)
<rsalvaterra>
(By "trying" I mean "I have patches pending review".)
black_ant has quit [Quit: simplicity does not kill]
Net147 has quit [Read error: Connection reset by peer]
<mrkiko>
rsalvaterra: why is it so brittle?
<mrkiko>
rsalvaterra: wasn't expecting it to be
<rsalvaterra>
mrkiko: You mean ubifs zstd compression? Because the filesystems of overlay mounts are treated as if they all supported the same mount options.
Net147 has joined #openwrt-devel
<rsalvaterra>
E.g.: If your overlay is ext4 and you select zstd compression in fstools, it will merrily pass compr=zstd as a mount option.
<rsalvaterra>
In other words, it's not user-proof (heck, not even developer-proof, I'd say).
dangole has quit [Remote host closed the connection]
<mrkiko>
rsalvaterra: ok, common options to all FSs, while you didn't want that. so some logic rework needs to be done
<mrkiko>
someone here may help with BCM63xx hardware?
<mrkiko>
I am awaiting noltari :D
dangole has joined #openwrt-devel
linzst has joined #openwrt-devel
<mrkiko>
and any help is really apreciated.
<rsalvaterra>
mrkiko: Yeah. And I haven't got to it, since I'm lazy and it works for me. :P
Nick_Lowe has quit [Ping timeout: 264 seconds]
<mrkiko>
rsalvaterra: fair enough :D
<rsalvaterra>
Also, if you plan to use any of my trees, be mindful that I'm constantly rebasing them.
dangole has quit [Remote host closed the connection]
black_ant has quit [Quit: simplicity does not kill]
black_ant has joined #openwrt-devel
black_ant has quit [Changing host]
black_ant has joined #openwrt-devel
feriman has joined #openwrt-devel
feriman has quit [Ping timeout: 256 seconds]
<mrkiko>
rsalvaterra: :D
valku has joined #openwrt-devel
pgjermshus_ has quit [Ping timeout: 272 seconds]
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
dorf has quit [Quit: ... gone ...]
kakaka has quit [Remote host closed the connection]
kakaka has joined #openwrt-devel
dorf has joined #openwrt-devel
pgjermshus_ has joined #openwrt-devel
goliath has joined #openwrt-devel
pgjermshus_ has quit [Ping timeout: 246 seconds]
feriman has joined #openwrt-devel
<rsalvaterra>
Rats. Seems like somehow I can't get airtime policy working…?
* dorf
confesses responsibility for said pull request.
pgjermshus_ has joined #openwrt-devel
<rsalvaterra>
dorf: 3,534 changes, 1,860 additions 1,674 deletions. In one patch?
<dorf>
in one patch :)
<rsalvaterra>
I don't use LuCI, but there's no way in hell I'd review that.
<dorf>
it's all css, shouldn't be that controversial :)
<dorf>
I mean, the easiest way to review it is just to deploy it, for an initial assessment.
linzst has quit [Quit: Leaving]
<dorf>
and a huge chunk of those changes are whitespace, so it's not as big as it looks.
<rsalvaterra>
Precisely. That's totally unrelated to code.
<dorf>
on a parallel note, the css and markup is a lot more complex than it needs to be in LuCI. why use flexbox when all you're doing is implementing tables?
<dorf>
sure, I could refactor the patch and make it easier to diff with upstream, and smaller. not a major issue. but no sense in doing that if the general consensus is "no, we don't like it".
<rsalvaterra>
I can only comment on the process (as I did), since I don't do porcelain, only plumbing. :)
<dorf>
:)
dorf has quit [Remote host closed the connection]
dorf has joined #openwrt-devel
pgjermshus_ has quit [Remote host closed the connection]
pgjermshus_ has joined #openwrt-devel
MichaelOF has joined #openwrt-devel
<philipp64>
Hi all. I’ve got an IPsec tunnel up but it’s not passing traffic. What are the tools/procedures for figuring out what’s happening to the traffic?
<jow>
dorf: I took a brief look and the changes seem to major to me to consider them an iterative improvement to the bootstrap theme
<jow>
I'd be okay with adding them as a new theme assuming you're willing to maintain it
pgjermshus_ has quit [Ping timeout: 246 seconds]
dangole has joined #openwrt-devel
<Tapper>
dorf hi how do I test out your changes to luci?. I want to test it with my screen reader.
<dorf>
Tapper: you'd just take the cacade.css file and upload to /www/luci-static/bootstrap/ but there's not much there of note for screenreaders that isn't in the existing theme.
<Tapper>
OK thanks
<philipp64>
jow: what in the firewall and routing would cause packets to be NAT’d instead of going through ip xfrm’s? I’m doing a “ping -I br-lan 192.168.1.1” from my 2nd router (br-lan is 192.168.3.1/24) and the packets are going out with the public address of that router, no ESP encapsulation. Not sure if it’s because my “vpn” zone not being configured or what…
<dorf>
jow: that sounds like a good idea. sure, I'd be willing to maintain the theme.
<dorf>
I was being reasonably conservative with the changes, but a new theme offers more scope for divergence.
<Tapper>
dorf I have uploaded it to my router will the changes just take affect when I next log in to luci or do I have to restart?
<dorf>
dorf: just issue a ctrl+shift+refresh in your browser to effect changes.
<dorf>
*Tapper
<Tapper>
OK thanks
<dorf>
*CTRL+SHIFT+R usually does the trick.
<Tapper>
dorf the tables on the frunt page do not show up as tables to my screen reader
<Tapper>
in bootstrap with and with out your changes.
<dorf>
Tapper: that's because they're not actually tables.
<Tapper>
the tables I am on about are Active DHCP Leases and Active DHCPv6 Leases
<Tapper>
Active DHCPv6 Leases they are not?
<dorf>
that's one of the issues I was raising earlier.. the html markup over-complicates everything by implementing tables and their children as divs.
<Tapper>
lol coppy and paste fail!
<Tapper>
dorf they are not tables?
adrianschmutzler has quit [Quit: ~ Trillian - www.trillian.im ~]
<dorf>
no, they're divs that implement pseudo-tables using flexbox.
<Tapper>
dorf well your changes don't break anything for me.
gch9812138 has joined #openwrt-devel
<Tapper>
It would be nice if the tables were proper tables.
<dorf>
yes, that's what I'm indirectly suggesting :)
Night-Shade has joined #openwrt-devel
<dorf>
aside from reducing complexity, you'd have better compatibility with non-css3 browsers that way. as it stands, you have to use a css-capable browser to view the console as intended.
<dorf>
*css3-capable
<Tapper>
one more thing I would like to have fixt in luci is the fact that when dilogs show up the don't grab focus and let the screen reader know that somthing has popped up
<dorf>
yeah, the modal dialogs need work, I agree.
<dorf>
no reason why escape and enter can't work there, either. tab-index should help with the default action.
<philipp64>
Is there an example of setting up a ‘vpn’ zone in the firewall?
<Tapper>
EG in Wireless > Wireless Overview you click on the Edit button the options show up in a dialog but to get to it I have to jump all the way down to the bottom of the page and press enter to access any of them. If you didn't know that is how luci behaved then you would not know that there was a way to change your wifi settings.
<Tapper>
When using a screen reader that is!
dangole_ has joined #openwrt-devel
<Tapper>
dorf You could make a new theme and call it rebootstrap
dangole has quit [Ping timeout: 240 seconds]
<Tapper>
dorf I will help out with the testing for a11y. I am shit at spelling but good at moaning! lol
<jow>
philipp64: I don't know, sorry
<philipp64>
who would be a good person to ask? I’d like to adequately test my changes to strongswan...
<jow>
dorf: the tables-as-divs is done to be able to decompose the tables into cards on mobile
<jow>
dorf: it wasn't done to overcomplicate things for no reason
<dorf>
jow: yeah, except there's no need to do it that way.
<jow>
how would it be done otherwise?
<jow>
restyling tables did not work for me
<dorf>
you'd do the inverse of what you're doing.
<dorf>
change the display type for tables, trs, tds etc.
<jow>
that didn't work here
<jow>
might retry it some time
<dorf>
no reason why it shouldn't. if you need a hand, holler.
<dorf>
if you manage to do tables and shift the flexbox styling to the mobile.css stuff, you've won :)
<dorf>
with tables as the default layout method, you can probably expect to shrink the existing css by at least 50% at a guess.
<dorf>
enforce id's for <body> and you should have what you need to customize individual pages, where required.
Namidairo has quit [Ping timeout: 246 seconds]
<dorf>
jow: also, just read your comment on the patch. bold fonts! install open sans ttf and you'll see a much improved, multi-weight implementation.
<dorf>
and thanks for the feedback, generally.
<dorf>
rebootstrap, Tapper!
<dorf>
haha..
<dorf>
the truncated heading in modal dialogs is intentional.. not so much truncated as top-aligned.
dangole_ is now known as dangole
pgjermshus_ has joined #openwrt-devel
dedeckeh has quit [Ping timeout: 245 seconds]
pgjermshus has joined #openwrt-devel
pgjermshus_ has quit [Ping timeout: 240 seconds]
<jow>
dorf: s/\.(table|tr|th|td)/\1/g worked indeed. Reason for the failure might have been IE compat
<jow>
will migrate the markup back to tables but leave the class names in place for a few months to allow for smooth transition
matteo has quit [Quit: ZNC 1.8.2+deb1+b2 - https://znc.in]
matteo has joined #openwrt-devel
pgjermshus has quit [Ping timeout: 264 seconds]
<dangole>
ping olmari
<dorf>
jow: excellent!
<dorf>
thankfully we don't need to bother with IE anymore.
<olmari>
dangole: repeat what was needed, if possible from downloaded snapshot :)
<olmari>
will test sat
<dangole>
olmari: download snapshot, install procd-ujail and procd-seccomp (the latter only if available), reboot
<dangole>
olmari: on wdr4900
<dangole>
olmari: so if everything works fine, dnsmasq should come up with ujail, ntpd should be started as user 'ntp' and 'umdns' should also be with 'ujail' and have seccomp indicated in /proc/`pidof umdns`/status
<olmari>
yeh :)
<dangole>
olmari: install umdns as well, just to also have an example of a service using seccomp jail
black_ant has quit [Ping timeout: 256 seconds]
linzst has joined #openwrt-devel
slh64 has joined #openwrt-devel
MichaelOF has quit [Quit: Konversation terminated!]
dwmw2_gone has quit [Quit: Coyote finally caught me]
Pepe has joined #openwrt-devel
dwmw2_gone has joined #openwrt-devel
dorf has quit [Remote host closed the connection]
dorf has joined #openwrt-devel
<jow>
Tapper: I pushed the luci change to turn tables into actual tables
<Tapper>
jow O nice thanks mate.
philipp64 has quit [Quit: philipp64]
dangole has quit [Remote host closed the connection]
<mrkiko>
thanks guys for your accessibility work in LUCI - very very much apreciated
dangole has joined #openwrt-devel
ivanich has quit [Quit: Konversation terminated!]
pgjermshus has joined #openwrt-devel
<Tapper>
mrkiko Hi mate how are you?
pgjermshus has quit [Ping timeout: 272 seconds]
<Tapper>
jow my build failed for my wrt3200. lol I will do a make dclean and try again
Ycarus has quit [Quit: Ycarus]
philipp64 has joined #openwrt-devel
voxadam has quit [Read error: Connection reset by peer]
voxadam has joined #openwrt-devel
veonik has quit [Ping timeout: 256 seconds]
veonik has joined #openwrt-devel
zjason has quit [Ping timeout: 256 seconds]
agb has quit [Ping timeout: 256 seconds]
zjason has joined #openwrt-devel
goliath has quit [Ping timeout: 256 seconds]
OutBackDingo has quit [Ping timeout: 256 seconds]
goliath has joined #openwrt-devel
agb has joined #openwrt-devel
OutBackDingo has joined #openwrt-devel
Borromini has joined #openwrt-devel
<Tapper>
jow the tables work grate. thanks
danitool has joined #openwrt-devel
<owrt-snap-builds>
build #706 of at91/sam9x is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/at91%2Fsam9x/builds/706 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Hannu Nyman <hannu.nyman@iki.fi>, Petr ?tetiar <ynezz@true.cz>, Paul Spooren <mail@aparcar.org>, Andre Heider
<owrt-snap-builds>
build #735 of sunxi/cortexa53 is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/sunxi%2Fcortexa53/builds/735 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Hannu Nyman <hannu.nyman@iki.fi>, Petr ?tetiar <ynezz@true.cz>, Paul Spooren <mail@aparcar.org>, Andre
<owrt-snap-builds>
build #457 of bcm47xx/generic is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/bcm47xx%2Fgeneric/builds/457 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<owrt-snap-builds>
build #732 of ipq806x/generic is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/ipq806x%2Fgeneric/builds/732 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Hannu Nyman <hannu.nyman@iki.fi>, Petr ?tetiar <ynezz@true.cz>, Andre Heider <a.heider@gmail.com>, Imran
<owrt-snap-builds>
Khan <gururug@gmail.com>
<owrt-snap-builds>
build #713 of arc770/generic is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/arc770%2Fgeneric/builds/713 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<owrt-snap-builds>
build #532 of ath79/tiny is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/ath79%2Ftiny/builds/532 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<zorun>
looks like some build hosts have network issues
<owrt-snap-builds>
build #527 of ipq40xx/generic is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/ipq40xx%2Fgeneric/builds/527 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<owrt-snap-builds>
build #532 of sunxi/cortexa8 is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/sunxi%2Fcortexa8/builds/532 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<dorf>
jow: did you try that theme with open sans ttf installed?
<dorf>
(the font families selected all support multiple font-weights except the fallback sans-serif)
<owrt-snap-builds>
build #582 of x86/legacy is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/x86%2Flegacy/builds/582 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<owrt-snap-builds>
build #547 of mxs/generic is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/mxs%2Fgeneric/builds/547 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<owrt-snap-builds>
build #615 of mvebu/cortexa9 is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/mvebu%2Fcortexa9/builds/615 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<owrt-snap-builds>
build #346 of ath79/mikrotik is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/ath79%2Fmikrotik/builds/346 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<owrt-snap-builds>
build #563 of octeon/generic is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/octeon%2Fgeneric/builds/563 blamelist: Klaus Kudielka <klaus.kudielka@gmail.com>, Rosen Penev <rosenp@gmail.com>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>, Petr ?tetiar <ynezz@true.cz>
<zorun>
I'm disabling the affected builders
<owrt-snap-builds>
build #396 of bcm47xx/mips74k is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/bcm47xx%2Fmips74k/builds/396 blamelist: Rosen Penev <rosenp@gmail.com>, Petr ?tetiar <ynezz@true.cz>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>
<owrt-snap-builds>
build #614 of x86/generic is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/x86%2Fgeneric/builds/614 blamelist: Rosen Penev <rosenp@gmail.com>, Petr ?tetiar <ynezz@true.cz>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>
<owrt-snap-builds>
build #634 of at91/sama5 is complete: Failure [failed] Build details are at http://buildbot.openwrt.org/master/images/builders/at91%2Fsama5/builds/634 blamelist: Rosen Penev <rosenp@gmail.com>, Petr ?tetiar <ynezz@true.cz>, Imran Khan <gururug@gmail.com>, Andre Heider <a.heider@gmail.com>
<mangix>
didn't know people actually use ccache...
<Borromini>
don't plenty do so?
<jow>
dorf: nope, I am using what Debian ships by default, and I suppose many users will be in a similar situation. So the design should degrate a little bit more gracefully
<dorf>
jow: ok. apt install fonts-open-sans may have you covered. and you're right, it's just a question of finding default fonts that support multi-weight. ubuntu sans does, segoe ui does, and obviously open sans does. so we'd need a family for other linuxes and os x, probably.
<dorf>
(or we use google fonts as a fallback source, though I don't know what the policy is on pulling in external resources)
<jow>
the policy is "don't"
<dorf>
ok, noted.
<jow>
it would be a privacy violation and in many cases you can't assume connectivity
<dorf>
the problem such as it is exists because without multi-weight support, semibold, bold and extrabold weights will all render as plain bold.
linzst has quit [Quit: Leaving]
dorf has quit [Remote host closed the connection]
dorf has joined #openwrt-devel
<karlp>
mangix: ccache makes a big difference for people rebuilding a small set of changing apps into a new image against a largely static openwrt branch
Namidairo has joined #openwrt-devel
goliath has quit [Quit: SIGSEGV]
Borromini has quit [Ping timeout: 264 seconds]
gch98121384 has joined #openwrt-devel
gch9812138 has quit [Read error: Connection reset by peer]
gch98121384 is now known as gch9812138
<mangix>
karlp: I've seen multiple problems with it in the past.
<karlp>
I've heard of peopel say they ahve "problems" but nothign concrete
<karlp>
and aren't you always the one fixing esoteric corner problems anyway? :)
<mangix>
one example, bonnie++ freezes when compiling with ccache
<karlp>
I've been using it for years, and I don't think I've ever had anything that could ever be attributed to a ccache failure.
<karlp>
that sounds, obviously, like a highly broken package?
<mangix>
oh yes, the makefile is totally broken
<karlp>
what brings this up btw? I can't see any other refs to ccache in the recent logs?
<mangix>
wonder if I should just get rid of it
<karlp>
I thnk I built it once years ago to try out perf testing a gpio spi connected sd card, but *shrugs*