ChanServ changed the topic of #haiku-3rdparty to: Third party devtalk for the Haiku® operating system | Main Haiku channel: #haiku | Bored? Tons of stuff to work on here: https://github.com/HaikuArchives | Logs: https://oftc.irclog.whitequark.org/haiku-3rdparty
<waddlesplash> NFSv2 or 4?
<waddlesplash> and nightlies or beta 5?
<waddlesplash> divad27182: there were a lot of NFS4 fixes since Beta5. please retest on a nightly build
<divad27182> how do I do that?
<divad27182> (it will take a while, my computer is down at the moment...)
<waddlesplash> you switch your system "haiku" repo in pkgman to be "master" instead of "r1beta5"
<waddlesplash> and then run "pkgman full-sync"
<waddlesplash> and then reboot
<divad27182> sounds do-able
<divad27182> is it reasonable that the r1beta5 haikuports would now be reporting "Bad data" when I try to do the pkgman add-repo ?
<divad27182> things went interesting...
<divad27182> waddlespash: I was able to make an empty file, and write to a file, create a directory (as my user).
<divad27182> then I tried to create a file as my user and it crashed again.
<divad27182> want a screenshot?
<divad27182> waddlesplash: sorry, got your name wrong...
<waddlesplash> divad27182: there's maintenance going on with haikuports so that error's not surprising
<divad27182> ok.
<divad27182> want the screenshot of the crash?
<waddlesplash> probably would be good yes
<waddlesplash> is https://dev.haiku-os.org/ticket/19656 still an issue or is that improved?
<waddlesplash> divad27182: if the screenshot looks more or less like https://dev.haiku-os.org/ticket/10925 then probably don't bother
<divad27182> much shorter, and it screeen shot capture failed. I didn't get to 19656 testing
<divad27182> yet
<divad27182> and my machine is down again :-(
<divad27182> OK. got a good capture
<divad27182> it is a "PANIC: ASSERT FAILED
<divad27182> it appears the non-zero userid may be relevant.
<divad27182> I'm going to try to send it through irc....
<waddlesplash> probably won't work
<divad27182> if you aren't seeing it, probably not.
<divad27182> is there a good site to upload to?
<waddlesplash> is the message about "NFS4ERR_BAD_STATEID"?
<divad27182> PANIC: ASSERT FAILED (../haiku-git/src/add-ons/kernel/file_systems/nfs4/InodeRegular.cpp:366): cookie != __null || fOpenState != __null
<divad27182> Thread 367 "NFSv4 Work Queue" running on CPU 0
<waddlesplash> probably you should open a new ticket
<waddlesplash> I see some potentially related tickets but not sure
<waddlesplash> Jim906 has been doing some good work on the NFS4 driver lately
<waddlesplash> Hopefully he'll be able to investigate
<divad27182> Would this be classed as R1/development ?
<waddlesplash> yes
<waddlesplash> but don't worry too much about the ticket fields
<waddlesplash> a dev (usually me lol) will come by and fill in whatever's missing
<divad27182> bug 19657 entered
<divad27182> is there anything that might be useful to try in KDL ?
<divad27182> or any questions I can answer?
<waddlesplash> I am not too knowledgeable about the inner workings of the NFS4 driver, so I don't know
<divad27182> doing a continue got me a second ASSERT FAILED.
<divad27182> reboot time I guess
<waddlesplash> yeah, ASSERTs are continuable, in theory
<waddlesplash> divad27182: if you really need reliable NFS, there is a package in the depots for a FUSE version I think
<waddlesplash> not as fast as the native one but more reliable I think
<waddlesplash> may require some fiddling to get it going though, our FUSE is not quite as smooth to configure as Linux FUSE
<divad27182> I think I saw a FUSE over ssh....
<divad27182> I didn't notice one over NFS
<divad27182> I can work with a local build directory and NFS read to get my source.
<divad27182> that was what I was trying when i reported 19656
<waddlesplash> the two issues mentioned in the commit I think have since been solved on the Haiku end
<waddlesplash> so it may be able to be enabled now
<waddlesplash> augiedoggie sometimes hangs out in #haiku, I'll see about asking him if he shows up one of these days
<divad27182> 19656 still happens
<waddlesplash> ah well
<divad27182> at this point, I'm testing by appending a short line to a short text file on the linux server, and then doing od -c on the file from haiku
<divad27182> haiku gets the changed length, but reads NUL bytes
<waddlesplash> yeah, the NFS driver really needs work
<waddlesplash> the codebase is very clean, it just has bugs
<divad27182> well, I would not be surprised if this happened in other network filesystems....
<divad27182> I think it involves a caching layer that is probably fairly generic
<divad27182> and very appropriate for reading a local unshared disk.
<waddlesplash> not really how Haiku's caching system works
<waddlesplash> at present anyway it's fully controlled by the filesystem itself
<divad27182> hmm.... then I guess it would be NFS screwing up.
<waddlesplash> the NFS4 driver that's native to Haiku is probably the issue
<waddlesplash> use a different driver... or sshfs or something
the_sea_peoples has quit []
the_sea_peoples has joined #haiku-3rdparty
the_sea_peoples has quit []
the_sea_peoples has joined #haiku-3rdparty
the_sea_peoples has quit []
jmairboeck has joined #haiku-3rdparty
Retropikzel has left #haiku-3rdparty [#haiku-3rdparty]
KitsunePrefecture has joined #haiku-3rdparty
KitsunePrefecture has quit []
<divad27182> Just found (and ticketed) that "mount -ro -t nfs4 -p ..." ignores the -ro
jmairboeck has quit [Quit: Konversation terminated!]