IRC channel logs

***Server sets mode: +nt

<damo22>youpi: is this correct? The kernel IRQ handler should always disable the line and the userspace driver has to handle the interrupt and then acknowledge it and finally reenable the line

<damo22>or does it ack before handling

<damo22>maybe it doesnt matter because the line will still be disabled until the end

<youpi>the userland driver has to ack after setting anything that will trigger the next irq

<youpi>the concern is missing an irq wakeup if checking for device requests is not done properly

<youpi>simplest is to ack the irq after checking everything that had to be done with the device

<youpi>that's what drivers already have to do anyway

<janneke>i'm getting "locking protocol" error with guix and sqlite

<janneke>i found https://sources.debian.org/patches/sqlite3/3.31.1-5/20-hurd-locking-style.patch

<janneke>but trying to use that with guix, i now get "Error: unable to open database file"

<janneke>(the sqlite error messages are terrible -- they translate/intepret/guess what actually happens)

<janneke>so i'm wondering, could i be missing some kinde of file_lock patch?

<janneke>*kind

<janneke>okay, i can reproduce it by running two sqlite3 instances and saying

<janneke>"sqlite> pragma synchronous = normal;" in both

<janneke>ah crap -- sqlite3 on debian/hurd fails with the same error :-(

<janneke>i fear the sqlite3 patch is broken

<janneke>so...iwbn to test the initial (an earlier working version) of the patch and from that understanding fix the current version...

<janneke>hmm, it's either already broken in 3.8.7.1 (jessie; the oldest i could easily find),

<janneke>or "something changed" with file-locking support

IRC channel logs

2020-07-01.log