Le 2023-02-22 13:33, Stuart Henderson a écrit : > Problem is that we don't know if other codepaths also result in using > old system calls - this avoids one known case but there may still be > others. Crashing at start is possibly not so bad; starting up ok and > crashing later could do much more damage. Especially in an area around > database files I'm getting a bit twitchy. There are a lot of syscalls > showing up in > go/pkg/mod/modernc.org/sqlite@v1.20.4/lib/sqlite_openbsd_*.go and IIUC > any using off_t will be in the set which was changed. > > Really my best suggestion at the moment is to run on 7.2 or a kernel > with those syscalls reinstated, but calling syscalls directly at all > is at odds with some of the possible future directions to avoid #3 in > https://marc.info/?l=openbsd-tech&m=167702364728145&w=2 (though quite > common in some ecosystems - in particular on Linux glibc has sometimes > been slow to add wrappers for syscalls in new kernels which downstream > software wishes to use, so not really surprising that people who have > learnt to code that way also do so when dealing with other OS). > > > -- > Sent from a phone, apologies for poor formatting. > > On 22 February 2023 11:47:36 Hukadan <m...@hukadan.org> wrote: >
Hi @ports, The error still occurs with the latest snapshot, and the changes in [1] make me think the problem is still present in the latest version upstream. Should gotosocial be marked as broken for the coming release? Regards, H. [1] https://gitlab.com/cznic/sqlite/-/commits/v1.21.0/lib/sqlite_openbsd_amd64.go