OpenBSD port no longer available: marked "broken" in tree

[TuggyNE]

Running pkg_add -v micro on a fresh install of OpenBSD 7.6 says Can't find micro. It turns out this is because it was marked broken due to something about syscall usage earlier this year. I'm not sure where to look for more information on that.

[niten94]

I do not know much about BSD systems and syscalls so I may be wrong somewhere, but I tried checking what can be done to not have the OpenBSD port marked as broken. There are few parts in micro module dependencies where syscall.Syscall* are called, but it is written in this part in OpenBSD 7.5 release page that syscall(2) is removed:

• Security improvements:
  • ...
  • Removed support for syscall(2), the "indirection system call," a dangerous alternative entry point for all system calls.
    Together with pinsyscalls(2) this change makes it impossible to perform system call through any other way than the libc system call wrapper functions.
    Users of syscall(2), such as Perl and the Go programming language were converted to use the libc functions.

The removal of syscall(2) is reported in golang/go#63900 and comments in #2442 but the latter issue states a different error. There only seems to be syscall.Syscall* indirect function calls with syscall.SYS_IOCTL passed in the main parts of micro, but syscall.Syscall seems to have been changed to call the libc stub of ioctl(2) in Go 1.23 at golang/go@8841f50.

There are no main parts where syscall.Syscall is called in modules, but syscall.Syscall6 is called in tscreen_bsd.go in the tcell fork used in micro. The libc stub is not called when using syscall.Syscall6 so I tried to test replacing calls of the function with syscall.Syscall and removing the last 3 arguments that are 0.

Micro was able to start up when I tested using an OpenBSD 7.6 VM, but I do not know why the terminal screen is blank and nothing happens even though it can be seen when running ktrace and kdump that read and write seems to be done on the terminal.

[niten94]

Sorry, I did not think there would be an email thread but this thread can be easily understood more. The situation was already figured out and there has been more attempted than I did: https://marc.info/?l=openbsd-ports&m=171418988805644&w=2

[niten94]

I haven't thought much about using my email address on public mailing lists so I do not know if I want to post email messages, but I was not able to find documentation in Go source code about the last argument being a pointer like said in the email thread. I think the cause of the bug is different but I explained it in the description of zyedidia/tcell#29.

There is a new version of micro but micro v2.0.13 (version distributed in port) seems to run properly on OpenBSD 7.6 when somehow changing go.mod and go.sum like in #3561 to update go-isatty and tcell. I do not think there will be a new release in a while, but the same can be done in the port using patches when zyedidia/tcell#29 is merged.

[niten94]

The port is still marked broken, but I made this issue close together with #3895 which I submitted and is now merged. The reason is that nightly builds should now work fine on OpenBSD, and did when I tested on OpenBSD 7.7 and 7.8.

I believe Micro's maintainers aren't assigned to patch the port and not expected to do so. However, I've sent a notification to the ports mailing list: https://marc.info/?l=openbsd-ports&m=176137734514024&w=2

2.0.14 ran on both (IIRC, or one of them) OpenBSD versions when patched like in my email message. Although currently latest OpenBSD versions seem fine, I forgot to address go-isatty there which wasn't updated yet in 2.0.14 and technically should be as well.

It's highly unlikely I will do it, but I may mention it if I will send a message again or patch on the list.