On 09/01/2026 12:05, Jishan Alam wrote:
Thank you for preparing and sharing the port patch. I’ve reviewed it
locally and can confirm that it builds and runs as expected. However,
since the upstream code is still unstable, I’d prefer to hold off on
committing it for now.
Yes, I agree, I did it just to be able to build it on our build machine.
It is not meant to be pushed to the ports tree now.
> Your fixed version of whowatch runs, but it has different UI from the
> old version 1.4, and the older version had colors for characters used
> for tree branches. As I understand, it has nothing to do with your utmpx
> fix, it was probably made by original author.
That matches my understanding as well. I haven’t used whowatch prior to
working on this update, so I don’t have historical context for the UI
changes beyond what’s present in the upstream code.
If you are interested I have access to one old machine with working
whowatch 1.4 so I can make a few screenshots with the old UI and send it
to you privately. I think there are a few details which were nicer in
the old version.
> One issue I found is that on one machine it does not display process
> [o]wner info, it shows -1 instead. I don't know why. On the other
> machine it shows the correct owner info. Botch machines are 14.3 amd64.
> On botch machines I tried to run it as normal user and as root.
Thanks for pointing this out. I’ll take a closer look at how the owner
information is parsed and will follow up once I have a better
understanding of what’s causing the discrepancy.
Do you think you will be able to make [d]etails also working on FreeBSD?
[d]etails and [s]ysinfo is new to this version, they were not there in
the version 1.4. But there was [c]md which turns On / Off showing long
command line options of the running processes or process name only. So
for shell scripts you see just "sh" or "/home/user/bin/myscript.sh -t -b
my-arg"
In parallel, I’ve been making progress on the sysinfo issue. It turns
out the existing implementation is largely Linux-specific, relying on /
proc/*, which doesn’t translate well to FreeBSD. To address this, I’m
redesigning the sysinfo layer to be more OS-neutral by introducing
abstract interfaces with OS-specific implementations. On Linux this will
continue to use /proc, while on FreeBSD it will rely on the sysctl API.
I’ve pushed the latest changes to my branch and will continue working
through the remaining sysinfo components. It’s a fairly broad change,
but now that the structure is in place, the rest should be straightforward.
I really appreciate your work on this!
Best regards
Miroslav Lachman
