IRC channel logs

<azeem>are the process selection differences between "ps a", or maybe rather between "ps ax" and "ps ex" compared to Linux intentional, or is that an implementation-specific corner-case based on what processes are considered to be attached to a terminal?

<azeem>or said otherwise, on Linux, I get the same output for "ps eux | grep ^$USER | wc -l" and "ps aux | grep ^$USER | wc -l", while on the Hurd, the former also shows processes not attached to the terminal

<azeem> * PS_USE_CLOBBER_ARGV

<azeem> * write over the argv and environment area

<azeem> * (Linux and most SysV-like systems)

<azeem>#elif defined(__linux__) || defined(__sun) || defined(__darwin__)

<azeem>#define PS_USE_CLOBBER_ARGV

<azeem>that should be __GNU__, right?

<youpi>yes

<youpi>but I don't think it will work like that for GNU/Hurd :)

<azeem>well, a trivial program worked

<youpi>oh?

<youpi>I'm really surprised since that'd be proc that remembers that

<azeem>hrm

<youpi>I mean, the process name change

<youpi>not the __GNU__ macro name

<azeem>let me test again then

<azeem> https://paste.debian.net/1393346/

<azeem>also,

<azeem>demo@debian:~/build-farm-19.1$ cat /proc/23815/cmdline

<azeem>A/process_title_testdemo@debian

<azeem>it seems to work

<azeem>I'll whip up a Postgres patch, this is about Postgres not being able to change the process title to something informative on the Hurd, should be easy to test

<youpi>I'm surprised that it works :)

<youpi>aaah, p_argv is indeed the address inside the process

<youpi>so indeed S_proc_getprocargs returns the current argv[] of the process

<youpi>fun

<azeem> demo 8276 8031 - 0:00.01 postgres: buildfarm regression [local] CREATE DATABASE

<azeem>nice

<youpi>nice :)

<azeem>but remind me, __GNU__ is the GNU system, so would have to be in addition to __linux__ or else I'd drop that right?

<youpi>__GNU__ is only GNU/Hurd

<azeem>aye

<azeem>I think process handling is different enough that they shouldn't be lumped together

<azeem>now that one can see what the processes are doing, it looks like the checkpointer takes up much more CPU times on the Hurd than (IME) on Linux:

<azeem> PID USER PR NI VIRT RES SHR S %CPU %MEM ZEIT+ BEFEHL

<azeem>11334 demo 20 0 372376 3084 0 R 90,1 0,1 0:43.55 postgres: checkpointer performing shutdown checkpoint

<azeem>not sure why that is

<youpi>possibly the checkpointer does various sync calls, which are more invasive in ext2fs

<azeem>oh yeah, but are those accounted for as cpu time?

<youpi>ah, no

<hwpplayer1>If I make a GNU , I mean GNU Hurd kernel system, can I name whatever I want or should I say GNU X Distribution ?

<hwpplayer1>What is the ideal way to call our distro ?

<identity>hwpplayer1: iirc ‘GNU Whatever’ is reserved for software that is part of the GNU project, you want something like ‘GNU/Whatever’ or just ‘Whatever’

<hwpplayer1>I see it is PSD(Procyberian Systems Distribution) by the way

<hwpplayer1>Should I mention the GNU and Hurd contributors in our documentation ?

<hwpplayer1>It will be ethical

<hwpplayer1>Can I add new codebase with AGPLv3-or-later ? To the core of Hurd ?

<hwpplayer1>I guess "yes"

<azeem>that doesn't look good, e2fsck has been going on for an hour now, I think I need to rebuild this again

IRC channel logs

2025-08-25.log