Linux Kernel Developers Fed Up With Ridiculous Bugs In Systemd
Phoronix: Linux Kernel Developers Fed Up With Ridiculous Bugs In Systemd
A patch was sent out today to the Linux kernel mailing list that would hide the "debug" string from showing up within the /proc/cmdline output. Why? To workaround a systemd bug. This has set off Linus Torvalds on another epic tirade...
It would have been funnier if this had happenend when debian was chosing systemd vs upstart. so much for omg systemd is just perfect.
No one, even its supporters, had ever claimed systemd was perfect. We just said it was better than the alternatives which apparently every distro out there other than Gentoo and Slackware agreed with. What I'm curious about is what is making systems crash from too much output. Buffer overflow maybe?
Originally Posted by Kiori
There is more than one Poettering?!
There is actually another person on this planet as abrasive as Poettering? Wow. And apparently even in the same development team. Wow again. And once again rudely ignoring users and other developers. Nothing new there
Fortunately he has found his match. What Linus wants, Linus gets Let's hope this is a humbling lesson for the systemd developers. Let's also hope their manager will step up and do something about this behaviour cause clearly these 5 year olds need some serious hand holding. And maybe a good spanking
Ah, the good old Phoronix way to dramatize non-issues...
If you followed the thread you'd know that :
- Greg KH submitted(or will do it) a patch to prevent systemd from reading the "debug" parameter (and will read "systemd.debug" instead)
- Linus (and other kernel devs) are trying to find a way to rate-limit the messages that could be sent to dmesg.
So not only the problem is almost resolved, but also the kernel is faulty for failing just because of a flooded dmesg.
I do however agree that Kay's behaviour in the bugtracker was really lame.
A tangent thought: /proc sure is great for scripting, but I'm still dumbfounded that native applications have to parse strings to get system information on Linux. Even the awful Windows API got it right here ... It's really awkward to see someone declare a character buffer or string, open a file, read the file into a buffer and then use <string> or <cstring> operations to get the information ... I mean, frankly, it's really stupid and the format of these files isn't always straightforward or even documented. Some of them are just space delimited numbers with no obvious meaning - I wouldn't even call that human-readable - at least a struct can have named members and comments.
Originally Posted by Scias
1) Greg shouldn't have to. Kay's attitude on the bugtracker is unacceptable and I hope someone on his team got in his face about it. Greg's babysitting, which Im thankful for, but still.
2) Kernel and systemd are both at fault it seems. Dmesg is getting flooded and just a tangent thought from: http://lists.freedesktop.org/archive...il/018378.html it kinda sounds like the socket that systemd provides before the Journal is set up is getting over-filled and there is apparently no code in place to handle that since it presumably was never expected. But this is just speculation.