Fedora 15 Lovelock Alpha Arrives

[phoronix]

Phoronix: Fedora 15 Lovelock Alpha Arrives

Dennis Gilmore has announced the official release of Fedora 15 Alpha. This next Fedora release, which is codenamed "Lovelock", brings a number of new features to this leading Linux distribution, including the GNOME 3.0 desktop...

http://www.phoronix.com/vr.php?view=OTE4Nw

[V!NCENT]

Can't wait for that SystemD goodness; nothing will ever crash on me anymore... ever...

(except the kernel)

[michal]

Can't wait for that SystemD goodness;

systemd, not SystemD

nothing will ever crash on me anymore... ever...

(except the kernel)

??

I'm not sure if I understand. I think that your idea about how systemd works and what it does maybe incorrect. Fortunately, there are available videos from presentations

http://0pointer.de/blog/projects/fosdem2011-video.html
http://0pointer.de/blog/projects/lca2011-video.html

[V!NCENT]

The answer to your indirect question/doubt is alctualy in the first video you listed around 20:00.

That makes we wonder if you even saw them (and remembered what was in them).

So for example we have the seemingly notorious KDE 4 series and D-bus and D-bus craches. What happens is that D-bus brings down the entire system (from an end-user perspective). Haters better be aware that this is what caused the nasty crashes...

But now we have systemd. systemd basically represents a collection of fake sockets that tracks processes with cgroups. Now D-bus crashes, systemd notices this, takes over with its fake D-bus socket, restarts D-bus, freezes up KDE whenever it reads out a sync, pass the socket back to the actual D-bus and everything resumes without the user even noticing.

Self-healing

[michal]

The answer to your indirect question/doubt is alctualy in the first video you listed around 20:00.

That makes we wonder if you even saw them (and remembered what was in them).

So for example we have the seemingly notorious KDE 4 series and D-bus and D-bus craches. What happens is that D-bus brings down the entire system (from an end-user perspective). Haters better be aware that this is what caused the nasty crashes...

But now we have systemd. systemd basically represents a collection of fake sockets that tracks processes with cgroups. Now D-bus crashes, systemd notices this, takes over with its fake D-bus socket, restarts D-bus, freezes up KDE whenever it reads out a sync, pass the socket back to the actual D-bus and everything resumes without the user even noticing.

Self-healing

Ok, actually yes in a certain sense it can work this way for dbus. I read your previous statement that systemd in some mysterious way stops applications from crashing, which of course is untrue.

You should just write that systemd can start service again. This only works for systemd services activated by socket.

[monraaf]

The answer to your indirect question/doubt is alctualy in the first video you listed around 20:00.

That makes we wonder if you even saw them (and remembered what was in them).

So for example we have the seemingly notorious KDE 4 series and D-bus and D-bus craches. What happens is that D-bus brings down the entire system (from an end-user perspective). Haters better be aware that this is what caused the nasty crashes...

I have never seen D-bus crash on gnome so it's probably just the Qt D-bus bindings that are buggy.

But now we have systemd. systemd basically represents a collection of fake sockets that tracks processes with cgroups. Now D-bus crashes, systemd notices this, takes over with its fake D-bus socket, restarts D-bus, freezes up KDE whenever it reads out a sync, pass the socket back to the actual D-bus and everything resumes without the user even noticing.

Self-healing

And because things usually crash for a reason (i.e. bugs) chances are it will crash again and before you know it you are stuck in an infinite loop of crashing-healing-crashing-healing...

[michal]

This only works for systemd services activated by socket.

Ok, it was not 100% true statement
http://0pointer.de/public/systemd-ma...md.socket.html

Services with Type=dbus are activated by dbus bus
http://0pointer.de/public/systemd-ma...d.service.html
(I did not write such service before, so I may be wrong here)

(Of course there are other ways to activate service, but here we are talking only about those which are directly activated by other applications)

[V!NCENT]

Ok, actually yes in a certain sense it can work this way for dbus. I read your previous statement that systemd in some mysterious way stops applications from crashing, which of course is untrue.

Lol no But I meant the OS. Apps are not part of the OS (or at least not by definition).

You should just write that systemd can start service again. This only works for systemd services activated by socket.

Yes. But then again what isn't using IPC? Only apps you can restart yourself by clicking an icon If not then it's either somewhere in the kernel.

This brings us to why the hell Linux can still nopt self-heal itself Even Windows does that nowadays (yes in kernel space).

[michal]

This brings us to why the hell Linux can still nopt self-heal itself

Maybe that could be possible to write services for desktop things like a task bar or other foo app to restart them if they crashes. Maybe it is, maybe not - I do not know. But I doubt that to happen anytime soon - so far there are systemd services only for a few apps - less than 10% services for Fedora.

[V!NCENT]

Maybe that could be possible to write services for desktop things like a task bar or other foo app to restart them if they crashes. Maybe it is, maybe not - I do not know. But I doubt that to happen anytime soon - so far there are systemd services only for a few apps - less than 10% services for Fedora.

I meant the kernel. Crashing the graphics driver (you can gues where that is ) in Windows will simply restart the driver and the rest of NT6.x (which is what Vista and 7 is build upon) is modular with defined interfaces which also restart on crashing (unless you run into some very fundamental stuff).

Even better would be to scrap the entire idea of IPC and use mobile code instead:
http://users.student.lth.se/cs07fh9/...d-mobility.pdf