Robert Schweikert wrote:

On 01/29/2014 07:48 AM, Ludwig Nussel wrote:

...

openSUSE has a packing policy of requiring rc* symlinks for each init script. So for e.g. /etc/init.d/foo there must be a symlink /usb/sbin/rcfoo -> /etc/init.d/foo. There is an rpmlint check for this policy (suse-missing-rclink). [...] I don't care much for this "feature" but will certainly introduce the symlinks for my packages where appropriate if that's the direction we're going.

While you are at it, maybe you want to create a macro that people can just use

%{createRCLinks} foo bar twix

And the symlinks are magically created for the packager.

Interesting idea. I'll keep that in mind. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.de/ SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 16746 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Dominique Leuenberger a.k.a. Dimstar wrote:

Quoting Ludwig Nussel :

...

openSUSE has a packing policy of requiring rc* symlinks for each init script. So for e.g. /etc/init.d/foo there must be a symlink /usb/sbin/rcfoo -> /etc/init.d/foo. There is an rpmlint check for this policy (suse-missing-rclink).

A very much appreciated feature I would like to add!

...

Personally I'd like to see us keeping the rc* feature also in the future so I'd like to propose introducing an rpmlint check to complain about missing rc* symlinks also for systemd unit files.

Any objections on that? Otherwise I'd extend the suse-missing-rclink rpmlint check to .service files at some point.

No objection.. the sooner the better! I'd even propose a staging repo where the check does not WARN, but FAIL... so we get an overview where it's all missing.

A bit extreme :-) Failing is actually not needed, one can search for the error in build logs. AFAIK there is no interface in obs that allows to fetch all build logs from a project in one request though. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.de/ SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 16746 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Dominique Leuenberger a.k.a. Dimstar wrote:

Quoting Ludwig Nussel :

...

Dominique Leuenberger a.k.a. Dimstar wrote:

...

Quoting Ludwig Nussel :

...

openSUSE has a packing policy of requiring rc* symlinks for each init script. So for e.g. /etc/init.d/foo there must be a symlink /usb/sbin/rcfoo -> /etc/init.d/foo. There is an rpmlint check for this policy (suse-missing-rclink).

A very much appreciated feature I would like to add!

...

Personally I'd like to see us keeping the rc* feature also in the future so I'd like to propose introducing an rpmlint check to complain about missing rc* symlinks also for systemd unit files.

Any objections on that? Otherwise I'd extend the suse-missing-rclink rpmlint check to .service files at some point.

No objection.. the sooner the better! I'd even propose a staging repo where the check does not WARN, but FAIL... so we get an overview where it's all missing.

A bit extreme :-) Failing is actually not needed, one can search for the error in build logs. AFAIK there is no interface in obs that allows to fetch all build logs from a project in one request though.

Yes; that's why I proposed to do the fail only in a staging project... so we get at least visibility...

Fine if someone volunteers to take care of that staging project :-) Personally I can't. So I would be fine with just introducing the check and look at the build logs in a few months. The following packages trigger the old check today already: kexec-tools openhpi scsirastools iprutils hp-drive-guard lsyncd irqd xsp openscap systemd php5 xdmsc istgt aaa_base glusterfs nbd cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.de/ SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 16746 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

On Thu, Jan 30, 2014 at 03:31:14PM +0100, Christian Boltz wrote:

...

The following packages trigger the old check today already:

kexec-tools openhpi scsirastools iprutils hp-drive-guard lsyncd irqd xsp openscap systemd php5 xdmsc istgt aaa_base glusterfs nbd

I'm a bit surprised by some packages on the list - and it's funny to see systemd listed ;-)

systemd does not have any rc links which points to /etc/init.d Werner -- "Having a smoking section in a restaurant is like having a peeing section in a swimming pool." -- Edward Burr

В Thu, 30 Jan 2014 18:14:58 +0100 Christian Boltz пишет:

The udev package (which is part of the systemd sources) contains another initscript, /etc/init.d/boot.udev.

Why would you want to manually call script which is supposed to run exactly once at system initialization? Do rc links to other boot.* scripts exist? -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 2014-01-30 22:04, Christian Boltz wrote:

...

Do rc links to other boot.* scripts exist?

Yes, boot.apparmor / rcapparmor ;-)

Or rccrypto, previously. It does not exist now. It is not clear to me how it would be added with systemd.

That said - I know that most /etc/init.d/boot.* scripts should run only once as boot, which also means they don't need a rc* symlink.

True. Only some exist. IIRC, the rccrypto link was added after I suggested it on some bugzilla. - -- Cheers / Saludos, Carlos E. R. (from 12.3 x86_64 "Dartmouth" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.19 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iEYEARECAAYFAlLqwNcACgkQtTMYHG2NR9VXpACfVOEJRGZBnLo5xSXwPgw8Pzpz LmsAnR7Att0DAKg0McNU1a7RZNeGGLdD =5wli -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

El 30/01/14 18:04, Christian Boltz escribió:

Yes, boot.apparmor / rcapparmor ;-)

Patches for native apparmor profile support in systemd has been posted but not yet merged. Unit directive is called AppArmorProfile= This is to switch profiles but not to load them, that part is still missing. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Christian Boltz wrote:

Am Donnerstag, 30. Januar 2014 schrieb Ludwig Nussel:

...

Dominique Leuenberger a.k.a. Dimstar wrote:

...

Quoting Ludwig Nussel :

...

Dominique Leuenberger a.k.a. Dimstar wrote:

...

...

...

A bit extreme :-) Failing is actually not needed, one can search for the error in build logs. AFAIK there is no interface in obs that allows to fetch all build logs from a project in one request though.>

Do you seriously think someone reads build logs of successful builds? ;-)

...

...

Yes; that's why I proposed to do the fail only in a staging project... so we get at least visibility...

Yes, fully agreed.

I'd even go a step further and say: Either create a rc* symlink, or add a rpmlintrc explicitely stating you do not want that symlink.

We need to be careful with that tool. It's already annoying enough. The rc symlinks are just some sugar, it's not really crucial to have. Some page that shows the current rpmlint results of the whole Factory project would be more useful IMO. Esp if it had a way to file bugs in batch mode :-) cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.de/ SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 16746 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org

Christian Boltz wrote:

Am Freitag, 31. Januar 2014 schrieb Ludwig Nussel: [...]

...

Some page that shows the current rpmlint results of the whole Factory project would be more useful IMO. Esp if it had a way to file bugs in batch mode :-)

Do you really think batch-filed bugreports are the better solution? I doubt ;-)

I think so, yes. At least it worked in the past. See e.g.: https://bugzilla.novell.com/showdependencytree.cgi?id=764093&hide_resolved=0

As a maintainer, I'd prefer to have a rpmlint message like

Your package contains foo.service, but no rcfoo symlink. You can create the rcfoo symlink by adding this line to your spec: ( mkdir -p %{buildroot}/usr/sbin && cd %{buildroot}/usr/sbin && ln -s service rcfoo ) and add the symlink to %files: /usr/sbin/rcfoo If you think a rcfoo symlink doesn't make sense for your package (for example because foo.service should never be started or stopped manually), add the following line to rpmlintrc: [copy&paste-ready line for rpmlintrc]

For a packager that would mean: build failure -> copy&paste from rpmlint message (to spec or rpmlintrc) -> success

I wouldn't call that annyoing - especially when the error message contains clear instructions what to do.

Ideally, yes. rpmlint messages in general do not contain detailed instructions to fix something though as it's kind of long winded to always change rpmlint code to match the current way of doing things. So the concrete instructions are usually here: http://en.opensuse.org/openSUSE:Packaging_checks

OTOH, I somehow doubt a manual bugreport (batchmode or not) will always contain that information - in other words: it's harder for the packager to handle it.

Also, especially with batch mode, you'll probably get a fresh bugreport some months later because you still didn't add a rc* symlink (for whatever reason) and the batch-reporter won't check bugzilla for old reports in 30 packages ;-)

You are already thinking one step ahead. I didn't even think of automating this batch filing. I'd be perfectely happy if we had a visual overview of the current state of rpmlint warnings, so humans could watch over this.

[1] for the packages I maintain, I could imagine to - list all "well-known" rpmlint warnings in rpmlintrc - add a special comment "# rpmlint-all-fatal" to the spec that makes _all_ _new_ rpmlint warnings fatal.

The advantage of this method would be to notice new warnings quickly [2], but I also know that it would mean lots of work to get it started. That's why I include "add a special comment to the spec to enable this behaviour" which means it's opt-in for each package(r).

What do you think about this idea?

I think a build failure is the wrong way to communicate non-fatal errors :-) A notification of new rpmlint warnings could be implemented without having to touch packages at all. You'd just need some service that fetches build logs regularly and notify the maintainers of changes in the rpmlint output.

[2] again: do you seriously think someone reads the build log if the build succeeds? ;-)

No, that's why there needs to be somebody that watches over the (interesting) non-fatal errors every once in a while. I'm sure such persons could be found if the information that allows to do the work is easily available. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.de/ SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 16746 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org