Cleaning up old scriptlets

[Scott Talbert]

I'm cleaning up some old scriptlets and have a few questions about whether 
these are needed anymore:

1) /sbin/ldconfig is no longer needed after installation a shared 
library in F28+, correct?

2) Is update-mime-database still needed after installing a file in 
%{_datadir}/mime/packages/ ?

3) Is 'udevadm control --reload-rules' still needed after installing a 
udev rules file?

Thanks,
Scott

[Jason L Tibbitts III]

> >>>> "ST" == Scott Talbert
> <swt(a)techie.net&gt; writes: 
ST> 1) /sbin/ldconfig is no longer needed after installation a shared
ST> library in F28+, correct?

That is correct.  It doesn't hurt anything, but can slow down package
installation. See
https://fedoraproject.org/wiki/Packaging:Scriptlets#Shared_Libraries for
a full explanation and description of macros which work across all
releases.

ST> 2) Is update-mime-database still needed after installing a file in
ST> %{_datadir}/mime/packages/ ?

This has not been needed since... Fedora 24 or 25 I think.  No
scriptlets for update-mime-database are currently listed in
https://fedoraproject.org/wiki/Packaging:Scriptlets

For EPEL, there is
https://fedoraproject.org/wiki/EPEL:Packaging#mimeinfo

ST> 3) Is 'udevadm control --reload-rules' still needed after installing
ST> a udev rules file?

Not sure at all; the packaging guidelines have never listed any
requirement for calling udevadm.

 - J<

[Robert-André Mauchin]

On mardi 12 juin 2018 19:23:00 CEST Jason L Tibbitts III wrote:

>  ST> 3) Is 'udevadm control --reload-rules' still needed
> after installing
>  ST> a udev rules file?
>  
>  Not sure at all; the packaging guidelines have never listed any
>  requirement for calling udevadm.
>   
This shouldn't be necessary as there is a transfiletrigger in Systemd to 
handle new rules:

%transfiletriggerin udev -- /usr/lib/udev/rules.d
# This script will automatically update udev with new rules if files
# have been installed or updated in /usr/lib/udev/rules.d.
if test -d /run/systemd/system; then
  %{_bindir}/udevadm control --reload
fi

https://src.fedoraproject.org/rpms/systemd/blob/master/f/triggers.systemd...

[Jason L Tibbitts III]

> >>>> "RM" == Robert-André Mauchin
> <zebob.m(a)gmail.com&gt; writes: 
RM> This shouldn't be necessary as there is a transfiletrigger in
RM> Systemd to handle new rules:

Note that this is in F28+ only.

 - J<

[Robert-André Mauchin]

On mardi 12 juin 2018 20:21:22 CEST Jason L Tibbitts III wrote:
>  >>>>> "RM" == Robert-André Mauchin
> <zebob.m(a)gmail.com&gt; writes:
>  
>  
>  RM> This shouldn't be necessary as there is a transfiletrigger in
>  RM> Systemd to handle new rules:
>  
>  Note that this is in F28+ only.
>  
>   - J< 
I've just discovered %udev_rules_update though a review. Equates to %nil in 
F28 and udevadm control --reload >/dev/null 2>&1 in F27. Pretty neat.

[Jason L Tibbitts III]

> >>>> "RM" == Robert-André Mauchin
> <zebob.m(a)gmail.com&gt; writes: 
RM> I've just discovered %udev_rules_update though a review. Equates to
RM> %nil in F28 and udevadm control --reload >/dev/null 2>&1 in
RM> F27. Pretty neat.

Would be really neat if these actually got documented somewhere.  And
unfortunately having a macro like that doesn't help all that much
because you still end up having to conditionalize the scriptlet
dependencies and usually the scriptlet itself for the distro version
(assuming you insist on keeping the same spec for each branch).  But at
least most of it will just go away after F27 goes EOL.

Handling the complete set of cases is what makes the ldconfig macros so
complicated.

 - J<

[Neal Gompa]

On Mon, Jun 18, 2018 at 5:37 AM Jason L Tibbitts III <tibbs(a)math.uh.edu&gt; wrote:
>  >>>>> "RM" == Robert-André Mauchin <zebob.m(a)gmail.com&gt;
> writes:
>
>  RM> I've just discovered %udev_rules_update though a review. Equates to
>  RM> %nil in F28 and udevadm control --reload >/dev/null 2>&1 in
>  RM> F27. Pretty neat.
>
>  Would be really neat if these actually got documented somewhere.  And
>  unfortunately having a macro like that doesn't help all that much
>  because you still end up having to conditionalize the scriptlet
>  dependencies and usually the scriptlet itself for the distro version
>  (assuming you insist on keeping the same spec for each branch).  But at
>  least most of it will just go away after F27 goes EOL.
>
>  Handling the complete set of cases is what makes the ldconfig macros so
>  complicated.
>  
In most cases, it was not necessary even before, since udev watches
the rules directory with inotify and reloads automatically. There are
only a couple of special cases where this doesn't kick in (mainly if
the rule called another program and that program was updated), and
that's where the file trigger helps with it.


-- 
真実はいつも一つ！/ Always, there's only one truth!