denials always win so they shouldn't really be used unless one is
suuuuuuuuuuuuper sure about it.
to that end we'll throw up a warning that tries to explain that denials
probably aren't what the user wants and also that denying usually isn't
necessary because filesystem permissions still apply, so even when using
a liberal share ACL the filesystem ACL would still prevent other users
from reading one's private data for example

this was a bit awkward to implement because sheets need to overlay
something yet overlaying the columnlayout messes with the layout size
itself compacting the tableview away.
equally we cannot use a regular page because pages come with a whole
bunch of padding that looks very weird here and I've not found a way to
disable it either. instead we now have a new toplevel item which merely
acts as the thing the sheet can overlay. the item then contains the
previous layout and elements so they retain the sizing even when
overlayed

BUG: 422554
FIXED-IN: 20.12

the order of entries within the ACL have meaning as samba aborts
permission lookup on matching denials but combines read and full access.
since our ACL table mimics windows' we'll follow its behavior as it's
also fairly easy to explain

"a denial always denies, no matter what other rules exist"

to that end we'll now sort the ACL to first list denials, then reads,
then full access

- if foo is denied they'll not be let in
- if bar is read they get let in
- if Everyone is denied nobody gets in

should we later add group support this further becomes

- if groupFoo is denied and bar is a member of it then bar is denied
- if groupBar is read and foo is a member of it then foo is still denied
- if groupFoo is fullacces and bar is a member they'll get fullaccess

CCBUG: 422554
FIXED-IN: 20.12

this fell over in the refactor so make it a bit more streamlined. the
context now gets a URL (file:///home/foo) and feeds that into the
resolveShare which constructs a path from it (/home/foo) to then feed to
ksambashare (which exclusively deals with local paths).
the path on the share instance is then the path to always use. to aid
with that we'll stop holding the url as a member

this unbreaks creating new shares which had no path set (this is now
done in resolveShare). also print that on applyChanges for easy
debugging

this is in preparation to add more "wizardy" behavior for scenarios
where samba isn't installed yet, such as adding the user to the samba
database.

the way the new ui works is that it's rendering a page stage, we push
pages on as they become relevant. longer term that'll be at least the
pages install->reboot->setpassword->acls

this also features some related improvements such as
- better const corectness for some member vars
- less ifdef samba_install
- the maximum share name is now capped at 60 characters (which is what
windows10 allows)
- there's a soft warning for the 60 character limit as well in the gui

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

as per new licensing policy

- use override instead of virtual; it implies virtual
- use =default where possible
- reduce slot lambda signatures to only used arguments
- getUsersList is now static - it uses no members
- reboot is now static - it uses no member
- use auto when the right hand side of an assignment shows the type
- make sure pointer members are init'd to nullptr lest we forget to
declare them in the ctor

previously we'd ignore the returned errors leaving the user hanging when
adding a share fails and not even telling them why.

we now route errors on add/remove through somewhat generic
kmessagebox-based reporting tech.
unfortunately this is fairly excessive WRT string mapping because
ksambashare sports one enum for everything so API-wise a call to save()
could return UserShareAclOk but in practice that of course cannot happen
because that'd mean nothing. as such many of the strings are overly
generic because they cannot practically appear on weak context actions
like save().
very meh.

this depends on the new lastSystemErrorString() API in KIO to stringify
the majority of failure scenarios here, which come out of samba's 'net'
command

trying to share /tmp now produces a kmessagebox informing the user that
the dir cannot be shared

BUG: 334618
FIXED-IN: 20.12

it's unreadable and also bound to break should the column indexes change

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

do not fail the build just because a new qt5/kf5 version deprecated
something...

This didn't make sense as the two are not connected.

Summary:
Compile without emit and foreach.
The private variables inside the model are prefixed with _, because
the userList variable shadowed class definition. This should clear things up.

And some better overloads and const qualifiers.

Test Plan: Compiles

Reviewers: #dolphin, ngraham

Reviewed By: #dolphin, ngraham

Differential Revision: https://phabricator.kde.org/D29118

Summary:
Fix compile warnings:
warning: zero as null pointer constant
warning: variable ‘result’ set but not used

Additionally some improved overloads and formatting which does not
deserve a sepearte patch ;-).

Test Plan: Compiles

Reviewers: #dolphin, ngraham, mlaurent

Reviewed By: #dolphin, ngraham

Differential Revision: https://phabricator.kde.org/D29105

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

Summary:
After Samba is installed, very frequently it will not work correctly until the machine is
rebooted. One potential reason is when the installed package has made group membership
changes, which only take effect after a reboot.

This patch implements a reboot recommendation along with a button to reboot the machine.
Both are shown immediately after Samba has been installed.

Because some expert users may understand the technical details of what's going on, the
reboot is only recommended, not required. If the window is closed and then re-opened,
the normal Samba sharing configuration UI is displayed instead of the reboot prompt.

Added a new function rather than using a Lambda because it may be useful for additional
purposes too (e.g. https://bugs.kde.org/show_bug.cgi?id=407846)

FEATURE: 407845
FIXED-IN: 20.04.0

Test Plan:
http://s1.webmshare.com/Ry55q.webm (not uploaded to Phab due to file size limit)

1. Remove Samba
2. Go to share a folder
3. Click "Install Samba"
4. After the installation has completed, click the "Restart" button and see that the machine reboots

5. Remove Samba again
2. Go to share a folder
3. Click "Install Samba"
4. Close the window and re-open it instead of rebooting as recommended
5. See that the Samba sharing config UI is all there

Reviewers: #vdg, #frameworks, #dolphin, apol, sitter

Reviewed By: sitter

Subscribers: anthonyfieroni, sitter, bruns

Differential Revision: https://phabricator.kde.org/D21466

Summary:
the max uid is now actually being used to cap the displayed uids (e.g.
I have a system user created beyond my MAX_UID of 60000 that were
erroneously shown because the detected maxuid was unused)

Test Plan: the excess user no longer shows up

Reviewers: ngraham

Reviewed By: ngraham

Differential Revision: https://phabricator.kde.org/D26964

Summary:
the pervious tech had shoddy coverage and was bound to eventually break
because of the rise of /etc/os-release to rule all the proprietary release
files that preceded it.

instead we now set sane default assumptions and attempt to load the actual
default for the specific system from /etc/login.defs (part of shadow source
and should be available on more or less all modern linux systems
https://github.com/shadow-maint/shadow).

Test Plan: max/min detection still works

Reviewers: ngraham

Reviewed By: ngraham

Differential Revision: https://phabricator.kde.org/D26963