GitLab

Summary:
When a QApplication is in it's destructor it calls window->destroy()
on all remaining QWindows. This deletes the platform window, whilst
keeping the real object alive. QtWayland doesn't like this.

Deleting the dialog fixes the issue. The deleteLater is fine because
QCoreApplication teardown is magic.

Test Plan: crashTest

Reviewers: #plasma, sitter

Reviewed By: sitter

Subscribers: anthonyfieroni, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D28692

the whatstring isn't actually entirely sufficient as it turns out. it is
derived from errorstring but that seems to sometimes not contain any
useful information as to what the actual protocol problem was. so, log
code, text and string when throwing a protocol exception

Summary:
KWallet::openWallet takes a windowId as a parameter.

QWidget->winId() is a dangerous call, it will create an underlying
platform window when called. This happens even if we're not the toplevel
widget. On X11 this passed winId must have always been broken, but not
in a way that caused a problem.

This leaves QtWayland in a very confused state with the side effect that
all scrolling widgets become broken.

BUG: 418755
BUG: 420035

Test Plan:
Bisected drkonqi till we found the cause
Verified fix by running crashtest (with some mods so I could report a bug)
I can't confirm the winId is correct on X11, but the code looks safe

Reviewers: #plasma, sitter

Reviewed By: sitter

Subscribers: sitter, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D28832

Summary:
previously this supposedly was

`#2  0x00007fefe9ef79e8 in ?? () at /usr/lib/dri/i965_dri.so`

nowadays the ?? is simply empty. the rating mechanism wasn't properly
handling this. treat an empty function name to mean the same as ?? and
rate the line as MissingFunction/Everything depending on whether
libraryName is also missing.

also update test expectations accordingly (could have actually seen that
discrepancy if one had paid attention to the outcoming rating :S)

CCBUG: 418538

Test Plan:
test passes. most lines from 417435 now rate `MissingFunction` and the whole trace is considered useless

```
Rating: 128 out of 364 Usefulness: Useless
90%: 327.6 70%: 254.8 40%: 145.6
```

Reviewers: cfeck, ngraham

Reviewed By: ngraham

Subscribers: ahmadsamir, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D28027

Summary:
As seen in https://bugs.kde.org/show_bug.cgi?id=417435 the previous fix
for .so detection actually falls flat on the face for most things
other than glib2. The original test here was flawed as libglib-2.0.so.0
has the complete suffix '0.so.0' which matched the original parsing
condition. most libraries however are libfoo.so.0 and for that the suffix
would be 'so.0' which did not match the original condition.

extend the condition to properly cover files without so-version, files with
awkward suffix (such as glib) and files with normal names and normal
versioning.
also extend test case accordingly

CCBUG: 418538

Test Plan: test passes. all lines from 417435 score less than `Good`

Reviewers: cfeck, ngraham

Subscribers: plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D28026

GIT_SILENT

Summary: These are less numbers, but more identifiers in the system.

Reviewers: #plasma, sitter

Reviewed By: sitter

Subscribers: plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D28142

Summary:
... and update a related test case

the original motivation here was garbage _start traces shouldn't be
considered valuable. of course we have way more advanced measurements for
whether a trace is valuable or not. notably the actual rating. so as
long as we have any frame for evaluation that frame's rating **is**
relevant.
drkonqi is smart enough to ignore lines that do not matter already.
e.g. everything above the crash handler is ignored, equally originator
functions main/start_thread/start/... are ignored. they do not impact
the rating at all, regardless of the amount of frames under review.
as such it is entirely possible to have fewer than 4 frames under review
and that those 4 frames are all we need (e.g. the crash is in main()
directly)

in fact, the very trace sample testing this was proving how the 4 line
limit made no sense. the frame about KCmdLineArgs is the only relevant
and only valuable frame in the sample and it is 100% useful making the
entire trace useful. it is simply a nullptr dereference in the main.

the frame limit is now 1 and the test sample has been replaced with
a real-life sample from https://bugs.kde.org/show_bug.cgi?id=193032

Test Plan: test passes again (was broken because _starts is now actively skipped)

Reviewers: apol, bcooksley

Reviewed By: apol

Subscribers: ahmadsamir, bcooksley, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D28098

GIT_SILENT

Summary:
we already (re)set the focus on failed login attempt but neglected to do
it when showing the page.

BUG: 418309
FIXED-IN: 5.18.3

Test Plan:
- empty kwallet
- try to report bug
- dialog appears and has focus on email input
- go back
- go forward
- still focus on email input
- provide wrong credentials
- again focus on email input + error message

Reviewers: #plasma, davidedmundson

Reviewed By: #plasma, davidedmundson

Subscribers: plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D27776

GIT_SILENT

GIT_SILENT

GIT_SILENT

Summary:
for unknown reasons gdb seems to behave inconsistently here. usually
'at' denotes a file, but there are backtraces where it doesn't
e.g. https://bugs.kde.org/show_bug.cgi?id=416923

I've add a trivial suffix check to ensure parsing of these lines works
correctly even when 'at' is used in combination with a library name.

this is a bit hackish, but in reality this entire parsing tech should
probably be replaced by a python plugin for gdb so we can get interactive
access to the frames and serialize them in a well defined format instead of
having to parse "random" text

on top of that our regex assumed we'd always have a function name, which is
also not true as that bug report shows. to mitigate this the matching group
has been made optional.

this commit also adds a test for the gdb line parsing unit with some
obvious line samples I could find just now

Test Plan: all tests pass

Reviewers: ngraham

Reviewed By: ngraham

Subscribers: ngraham, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D27041

Fabian Vogt authored Jan 17, 2020 and Christophe Giboudeaux committed Jan 17, 2020

Summary:
This way the repository with debuginfo does not have to be enabled
manually before clicking on the button.
This option works again now, so it should be passed as default.

BUG: 416038

Test Plan: Works locally. Reported success on the linked bug.

Reviewers: lbeltrame, cgiboudeaux

Reviewed By: cgiboudeaux

Subscribers: plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D24751

GIT_SILENT

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: Adds the KIOFuse binary to the mapping for bug reporting.

Test Plan: None

Reviewers: fvogt, ngraham

Reviewed By: ngraham

Subscribers: plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D26104

# Conflicts:
#	CMakeLists.txt

since the test relies on uname (and the relevant code inside sysinfo
also conditions on uname) there is little point in building it on platforms
without uname e.g. windows.

spec-wise inheriting constructors is an all or nothing sort of affair which
means NewBug needs to manually implement the default constructor because
it actually does needs a custom copy constructor (this may have to do with
qobject having a q_disable_copy call for the cctor, so probably msvc
dislikes using for that reason)

GIT_SILENT

this is now implemented on a connection level, so we need to test it
there.

this applies 4e29b8a7 to everything

I've since gotten an account with +foobar@gmail.com suffix and confirmed
it too needs encoding to properly log in as well.
So does a product with the name 'foo+bar'. I've deduced that bugzilla's
query handling simply always assumes full encoding WRT the plus character
and so we'll force full encoding through the HTTPConnection to always
use full encoding (i.e. including the plus character)

CCBUG: 413920

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"

# Conflicts:
#	CMakeLists.txt

QUrlQuery doesn't encode + by default while bugzilla wants it to be.
This prevented passwords that include a plus from working.
Simply force-encode the password string to be fully encoded.

I've confirmed this working with passwords entailing spaces, plus, and
percent characters, all at the same time.

BUG: 413920
FIXED-IN: 5.17.3

memfd files for QML JIT are /memfd:JITCode:/path/to/libQt5Qml.so.5 which
results in false positives for the deletion regex we use to detect
updated software

Summary:
this actually improves UX more than anything. when there was an update it
is fairly impossible to install debug symbols, so drkonqi may get stuck in
a loop between saying this crash is garbage and suggesting the user install
some more symbols. what's more is that the crash can be the result
of incompatibilities in runtime-loaded plugins (a notorious example are
KIO slaves, which get forked from klauncher which I think can lead
to problems when loading the newer libQt5Core.so etc.).

all in all crashes coming from a half-update runtime are hard to trace
and also at risk of being worthless. so, instead disable the report and
the install symbols features and inform the user about what
went wrong in both cases.

this is pretty much exclusively working for linux where the kernel will
actually add " (deleted)" hints to all symlinks in procfs and have a
map_files directory where all mmap'd regions are symlinked to their
respective paths... also with that suffix as needed.

the way it works is super easy: iter all map_files, get all deleted paths,
make sure none of the .so or the executable itself are marked deleted.
if anything is marked deleted, the report feature gets disabled

Test Plan: update dolphin -> attach drkonqi -> drkonqi whines about dolphin being updated

Reviewers: #plasma, apol

Reviewed By: apol

Subscribers: apol, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D25002

Summary:
This allows devs to run: "make clang-format" and format all files easily
using the preset KDE clang format style

This patch adds support so devs can easily test the intended formatting
now, it doesn't change the code yet.

Doing an initial run to commit everything will happen later.

Test Plan:
Ran script
git diff was full of amazingly cleaned code

See D25134

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"

GIT_SILENT

# Conflicts:
#	CMakeLists.txt