Sometime in December 2025 I noticed qbittorrent-nox 5.1.4 would no longer stay running for me.  Sometime it would run for a short time, but often it would just quit straight away after a "service qbittorrent restart".  I don't know whether this coincided with my upgrading to FreeBSD/amd64 15-STABLE from 14-STABLE, or due to other package updates.  (I build all my packages locally via Poudriere and rebuilt and reinstalled them all as part of the FreeBSD upgrade.)

I would notice something like this sometimes in my daily security run output when having the problem with qbittorrent-nox:

pid 3111 (qbittorrent-nox), jid 0, uid 922: exited on signal 5 (no core dump - other error)

I updated to qbittorrent-nox 5.1.4_1 today and now it appears to run successfully again, so whatever was causing it to quit before no longer is doing that now for me.

Comment 2 Jason E. Hale 2026-01-13 06:42:06 UTC

I just installed the qbittorrent 5.1.4 package on 15.0-STABLE amd64 (which was not built with WITH_DEBUG) and couldn't get the crash and there is nothing outputted to the console upon startup.

(In reply to Jason E. Hale from comment #2)

Are you saying qbittorrent 5.1.4 runs properly for you on 15-STABLE?  For me, I could not get it to run for more than a couple of minutes before it would mysteriously quit (with the signal 5 error in the system log).  My problem is that 5.1.4 would not stay running, whereas it had before.  Qbittorrent-nox 5.1.4_1 restores the behaviour for me where it will stay running.

Comment 4 Yuri Victorovich 2026-02-01 01:11:02 UTC

(In reply to Paul Mather from comment #3)

qbittorrent runs for a very long time when it is built with WITH_DEBUG=yes.
This is currently a workaround in this port.

Same in multimedia/obs-studio: it crashes when built with debug info and works fine when built with debug info.

In multimedia/obs-studio I narrowed it down to having this if statement fail:
  945│     if (Q_UNLIKELY(!qobject_cast<QApplication *>(QCoreApplication::instance())))
  946├───────> qFatal("QWidget: Cannot create a QWidget without QApplication");
  947│

in /usr/ports/devel/qt6-base/work/qtbase-everywhere-src-6.10.1/src/widgets/kernel/qwidget.cpp
despite the fact that QApplication was created and is present.

Comment 5 Yuri Victorovich 2026-02-01 01:11:54 UTC

(In reply to Yuri Victorovich from comment #4)

* obs-studio aborts w/out debug info

(In reply to Yuri Victorovich from comment #4)

> Same in multimedia/obs-studio: it crashes when built with debug info and works fine when built with debug info.

Something about Qt's inheritance checks was broken in https://github.com/qt/qtbase/commit/a3d2e45cb7cc61f8f9bb7ff968d5ca8247a15bd4 (due to -Bsymbolic-functions). This can worked around with CMAKE_OFF+= QT_FEATURE_reduce_relocations, but really should be investigated further.

(In reply to Alex S from comment #6)

I can't confirm this being relevant to qBittorrent, though. Doesn't crash for me.

The repro looks like this:

~/qt-casting $ cat CMakeLists.txt 
cmake_minimum_required(VERSION 3.16)

project(qt-casting)

set(CMAKE_INTERPROCEDURAL_OPTIMIZATION_RELEASE ON)

set(CMAKE_AUTOUIC ON)
set(CMAKE_AUTOMOC ON)
set(CMAKE_AUTORCC ON)

find_package(Qt6 REQUIRED COMPONENTS Core Widgets)

add_executable(qt-casting main.cpp)
target_link_libraries(qt-casting PRIVATE Qt::Core Qt::Widgets)
~/qt-casting $ cat main.cpp 
#include <cstdio>
#include <QApplication>

class Whatever : public QObject {
    Q_OBJECT
};

#include "main.moc"

int main(int argc, char *argv[])
{
    QApplication a(argc, argv);

    printf("qobject_cast:      %s\n", qobject_cast<QApplication *>(QCoreApplication::instance()) != nullptr   ? "ok" : "fail");
    printf("inheritance check: %s\n", QCoreApplication::staticMetaObject.inherits(&QObject::staticMetaObject) ? "ok" : "fail");

    Whatever whatever;
    printf("%p\n", &whatever);

    return 0;
}
~/qt-casting $ mkdir build
~/qt-casting $ cd build/
~/qt-casting/build $ cmake .. -DCMAKE_BUILD_TYPE:STRING="Release"
...
~/qt-casting/build $ make
...
~/qt-casting/build $ ./qt-casting
qobject_cast:      ok
inheritance check: fail
0x8211dc538

Comment 8 Yuri Victorovich 2026-02-01 16:57:13 UTC

Should CMAKE_OFF+= QT_FEATURE_reduce_relocations be added to all */qt6-* ports or only to qt6-base?

IMO this should be reported as a bug to the Qt upstream.

Comment 9 Jason E. Hale 2026-02-02 06:12:23 UTC

(In reply to Yuri Victorovich from comment #8)
> Should CMAKE_OFF+= QT_FEATURE_reduce_relocations be added to all */qt6-* ports or only to qt6-base?

If this is indeed the fix, it would only be needed in qt6-base. The other qt modules would inherit this feature, but would all likely need to be rebuilt.

I'm preparing Qt 6.10.2 now, so if I can verify that this works, I'll include it in the update so folks don't have to rebuild all of Qt twice in a short period of time.

Comment 10 commit-hook 2026-02-08 06:44:46 UTC

A commit in branch main references this bug:

URL: https://cgit.FreeBSD.org/ports/commit/?id=b4a5f8275569ab3abf2962939fd19c8bc91e5760

commit b4a5f8275569ab3abf2962939fd19c8bc91e5760
Author:     Jason E. Hale <jhale@FreeBSD.org>
AuthorDate: 2026-02-05 21:11:37 +0000
Commit:     Jason E. Hale <jhale@FreeBSD.org>
CommitDate: 2026-02-08 06:21:14 +0000

    Qt6: Update to 6.10.2

    January 29, 2026 by Jani Heikkinen

    Qt 6.10.2 is now available for download. As a patch release, Qt 6.10.2
    doesn’t introduce new features, but it delivers around 300 bug fix,
    security improvements, and quality enhancements on top of Qt 6.10.1.
    For a full overview of the most notable changes, take a look at the Qt
    6.10.2 release notes.

    Announcement:
    https://www.qt.io/blog/qt-6.10.2-released
    Release notes:
    https://code.qt.io/cgit/qt/qtreleasenotes.git/about/qt/6.10.2/release-note.md

    FreeBSD ports changes:
    devel/qt6-base:
    - Disable the reduced_relocations feature. Prior to Qt 6.10.0, this was
      only enabled if the compiler was GCC, but we are seeing crashes in
      certain ports such as misc/bibletime, multimedia/obs-studio, and
      net-p2p/qbittorrent. [1]
    - Rebase patches

    graphics/qt6-wayland:
    - Add a pre-configure check to test if devel/qt6-base was built with
      the WAYLAND option. This module cannot be built if the WAYLAND
      option is disabled in devel/qt6-base. [2]

    www/qt6-webengine:
    - Patched with security patches up to Chromium version: 144.0.7559.109
    - Define NODEJS_EXECUTEABLE in the configure environment after [3]. This
      is forward porting for the post-release security rollup and not part of
      the official release. Future releases [3][4] will likely require this.
    - Remove upstreamed patches

      [3] https://code.qt.io/cgit/qt/qtwebengine-chromium.git/commit/?h=134-based&id=cf510ebd4e870eb24fb5bc1d1ddcc5442247672d
      [4] https://code.qt.io/cgit/qt/qtwebengine-chromium.git/commit/?h=140-based&id=aab0f6f0928e7e4f73832abba9a00b7946b7e459

    x11-toolkits/qt6-declarative
    - Remove upsteamed patches

    The Qt python bindings have also been updated along with their
    respective tools.

    PySide6: Update to 6.10.2
    PyQt6: Update to 6.10.2

    PR:             292376 [1], 292273 [2]
    Reported by:    yuri [1], jbo [2]
    MFH:            2026Q1
    Security:       73ff246b-04b2-11f1-84fc-4ccc6adda413

 Mk/Uses/pyqt.mk                                    |    6 +-
 Mk/Uses/qt.mk                                      |    4 +-
 accessibility/qt6-speech/distinfo                  |    6 +-
 comms/qt6-connectivity/distinfo                    |    6 +-
 comms/qt6-sensors/distinfo                         |    6 +-
 comms/qt6-serialbus/distinfo                       |    6 +-
 comms/qt6-serialport/distinfo                      |    6 +-
 databases/qt6-base_sqldriver/distinfo              |    6 +-
 devel/py-qt5-sip/distinfo                          |    6 +-
 devel/py-qt6-pyqt/distinfo                         |    6 +-
 devel/py-qt6-sip/distinfo                          |    6 +-
 devel/pyside6-tools/distinfo                       |    6 +-
 devel/pyside6/distinfo                             |    6 +-
 devel/qt6-5compat/distinfo                         |    6 +-
 devel/qt6-base/Makefile                            |    4 +-
 devel/qt6-base/distinfo                            |    6 +-
 .../files/patch-cmake_QtPkgConfigHelpers.cmake     |    4 +-
 .../files/patch-src_corelib_CMakeLists.txt         |    4 +-
 .../patch-src_corelib_global_qcompilerdetection.h  |    4 +-
 .../files/patch-src_corelib_io_qlockfile__unix.cpp |    4 +-
 .../patch-src_corelib_thread_qthread__unix.cpp     |   31 +-
 ...rc_network_socket_qnativesocketengine__unix.cpp |    4 +-
 ...hellintegration_xdg-shell_qwaylandxdgshell__p.h |    4 +-
 devel/qt6-base/pkg-plist                           |   13 +-
 devel/qt6-grpc/distinfo                            |    6 +-
 devel/qt6-languageserver/distinfo                  |    6 +-
 devel/qt6-location/distinfo                        |    6 +-
 devel/qt6-positioning/distinfo                     |    6 +-
 devel/qt6-remoteobjects/distinfo                   |    6 +-
 devel/qt6-scxml/distinfo                           |    6 +-
 devel/qt6-tools/distinfo                           |    6 +-
 devel/qt6-translations/distinfo                    |    6 +-
 devel/qt6-translations/pkg-plist                   |    5 +
 devel/shiboken6/distinfo                           |    6 +-
 graphics/qt6-3d/distinfo                           |    6 +-
 graphics/qt6-imageformats/distinfo                 |    6 +-
 graphics/qt6-lottie/distinfo                       |    6 +-
 graphics/qt6-quickeffectmaker/distinfo             |    6 +-
 graphics/qt6-svg/distinfo                          |    6 +-
 graphics/qt6-wayland/Makefile                      |    8 +-
 graphics/qt6-wayland/distinfo                      |    6 +-
 misc/qt6-doc/distinfo                              |    6 +-
 misc/qt6-doc/pkg-plist                             |   11 +-
 misc/qt6-examples/distinfo                         |    6 +-
 misc/qt6-examples/pkg-plist                        |  159 ++-
 multimedia/qt6-multimedia/Makefile                 |    1 -
 multimedia/qt6-multimedia/distinfo                 |    6 +-
 net/qt6-coap/distinfo                              |    6 +-
 net/qt6-mqtt/distinfo                              |    6 +-
 net/qt6-networkauth/distinfo                       |    6 +-
 science/qt6-quick3dphysics/distinfo                |    6 +-
 www/qt6-httpserver/distinfo                        |    6 +-
 www/qt6-webchannel/distinfo                        |    6 +-
 www/qt6-webengine/Makefile                         |    7 +-
 www/qt6-webengine/distinfo                         |    6 +-
 www/qt6-webengine/files/patch-QTBUG-139335 (gone)  |  610 ----------
 .../files/patch-security-rollup (new)              | 1261 ++++++++++++++++++++
 www/qt6-websockets/distinfo                        |    6 +-
 www/qt6-webview/distinfo                           |    6 +-
 x11-toolkits/qt6-charts/distinfo                   |    6 +-
 x11-toolkits/qt6-datavis3d/distinfo                |    6 +-
 x11-toolkits/qt6-declarative/Makefile              |    1 -
 x11-toolkits/qt6-declarative/distinfo              |    6 +-
 .../files/patch-QTBUG-142331 (gone)                |  229 ----
 .../files/patch-QTBUG-142514 (gone)                |  122 --
 ...ch-src_3rdparty_masm_assembler_ARM64Assembler.h |    4 +-
 ...ch-src_3rdparty_masm_assembler_ARMv7Assembler.h |   12 +-
 .../files/patch-src_qml_configure.cmake            |   10 +-
 x11-toolkits/qt6-declarative/pkg-plist             |    1 +
 x11-toolkits/qt6-graphs/distinfo                   |    6 +-
 x11-toolkits/qt6-quick3d/distinfo                  |    6 +-
 x11-toolkits/qt6-quicktimeline/distinfo            |    6 +-
 x11-toolkits/qt6-shadertools/distinfo              |    6 +-
 x11-toolkits/qt6-virtualkeyboard/distinfo          |    6 +-
 74 files changed, 1603 insertions(+), 1208 deletions(-)

(In reply to commit-hook from comment #10)
Reporting here as the commit doesn't have PR dedicated for this upgrade of this.

At least x11/qt6-declarative fails to package with a plenty of unaccessible staged files. If it's OK for you, isn't pkg-plist properly pushed?

Created attachment 267905 [details]
Error list (install stage with pkg_replace) of qt6-declarative

Error list of x11-toolkits/qt6-declarative.
Actually, this is the install part of pkg_replace, but the result of `make package` after `make` on /usr/ports/x11-toolkits/qt6-declarative looks the same.

Note that current pkg-plist is 344058 bytes in size, while the output of `make makeplist`is 83355 bytes in size (including message at the 1st line).

Comment 13 Jason E. Hale 2026-02-08 13:29:40 UTC

(In reply to Tomoaki AOKI from comment #12)
I've just re-tested x11-toolkits/qt6-declarative 6.10.2 in poudriere and the pkg-plist is correct as far as I can tell. Only one line was inserted into the pkg-plist 6.10.1 -> 6.10.2, so I don't know what your doing to come up with 83355 bytes, but 344058 bytes is correct size of the pkg-plist. The size for the 6.10.1 version of qt6-declarative's pkg-plist was 344003 bytes, so the 55 added characters maths out to 344058.

(In reply to Jason E. Hale from comment #13)
Possibly it is specific for bare-metal builds.

The environment I've bitten is bare-metal environment using ports-mgmt/pkg_replace.
main at commit base a25b12c6ce63be1fefb31d14daca332682fd31fc, amd64.

Try deleting everything under /usr/local/lib/compat/pkg didn't help (expected, as there were nothing related with qt6).

Why I don't use poudriere on main is because it's too much sensitive with OSVERSION.
Forcing full rebuilds even on bumps in least 2 digits is too much a pain on main,
and my daily driver is stable/15 running on exactly the same computer (on another SSD), which means I cannot give too much time for upgrading main (test and emergency environment).

Sorry, time I can give for main this weekend is over and need to switch back to stable/15 shortly. I'll retry (hopefully) next weekend.

Switched back to stable/15 at commit base 58a369a6f6ce6786ef7aa4715c96548f36ae396 on exactly the same computer that has issues on main.

poudriere build (not yet finished the whole upgrades) passed x11-toolkit/qt6-declarative and currently no fallouts.

So the issue would be specific for bare-metal upgrade builds.
Possible reason I can think of for now is *.pc files of previous version is affecting. (As noted on Comment 14, the issue persists even after emptying /usr/local/lib/compat/pkg/.)
Or insufficient build order restrictions?

Poudriere builds finished without fallouts, but intentionally ran bare-metal upgrades using pkg_replace without using built pkgs, come to reproduce failure on stable/15, too.

Not base branch dependent, bare-metal upgrades specific failure.

(In reply to Tomoaki AOKI from comment #16)
Proceeded upgrade using pre-built pkgs with local poudriere on stable/15, then, retried `make` and `make package` at /usr/ports/x11-toolkits/qt6-declarative and it went fine.

So I think I could confirm this issue is bare-metal upgrade specific.
But it shold be better fixed.

Comment 18 Torsten Zuehlsdorff 2026-03-22 19:14:52 UTC

I am closing this as the qbittorrent crash caused by `-Bsymbolic-functions` breaking Qt's inheritance checks was fixed in the Qt6 6.10.2 update (commit b4a5f827), which disables the `reduced_relocations` feature. The fix was identified by iwtcex@gmail.com and committed by the KDE team. The current port version is 6.10.2.

Comment 19 Jesper Schmitz Mouridsen 2026-04-05 19:42:52 UTC

FWIW
diff --git a/multimedia/obs-studio/Makefile b/multimedia/obs-studio/Makefile
index 02d459f69..1b6138847 100644
--- a/multimedia/obs-studio/Makefile
+++ b/multimedia/obs-studio/Makefile
@@ -69,9 +69,9 @@ CMAKE_OFF=    CMAKE_COMPILE_WARNING_AS_ERROR # no -Werror
 CFLAGS+=       -Wno-error=attribute-warning
 CFLAGS_powerpc64=      -Wno-error=deprecated-altivec-src-compat -Wno-error=pass-failed
 CFLAGS_powerpc64le=    -Wno-error=deprecated-altivec-src-compat
-CFLAGS+=       -fPIC # otherwise we get: relocation R_X86_64_PC32 cannot be used against symbol '__stack_chk_guard'; recompile with -fPIC
-CXXFLAGS+=     -fPIC # otherwise we get: relocation R_X86_64_PC32 cannot be used against symbol '__stack_chk_guard'; recompile with -fPIC
-LDFLAGS+=      -fPIC # pending https://github.com/obsproject/obs-studio/issues/3436
+CFLAGS+=       -fPIC  -fvisibility=hidden -fvisibility-inlines-hidden  # otherwise we get: relocation R_X86_64_PC32 cannot be used against symbol '__stack_chk_guard'; recompile with -fPIC
+CXXFLAGS+=     -fPIC -fvisibility=hidden -fvisibility-inlines-hidden  # otherwise we get: relocation R_X86_64_PC32 cannot be used against symbol '__stack_chk_guard'; recompile with -fPIC
+LDFLAGS+=      -fPIC -fvisibility=hidden -fvisibility-inlines-hidden # pending https://github.com/obsproject/obs-studio/issues/3436


Makes obs work start on qt6-base-10.2 with QT_FEATURE_reduce_relocations ON

Comment 20 Jesper Schmitz Mouridsen 2026-04-08 21:53:13 UTC

Created attachment 269518 [details]
reproducer of obs-studio issue.

cmake -DCMAKE_VERBOSE_MAKEFILE=On -DCMAKE_BUILD_TYPE=Release .. 
with lto it exposes the issue of obs-studio remove "set(CMAKE_INTERPROCEDURAL_OPTIMIZATION ON)" and it works, without having the QT_FEATURE_reduce_relocations set to OFF as in the original workaround.
smt is broken in  lto?

Comment 21 Jesper Schmitz Mouridsen 2026-04-10 20:32:25 UTC

(In reply to Alex S from comment #6)
Your reproducer seems to be fine with cmake -DCMAKE_VERBOSE_MAKEFILE=ON -DCMAKE_CXX_COMPILER=g++14 -DCMAKE_BUILD_TYPE=Release ..

So the lto of llvm is not healthy for qt with -Bsymolic-functions? Or is this expected somehow?  


/srepro/gccbuild]# ./qt-casting:
qobject_cast:      ok
inheritance check: ok
srepro/clangbuild
qobject_cast:      ok
inheritance check: fail

(In reply to Jesper Schmitz Mouridsen from comment #21)

> So the lto of llvm is not healthy for qt with -Bsymolic-functions?

I don't know. Try comparing symbol tables, I suppose. I'm not particularly familiar with Qt and was only doing basic reduction/bisection steps because the issue was blocking my patch for the OBS build configuration.

> Or is this expected somehow?

Nothing should fail. (Actually, the qobject_cast check never fails with this repro, the trigger seems to be no longer there after a few rounds of simplification, so it's just the inheritance check.)