October 2026 - Wine-Bugs - list.winehq.org

[Bug 60554] Continuum 0.40 hangs with 100% CPU when joining a zone

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60554 --- Comment #1 from winehq-bugzilla(a)tobias.preclik.de --- Created attachment 82422 --> http://bugs.winehq.org/attachment.cgi?id=82422 relay log excerpt (registry calls) -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 60554] New: Continuum 0.40 hangs with 100% CPU when joining a zone

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60554 Bug ID: 60554 Summary: Continuum 0.40 hangs with 100% CPU when joining a zone Product: Wine Version: 6.15 Hardware: x86-64 OS: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: advapi32 Assignee: wine-bugs(a)list.winehq.org Reporter: winehq-bugzilla(a)tobias.preclik.de Target Milestone: --- Distribution: --- Created attachment 82421 --> http://bugs.winehq.org/attachment.cgi?id=82421 terminal output Continuum 0.40 installs and starts fine and shows the zone list. As soon as you join a zone, the window stops responding and Continuum.exe uses 100% of one CPU core. Steps to reproduce: 1. Download https://www.getcontinuum.com/downloads/continuum/Continuum040Setup.zip and unzip it. SHA1 of the zip: d95283d34d9a88eadbc4566abd0a411fb9da8953 (Continuum040Setup.zip) SHA1 of the installer: c98e42a92b1f2c3233bf89d597cb22b0162b2668 (Continuum040Setup.exe) 2. `wine Continuum040Setup.exe` and accept the defaults. 3. Start Continuum.exe, create or log in with a user name, select any zone and click Play. Analysis: when entering a zone, the client enumerates the registry. Opening HKLM\Software\Microsoft\Windows NT\CurrentVersion\Perflib\009 returns the pseudo-handle HKEY_PERFORMANCE_TEXT (0x80000050). RegEnumValueA() on that handle returns ERROR_INVALID_HANDLE for every index, and Continuum only stops on ERROR_NO_MORE_ITEMS, so it keeps incrementing the index forever (see the attached relay log: about 3 million calls before the log was stopped). The todo_wine tests in test_performance_keys() (dlls/advapi32/tests/registry.c) expect ERROR_MORE_DATA from RegEnumValueA() and ERROR_NO_MORE_ITEMS from RegEnumKeyA() on these handles. Regression note: not bisected. The handle has been returned since 1ab13117081 ("kernelbase: Return HKEY_PERFORMANCE_TEXT when trying to open the Perflib\009 key.", wine-6.15). This matches ProtonDB reports, where Proton 5.13 works and later versions hang when joining a server. Tested on Ubuntu 22.04 x86_64 with wine-11.20 built from the release tag, in a clean prefix with a fresh install, and earlier with wine-11.0 from the WineHQ stable packages. It hangs in both. The terminal output only contains a few fixme lines (attached). -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 27559] ImgBurn 2.5.5.0 and earlier needs shell32.SHGetKnownFolderIDList() implementation when run with WinVer >= Vista

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=27559 Bernhard Übelacker <bernhardu(a)mailbox.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |bernhardu(a)mailbox.org --- Comment #31 from Bernhard Übelacker <bernhardu(a)mailbox.org> --- Created attachment 82420 --> http://bugs.winehq.org/attachment.cgi?id=82420 shell32: Respect the flag KF_FLAG_NO_ALIAS in SHGetKnownFolderIDList. By attempting to investigate bug #60540, I got also this error: err:virtual:virtual_setup_exception stack overflow 1472 bytes ... I want to share a few interesting things: ImgBurn.exe seems to be UPX compressed. By "upx -d ImgBurn.exe" the uncompressed executable can be retrieved and with that gdb is able to show a slightly better backtrace: (rr) bt 0 0x0089813c in ImgBurn!@Mpshellutilities@TNamespace@GetNameForParsing$qqrv 1 0x0089815f in ImgBurn!@Mpshellutilities@TNamespace@GetNameForParsing$qqrv ... 5 0x0089815f in ImgBurn!@Mpshellutilities@TNamespace@GetNameForParsing$qqrv 6 0x0089114e in ImgBurn!@Mpshellutilities@PotentialMappedDrive$qqrp27Mpshellutilities@TNamespace 7 0x0089cefa in ImgBurn!@Mpshellutilities@TNamespace@GetValid$qqrv 8 0x009c532c in ImgBurn!@@Updownedit@Finalize ... 31 0x0070c1e3 in ImgBurn!@@Imagequeue@Finalize 32 0x00a80acf in madTraceProcess 33 0x00000000 in ?? (rr) This GetNameForParsing line gets invoked over and over again until the stack is exhausted. The source of it or a similar version seems to be available here: https://github.com/TurboPack/MustangpeakCommonLib/blob/master/Source/MPShel… This shows if the "IsDesktop" call returns true another call to NameForParsing/GetNameForParsing, that seems to match our recursion. In the end, this "IsDesktop" function seems to be only a check, if "Result := APIDL.mkid.cb = 0". https://github.com/TurboPack/MustangpeakCommonLib/blob/master/Source/MPComm… Then I noticed this bug report is about SHGetKnownFolderIDList, which threw a warning to me in the console output: fixme:shell:SHGetKnownFolderIDList unsupported flags: 0x00001000 This flag is KF_FLAG_NO_ALIAS and it looks like this is where Wine behaves differently than native. Wine always returns the pidl generated by _ILCreateDesktop, which has mkid.cb == 0. With attached patch applied, versions 2.5.5.0 and 2.5.8.0 do no longer show this stack overflow, maybe you can verify. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 60115] d2d_cdt_insert_segment() recurses without a global termination guard, stack overflow on degenerate geometry

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60115 --- Comment #4 from Seth2313(a)gmail.com <Seth2313(a)gmail.com> --- Thank you - that's exactly the independent confirmation this bug needed. Summary for whoever picks this up: - Reproduced on two unrelated setups: NVIDIA (proprietary driver, wine-staging 11.19/11.20) and AMD RX 9070 XT (Mesa/RADV, kron4ek 11.15). So it is not driver- or build-specific. - Triggers in Serum 2: an LFO modulating an EQ parameter while a note plays, and dragging the wavetable position. On the second setup Serum 1 also crashed as soon as its window opened. - Bridging the plugin in FL doesn't help (same result here). - With the patch, built from source on both setups, all of these are gone, without visible rendering artefacts. Unless someone prefers otherwise, my plan is to split the patch into one commit per defect (insert_segment, cut_edges, fixup, resolve_beziers, plus the optional hollow-figure change kept separate), add regression tests derived from the fuzzer seeds to dlls/d2d1/tests, and open a merge request on gitlab.winehq.org referencing this bug. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 60115] d2d_cdt_insert_segment() recurses without a global termination guard, stack overflow on degenerate geometry

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60115 --- Comment #3 from synatedic(a)gmail.com --- Confirmed working. I built d2d1.dll from your patch against vanilla wine-11.19 (applied cleanly, no fuzz) and dropped it into my Bottles prefix running kron4ek-wine-11.15-amd64 with d2d1 = native,builtin. You asked whether it survives on that runner — it does. Setup: CachyOS (Arch), AMD RX 9070 XT, Mesa/RADV, DXVK and VKD3D enabled, 64-bit prefix, Windows 11 mode, FL Studio 2025, Serum 2 (VST3). Before the patch: Serum 2: crash when assigning an LFO to an EQ parameter and playing a note, and when dragging the wavetable position on an oscillator. Static GUI rendering was fine. Serum 1: crashed immediately on opening the plugin window, so the whole Serum line was unusable here. Bridging the plugin in FL did not contain it — FL went down with it either way. After the patch: both of the Serum 2 triggers above are gone, and Serum 1 opens normally. No visual artifacts in the plugin UI that I can see. Thanks for the patch and the detailed writeup — the step-by-step in the repo made it straightforward to build this myself rather than use the prebuilt DLL. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 60115] d2d_cdt_insert_segment() recurses without a global termination guard, stack overflow on degenerate geometry

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60115 --- Comment #2 from Seth2313(a)gmail.com <Seth2313(a)gmail.com> --- hanks for confirming, and especially for the steps - "LFO assigned to an EQ parameter in Serum 2, then play a note" is the most reliable trigger anyone has reported for this bug. Your log matches mine exactly: hundreds of "Triangle not found" lines followed by the stack overflow at stack 0x108c0 (0x10000-0x11000-0x110000). The animation matters because every redraw builds new path geometry, so every frame is another chance to hit a degenerate one. UPDATE SINCE THE ORIGINAL REPORT 1. Still present upstream. dlls/d2d1/geometry.c is byte-identical in 11.15, 11.19 and 11.20, so your build and current Wine are affected alike. 2. The workaround in the description is not enough, and it has a bug of its own. Fuzzing the triangulation with random degenerate paths found four more places that don't terminate or crash: - d2d_cdt_insert_segment: the skip paths of my first workaround returned TRUE without setting *edge, so the caller carried on with a stale or freed edge (page fault). Fixed by returning a valid edge at the target vertex, or failing cleanly. Please don't use the snippet from the description any more. - d2d_cdt_cut_edges: unbounded non-tail recursion on cyclic topology (stack overflow). Now iterative - collect the crossed edges, delete them in reverse order (same order as the recursion), bounded by edge_count. - d2d_cdt_fixup: the face walk never gets back to its origin (infinite loop). Now bounded, returns FALSE like the existin path. - d2d_geometry_resolve_beziers: overlapping control forever and memory grows without bound. Now a budg 16 * bezier_count + 1024 splits per geometry; any gives a slightly imprecise fill instead of a hang. Every bail-out ends in an existing error path (one segment skipped, or that single geometry's fill fails with E_FAIL) - nev 3. Results: stock 11.19 overflows the stack or hangs in 10,000 fuzz seeds; with the patch, 240,000 seeds run conformance tests (17,475 checks) give identical res without the patch. REPRODUCER WITHOUT FL STUDIO The original report had no minimal test case. There is one now: tools/d2d1_fuzz.cpp in https://github.com/Sethtekk/fl-studio-wine-linux It is a ~30-line MinGW program that builds random fille (many duplicate and collinear points, 1-4 figures, line fill modes) and closes them. It prints "seed N" before last line before a crash or hang identifies the input. x86_64-w64-mingw32-g++ -O2 d2d1_fuzz.cpp -ld2d1 -stat wine d2d1_fuzz.exe 1 10000 THE FIX Patch against vanilla 11.19 (applies unchanged to 11.15 https://github.com/Sethtekk/fl-studio-wine-linux/bloby-fix.patch Technical notes per defect, measurements and test-suite https://github.com/Sethtekk/fl-studio-wine-linux/bloby-fix.md The patch also contains an optional, separate hunk that d2d_geometry_intersect_self() for D2D1_FIGURE_BEGIN_HOL strokePath() creates one on every call and only draws i micro-benchmark that cuts Close() from 3.32 ms to <0.01 but inside FL Studio I could not measure a difference, so it is not part of the bug fix. I'm happy to split this into separate patches and turn seeds into conformance tests for dlls/d2d1/tests if tha upstream. USING THE FIX TODAY (ANY WINE PREFIX) Until this is fixed upstream, the patched DLL can be dr existing 64-bit prefix. d2d1 is a pure PE DLL, so it do exact Wine build: the DLL built from 11.19 runs here on and the earlier one built from 11.9 ran unchanged up to version of the fix has been in daily use with FL Studio May, the current one since mid-October. Easiest way, with the script from the repository (backs DLL, installs the patched one and sets the override): git clone https://github.com/Sethtekk/fl-studio-wine- cd fl-studio-wine-linux WINEPREFIX=/path/to/your/prefix bash scripts/04-apply Or by hand - prebuilt x86_64 DLL from 11.19 with the pa fd197f30b36451714e5fc5a7afc148c9: https://github.com/Sethtekk/fl-studio-wine-linux/raw/ed cp $WINEPREFIX/drive_c/windows/system32/d2d1.dll d2d1 cp d2d1.dll.patched $WINEPREFIX/drive_c/windows/syste WINEPREFIX=... wine reg add 'HKCU\Software\Wine\DllOvuiltin /f The "Wine builtin DLL" marker at offset 0x40 is already If you build the DLL yourself, zero it, otherwise Wine copy: dd if=/dev/zero of=d2d1.dll bs=1 seek=64 count=16 conv=notrunc To confirm the guard is working, run once with WINEDEBU "Iteration limit exceeded, skipping segment" lines inst it caught the degenerate geometry. @synatedic - in Bottles it is the same: copy the DLL into the bottle's drive_c/windows/system32 and add d2d1 = native,builtin DLL overrides. It should work with your kron4ek 11.15 r haven't tested that exact runner, so I'd be glad to hea test survives with it. Two related notes: - Wine's d2d logs "fixme:d2d:d2d_device_context_DrawGeo style" for every line Serum 2 draws (I measured ~11,000 lines/s with the plugin window open). Not a crash, but the logging cos WINEDEBUG=fixme-all is worth setting. - I haven't tested Serum 2 with DXVK disabled, so I can part. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 59885] Sonic Boom (1995 Klik & Play fangame) locks up in new WoW64 build

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=59885 --- Comment #5 from Janne <janne.kekkonen(a)gmail.com> --- Hello, I retested this with Wine 11.20 on fresh prefixes. In all prefixes, the game installed successfully and I was able to play it. During installation, the following lines were printed to the console: 0134:fixme:nstc:NSTC2_fnSetControlStyle2 mask & style (0x00000004) contains unsupported style(s): 0x00000004 0134:fixme:shell:IShellBrowser_fnSendControlMsg stub, 000000000081F860 (2, 1026, a003, 0, 000000000021F378) 0134:fixme:shell:IShellBrowser_fnSendControlMsg stub, 000000000081F860 (2, 1026, a004, 1, 000000000021F378) 0134:fixme:shell:IShellBrowser_fnSendControlMsg stub, 000000000081F860 (2, 1025, a003, 1, 000000000021F378) 0134:fixme:shell:IShellBrowser_fnSendControlMsg stub, 000000000081F860 (2, 1025, a004, 1, 000000000021F378) WINEARCH options tested: win32, win64, and wow64 (I have no idea if wow64 differs from win64). The prefixes were created using these commands: WINEPREFIX=~/11.20win32 WINEARCH=win32 winecfg WINEPREFIX=~/11.20win64 WINEARCH=win64 winecfg WINEPREFIX=~/11.20wow64 WINEARCH=wow64 winecfg -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 60115] d2d_cdt_insert_segment() recurses without a global termination guard, stack overflow on degenerate geometry

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60115 synatedic(a)gmail.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |synatedic(a)gmail.com --- Comment #1 from synatedic(a)gmail.com --- Created attachment 82419 --> http://bugs.winehq.org/attachment.cgi?id=82419 Wine console log: FL Studio 2025 + Serum 2 (VST3) crash on animated LFO rendering Confirming this on Wine 11.15 (kron4ek build) running Serum 2 (VST3) inside FL Studio 2025 under Bottles/Flatpak on Arch (CachyOS), 64-bit prefix, Windows 11 mode, DXVK and VKD3D enabled. Steps to reproduce: Load FL Studio, add Serum 2 as a VST3 instrument. Assign an LFO to an EQ parameter inside Serum. Play a note so the LFO animation starts rendering. Result: immediate crash. Static GUI rendering works fine — the crash only occurs when the modulation animation redraws continuously. Log output: hundreds of repeated err:d2d:d2d_cdt_insert_segment Triangle not found. lines, followed by: err:virtual:virtual_setup_exception stack overflow 1856 bytes addr 0x6ffffff75c2a stack 0x108c0 (0x10000-0x11000-0x110000) err:seh:call_seh_handlers invalid frame 00007FFFFE21E210 (0000000001D32000-0000000001E30000) err:seh:NtRaiseException Exception frame is not in stack limits => unable to dispatch exception. Switching the wavetable display from 3D to 2D does not help. Disabling DXVK/VKD3D prevents Serum 2 from loading at all. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 60525] FindVUK breaks in Wine 11.20 - SetFilePointer fails on Blu-ray/DVD drive

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60525 Zeb Figura <z.figura12(a)gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Ever confirmed|0 |1 Status|UNCONFIRMED |NEW --- Comment #6 from Zeb Figura <z.figura12(a)gmail.com> --- Okay, I see the problem, I tested using an unaligned size. Passing 2048 makes things work. I'll fix this anon. -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026

[Bug 60539] Mod Organizer 2 needs RtlIsNameInExpression()

by WineHQ Bugzilla

http://bugs.winehq.org/show_bug.cgi?id=60539 Zeb Figura <z.figura12(a)gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Summary|ntdll: |Mod Organizer 2 needs |RtlIsNameInExpression is |RtlIsNameInExpression() |not exported | -- Do not reply to this email, post in Bugzilla using the above URL to reply. You are receiving this mail because: You are watching all bug changes.

Oct. 21, 2026