Activity for Michael Zeilfelder
-
Michael Zeilfelder committed [r6730] on Code
ComboBox now marks all mouse-wheel events as handled
-
a CGUITable.cpp bug
-
Closed as like mentioned above it's not reproducible and there was no further feedback.
-
-std=gnu++26 CLMOMeshFileLoader::SObjMtl::operator=(const irr::scene::CLMOMeshFileLoader::SObjMtl&)’ is deprecated [-Wdeprecated-copy]
-
Those should have been fixed in svn 1.8 and svn trunk a while ago already. I just checked again and found some other warnings in 1.8 with newer compilers (fixed now), but that one didn't show up even with gnu++26 (and I think it was already there a few versions earlier). Not updating hc1 for now. Maybe taking another look at that one if I'm ever retired.
-
Avoid compiler warnings with newer compilers (backport from trunk)
-
Cleanup: Use some double instead of float constants in places where we work with doubles
-
Cleanup: removing some unnecessary destructors for structs
-
Minor internal string type cleanup
-
string::insert now expects template type instead of always char
-
Merging r6700 through r6723 from trunk to ogl-es branch
-
Add function aabbox3d::getPlanes
-
Comment spelling fix
-
Optimize CSceneCollisionManager::getCollisionPoint with better early out
-
Minor optimization in CTriangleSelector::getTriangles
-
Add default cases to avoid warning C26818: Switch statement does not cover all cases.
-
Add default cases to avoid warning C26818: Switch statement does not cover all cases.
-
Reduce the amount of places where const got removed with c casts
-
const IVertexBuffer::operator[] now returns a const S3DVertex reference
-
Reduce the amount of places where const got removed with c casts
-
Add flag irr::gui::EFF_HOVERED_HANDLES_UNFOCUSED_MOUSE_EVENTS to allow disabling this UI behaviour
-
Allow compiling with -Wformat-overflow=2
-
Cleanup: Remove unnecessary semicolons
-
Fix recently broken character input for some extended key characters on Windows
-
IGUIElement::isMyChild can now also check const IGUIElement* parameter
-
Inquiry Regarding Forking the Irrlicht Engine
-
The engine is (mostly) under the zlib license, so forking is absolutely allowed. Just check the license file, it 's rather short. Actually there are quite a few forks out there already. For example Luanti (formerly Minetest) use their own fork now. Or the Nabla engine once started out as a fork (originally as build-a-world engine) and has similar targets you mention here, so maybe take a look (it's very advanced and under a lot more development currently than Irrlicht, thought maybe a bit harder...
-
Fix some input key on Windows by breaking Windows 98 compatibility
-
Add AutoRepeat and Extended flags to SKeyInput to pass on that info on Windows
-
Remove unused union in ply loader
-
The loop is: for ( u32 i = 0 ; i < Rows.size() ; ++i ) So the loop itself checks 'i' already. A second check won't change anything as neither 'Rows' nor 'i' change inside the loop. The second check is a bit more interesting as the expectation is that every Row has an "Items" array which has the same size as "Columns". This also seems to be true to me at first view. But there are more places where it could in theory go wrong (but I didn't see anything right now checking this). You could add for testing...
-
The loop is: for ( u32 i = 0 ; i < Rows.size() ; ++i ) So the loop itself checks 'i' already. A second check won't change anything as neither 'Rows' nor 'i' change inside the loop. The second check is a bit more interesting as the expectation is that very Row has an "Items" array which has the same size as "Columns". This also seems to be true to me at first view, but there are more places where it could in theory go wrong (but I didn't see anything right now checking this). You could add for testing...
-
That crash seems to be inside a non-Irrlicht class (irr::core::ustring16). Not native Irrlicht at least, probably from the Irrlicht-spinoff engine you are working with.
-
That crash seems to be inside a non-Irrlicht class (irr::core::ustring16).
-
a CFileSystem bug
-
Thanks for reporting. Code changed in r6706 in svn trunk Wasn't a real problem as the unitialized variable was only used once in constructor where it was ignored (only used as return result) and afteward stuff was initialized. But cleaner to initialize the variable directly (and it won't trigger static code analysis tools that way anymore).
-
Initialize CFileSystem::FileSystemType earlier to avoid triggering code analysis tools
-
Add option to use parent transformation for absolute transformation of nodes
-
I haven't even found time yet to look at this ... (I spend my spare-time yesterday on the other bug you reported, I'm not doing Irrlicht round-the-clock or even every evening). And nothing is ever simple, especially when you have to make things work for all users insteads of just one application. My first thoughts on a very quick view go like this: - It mixes device abstraction code (Irrlicht device) with gui. Maybe can't be avoided (hard to tell without spending a lot more time on it), but it's...
-
I haven't even found time yet to look at this ... (I spend my spare-time yesterday on the other bug you reported, I'm not doing Irrlicht round-the-clock or even every evening). And nothing is ever simple, especially when you have to make things work for all users insteads of just one application. My first thoughts on a very quick view go like this: - It mixes device abstraction code (Irrlicht device) with gui. Maybe can't be avoided (hard to tell without spending a lot more time on it), but it's...
-
examples/Demo fullscreen switch X Error
-
Closing now due to lack of feedback
-
a small bug
-
OK, I fixed it in svn trunk r6704 by simply getting rid of the 'long' branch. I see no problem with that. If there ever is I guess someone can complain again. Thanks for the report! Also a tiny plea for future bug reports - in your bug report titles please don't always say something like "it's a tiny bug". Write instead what the bug is about, something like "CFileSystem::createFileList() uses wrong type on 64 bit systems". Otherwise browsing bug-reports becomes a bit confusing ;-)
-
CFileSystem::createFileList() now using intptr_t instead of long for _tfindfirst on MinGW as well
-
Yeah, I got the problem. But your solution changes Visual Studio types as well (for 32-bit) and that always used intptr_t as well for 32-bit, so while it would work (long is enough on 32-bit), it's not exactly nice. So I'm currently wondering if that branch is still needed at all. Because all headers I looked at so far use intptr_t - Microsoft headers anyway, but also the original ones from MinGW as well as the clang headers. So only reason I can think of is that there are some old MinGW headers...
-
Yeah, got it updating later on (did indeed not update for a long time and needed the -Syu) and got clang 20. And realized by now we already used intptr_t for this (I had thought earlier you introduced that, but turns out it's the one place Irrlicht uses this). But what I couldn not figure out is why we ever used long. It says it's because of mingw, but all headers I checked there so far also used intptr_t. Mabye this is something really old, I'll have to try if I can get an old 32-bit system booting...
-
In theory simple... but unfortuantely I'm getting download errors. Maybe I have to re-install my msys2 system completely or something like that. Not sure yet. Also the version it tries to download by default seems to be llvm 14 and not 20.
-
Oh, a runtime error, not a compiler error, interesting. Thanks for the info! I'll try to install Clang on Windows (but might take a few days until I get to it).
-
Can you please copy-paste me the exact error you are getting? And if you use any special compile flags for clang then please add those as well. I'd like to avoid adding intptr_t as Irrlicht otherwise still compiles without c++11 enabled.
-
Add check in quaternion::normalize() for broken quaternions
-
Add (vertex) color material option to MaterialViewer
-
a small bug
-
Thank you for the report. It's fixed now in r6700 in Irrlicht 1.8 branch and in r6701 in svn trunk.
-
Merging r6686 through r6700 from branch releases/1.8 to trunk
-
Fix CGUIPanel (in GuiEditor) so horizontal scrollbar can get focus
-
Merging r6688 through r6698 from trunk to ogl-es branch
-
Fix bug in CMeshSceneNode::render when it's called without previous OnRegisterSceneNode call
-
Fix 2 more cases of variable shadowing in CB3DMeshWriter
-
Avoid variable shadowing in CB3DMeshWriter::writeMesh
-
CIndexBuffer now uses explicit calls to it's own functions in constructor
-
Fix spelling in comments
-
No problem supporting both. Right now clang is the only compiler which sets utf-8 as default to my knowledge (and it didn't have a problem with those characters...), so independent of the year it's a bit early to assume all people use this.
-
Update changes.txt
-
compile warning caused by non-Unicode char
-
Thanks for reporting. Fixed in svn trunk r6692. Note that this bug only happens when compiling with the /utf-8 flag which is not c++ standard. So the patches you linked above will break that Irrlicht version for most users. If you check the new commit it's now only using characters which will work in both cases. Also there had been a few more (mostly the ° character). Found some nice regular expression on StackOverflow to find them all: https://stackoverflow.com/questions/64093078/how-can-i-find...
-
Remove characters in source which are not in the utf-8 range
-
Allow compiling Irrlicht with Unicode characters set without _IRR_WCHAR_FILESYSTEM
-
Unify settings for different targets in VS 2022 project file
-
Backport VS 2010 project files from 1.8 branch
-
You are probably correct. Thought strangely we do have a few Chinese users, not sure how they do it. I've never seen that patch or that Irrlicht clone. On first view I don't like the copy-paste changes in that patch too much as they break the interface instead of extending it and keeping downward compatibility. That's fine to do for your own project, but a pretty bad thing for a library. Can't say much about the changes they do with key-input without spending some time on that. Bit interesting on...
-
update zlib
-
Thanks. I've updated the branches 1.8 (r6683), trunk (r6685) and ogl-es (r6687). Caused some new C99 warnings in Android, but I guess nothing I can do about that.
-
zlib updated from 1.2.11 to 1.3.1
-
Merging r6674 through r6686 from trunk to ogl-es branch
-
Add missing override keywords in CShadowVolumeSceneNode
-
Merging r6607 through r6684 from branch releases/1.8 to trunk
-
Update changes.txt
-
Update to zlib 1.3.1
-
Add plane3d::flip function
-
NULL pointer dereference in CIrrDeviceLinux::createInputContext
-
Thank you for the detailed report. I definitely meant to close the input method and not the context. And that was also missing in the later failure case. Fixed in svn trunk r6681.
-
Fix bug #476: Crash in CIrrDeviceLinux::createInputContext
-
Add missing ILightManager::OnRenderPassPostRender calls
-
Stop OpenGL from updating enabled clipping planes every time ETS_WORLD is set
-
Bugfix: IMeshManipulater functions recalculateNormals and recalculateTangents no longer create nan's
-
Fix bad stencil draw in D3D9 when shadow nodes existed, but no shadow volumes got drawn
-
Add GLError test in COpenGLSLMaterialRenderer::init
-
Unify eol in headers
-
WebGL1 driver now draws npot textures with ETC_CLAMP_TO_EDGE
-
Merging r6659 through r6672 from trunk to ogl-es branch
-
Add isPowerOfTwo function
1 >
