[RFHS3] Remove mention of RapidYML backend, which is not supported. Due to lack of time, only one RooFit backend can be supported. On CMake configure, it, however, looked like RapidYML should be found. Here, the corresponding YML messages and mentions throughout the documentation are removed. If the backend should be revived, this commit can be reverted. Fix root-project#18128. Introduce TObject::SavePrimitiveVector Save array of doubles in C++ macro as std::vector<Double_t>. Should replace in code use of SavePrimitiveArray - while stack has limited memory and therefore plain arrays in macro can exhaust resources Use SavePrimitiveVector for TH1/TProfile classes Use SavePrimitiveVector in TGraph classes Use SavePrimitiveVector for TCutG/TScatter/TEfficiency Use SavePrimitiveVector for TF1/TF2 Use SavePrimitiveVector in TGraph2D Use SavePrimitiveVector in TPolyLine/Mraker/3D classes Remove SavePrimitiveArray Was replaced by SavePrimitiveVector. No longer used in the code, was not in released code [ntuple] Move RNTupleView out of Experimental No constraints to torch version in `requirements.txt` Now that pytorch/pytorch#138333 is fixed, commit 24fd0d8 can be reverted. [tree] check for return values in WriteBaskets Fixes https://its.cern.ch/jira/browse/ROOT-5137 [nfc] clangformat [nfc] modernize Co-authored-by: Philippe Canal <pcanal@fnal.gov> [core] escape pluses in TRegexp::MakeWildcard Fixes https://its.cern.ch/jira/browse/ROOT-7626 [Python] Format `rdf_filter_pyz.py` to pass ruff checks [Python] Fix RDF Pythonization tests with `builtin_llvm=OFF` On my machine, the RDF pythonization tests don't run when I build ROOT with `builtin_llvm=OFF`. There is a crash unless I do `import numba` in the beginning. I guess it's not a priority to understand the underlying problem because not many people build ROOT like this and the workaround is easy, but at least the tests should be green. We do the same "magically ordered imports" also in other Python tests where ROOT doesn't work together with xgboost because of `std::regexp` symbol clashes, unless you do the import in a certain order: root-project#15183 [hist] fix red highlight box in log mode Fixes root-project#12497 [ci] Disable TMVA on Alma8 [RF] Fix implementation of HS3 importers Fix by @guitargeek. The RooWSFactoryTool expression handler was not correctly matching branckets, leading to failures when importing valid JSON filed with RooProdPdf objects. [df] Avoid bad integer substitution in string formatting The '%lu' token used in RLoopManager::RunDataSource does not correspond to the correct integer representation of `std::uint64_t` on all platforms. We use a stringstream instead to construct the warning message. Introduce TObject::SavePrimitiveDraw method Store typical object->Draw() invocation. Correctly handle "nodraw" arguments in options Also replace special symbols if option string is not empty Use SavePrimitiveDraw in graf2d/graf classes Improve TGraphStruct::SavePrimitive Use new method to store constructor and attributes Simplify loops over nodes and edges Use fixed variable name for TGraphEdge Use SavePrimitiveDraw in hist classes Use SavePrimitiveDraw in other classes Fix TPave::SavePrimitive for web case In case when canvas created in the batch mode, it never rendered on server side. And thus coordinates of TPave object remains not-initialized. But there are macros where coordinates of TPaveStats are modified. Such modification stored in NDC coordinates and should be taken into account when TPave stored in C macro Adjust stressGraphics_web.ref after TPave adjustments Now positions of stats box modified in batch stored correctly [RF] fix segfault in fillLegacyCorrMatrix shown in https://root-forum.cern.ch/t/extracting-global-correlation-causes-a-seg-fault-with-sumw2error/55138 Implementing changes requested by @dpiparo [webgeom] resize workaround for qt6 Rendering inside qt5/qt6 does not provide correct size, therefore add resize handler in both cases Do not use `source_dir` in widgets This is location of JSROOT scripts, but normally one should use `import {something } from 'jsroot'` statement Now jsroot_importmap used in all widgets anyway Reduce source_dir usage in eve7 [ntuple] Move RNTupleWriter out of Experimental [hist] revert changes to v5 GetExpFormula fl_format Since it breaks backward compatibility Fixes root-project#18214 thanks guitargeek for finding out [skip-ci] cleaner docu Implementing changes requested by @dpiparo Move qt5web classes to qt6webdisplay They were used by both qt5 and qt6, now only for qt6 Completely delete qt5webdisplay sources Delete/deprecate qt5web build option Qt5 web engine no longer works with newest JSROOT and makes no sence to support. Everybody should be able to switch to comparable qt6 version Remove kQt5 from RWebDsiaplayArgs It was display kind for --web=qt5 argument Adjust tutorials/visualisation/webgui/qtweb Only support qt6 for building this tutorial Remove qt5 workaround in THttpServer For usage of qt5 webengine importmap has to be emulated Now such emulation is no longer necessary Introduce RWebDisplayArgs::GetQtEmbedQualifier Keep old GetQt5EmbedQualifier to backward compatibility, but prefer to use newer approach Use GetQtEmbedQualifier in qtweb tutorial Last place where qt5 was checked Remove qt5 from documentation and comments Update README/ReleaseNotes/v636/index.md Co-authored-by: Jonas Rembser <jonas.rembser@cern.ch> Use view framework with newest CEF Latest CEF 134 shows URL/tabs with native widgets Therefore switch to use of view framework starting from CEF 131 Update CEF readme with current CEF version Let configure usage of CEF views framework via rootrc parameter Use "WebGui.CefUseViews" parameter for this. Default is platform-dependent Mention CEF parameters in documentation [jsroot] dev 1/04/2025 1. Introduce `settings.FilesTimeout` to configure global timeout for file reading operations 2. Introduce `settings.FilesRemap` to let provide fallback address for http server, used for `root.cern` 3. Remove support of qt5 webengine, only qt6web is supported 4. Fix - correct axis range in `TScatter` drawing More use of RWebWindowWSHandler::GetBoolEnv Update docu Fix warning caused by RPadLength arithmetic operators Use `const` references as one should for `operator+` and `operator-`. This fixes the following warning I get when building ROOT with tests: ```txt In file included from /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/x86_64-unknown-linux-gnu/bits/c++allocator.h:33, from /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/allocator.h:46, from /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/memory:65, from /nix/store/qf61vad876jgyd4vnn147cz2jyvy23vl-gtest-1.16.0-dev/include/gtest/gtest.h:55, from /home/rembserj/code/root/root_src/graf2d/gpadv7/test/coords.cxx:12: In member function ‘void std::__new_allocator<_Tp>::deallocate(_Tp*, size_type) [with _Tp = double]’, inlined from ‘static void std::allocator_traits<std::allocator<_Tp1> >::deallocate(allocator_type&, pointer, size_type) [with _Tp = double]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/alloc_traits.h:544:23, inlined from ‘void std::_Vector_base<_Tp, _Alloc>::_M_deallocate(pointer, std::size_t) [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:389:19, inlined from ‘void std::_Vector_base<_Tp, _Alloc>::_M_deallocate(pointer, std::size_t) [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:385:7, inlined from ‘std::_Vector_base<_Tp, _Alloc>::~_Vector_base() [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:368:15, inlined from ‘std::vector<_Tp, _Alloc>::~vector() [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:738:7, inlined from ‘ROOT::Experimental::RPadLength::~RPadLength()’ at /home/rembserj/code/root/root_src/graf2d/gpadv7/inc/ROOT/RPadLength.hxx:31:7, inlined from ‘virtual void PadCoord_AddSubtract_Test::TestBody()’ at /home/rembserj/code/root/root_src/graf2d/gpadv7/test/coords.cxx:36:26: /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/new_allocator.h:172:33: warning: ‘void operator delete(void*, std::size_t)’ called on pointer ‘<unknown>’ with nonzero offset [1, 9223372036854775800] [-Wfree-nonheap-object] 172 | _GLIBCXX_OPERATOR_DELETE(_GLIBCXX_SIZED_DEALLOC(__p, __n)); | ^ In member function ‘_Tp* std::__new_allocator<_Tp>::allocate(size_type, const void*) [with _Tp = double]’, inlined from ‘static _Tp* std::allocator_traits<std::allocator<_Tp1> >::allocate(allocator_type&, size_type) [with _Tp = double]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/alloc_traits.h:509:28, inlined from ‘std::_Vector_base<_Tp, _Alloc>::pointer std::_Vector_base<_Tp, _Alloc>::_M_allocate(std::size_t) [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:380:33, inlined from ‘void std::_Vector_base<_Tp, _Alloc>::_M_create_storage(std::size_t) [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:398:44, inlined from ‘std::_Vector_base<_Tp, _Alloc>::_Vector_base(std::size_t, const allocator_type&) [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:334:26, inlined from ‘std::vector<_Tp, _Alloc>::vector(const std::vector<_Tp, _Alloc>&) [with _Tp = double; _Alloc = std::allocator<double>]’ at /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/stl_vector.h:603:61, inlined from ‘ROOT::Experimental::RPadLength::RPadLength(const ROOT::Experimental::RPadLength&)’ at /home/rembserj/code/root/root_src/graf2d/gpadv7/inc/ROOT/RPadLength.hxx:31:7, inlined from ‘virtual void PadCoord_AddSubtract_Test::TestBody()’ at /home/rembserj/code/root/root_src/graf2d/gpadv7/test/coords.cxx:36:26: /nix/store/aw0qxjd1phf16qhlwpdb4x87yymfv9rp-gcc-14-20241116/include/c++/14-20241116/bits/new_allocator.h:151:55: note: returned from ‘void* operator new(std::size_t)’ 151 | return static_cast<_Tp*>(_GLIBCXX_OPERATOR_NEW(__n * sizeof(_Tp))); | ^ ``` Update WebDisplay.md file Modify RGeomViewer::SaveAsMacro Do not use RDirectory::Heap(). Just create shared_ptr variable outside of function scope Use "\n" instead of std::endl [Python] Correctly manage NumPy dependencey in pythonization tests The dependency on Python libraries like NumPy needs to be declared in the CMakeLists.txt, such that the test gets disabled when the library is not available. In some cases, the dependency on NumPy could be avoided by using the builtin "array" library. [Python] Apply suggestions by ruff [CPyCppyy] Don't use deprecated `PyErr_Fetch` and `PyErr_Restore` API Also, in some places the same logic could be implemented with just `PyErr_Clear()`. [Python] Update `sync-upstream` script for cppyyy syncups SearchInstalledSoftware.cmake: correctly report name of option to disable [skip-ci] cpu backend specify it's single threaded See root-project#17344 [DF] Retrieve RDatasetSpec for RDF built from JSON [CPyCppyy] Drop `__array__` from std::vector pythonizations The addition of the __array__ utility to std::vector Python proxies causes a bug where the resulting array is a single dimension, causing loss of data when converting to numpy arrays, for >1dim vectors. The recursive nature of this function, passes each subarray (pydata) to the next call and only the final buffer is cast to a lowlevel view and resized (in VectorData), resulting in only the first 1D array to be returned. See root-project#17729 Since this C++ pythonization was added with the upgrade in 6.32, and is only defined and used recursively, the safe option is to disable this function and no longer add it. It is temporarily removed to prevent errors due to -Wunused-function Revert "[ci] Disable TMVA on Alma8" This reverts commit ebab348. Protect PyMva tutorials for a failurte in importing tensorflow Disable also batchgenerator tests for Python version less than 3.9 since tensorflow is not working for these lower python versions Fix build issue on macOS 15.4 / XCode 16.3 `less` is defined in functional according to the C++ standard. cling: Add ROOT lock to LookupHelper::findType This fixes root-project#18236 As seen in cms-sw/cmssw#47763 (comment) there are cases where the ROOT global lock is taken too late: In the stack trace below the lock is requested on frame root-project#7 where as it should really be (also) requested on frame root-project#12 as soon as the transaction is being generated. ``` root-project#6 0x00001501421f55ea in ROOT::TVirtualRWMutex::Lock() () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCore.so root-project#7 0x0000150133a680ca in TCling::HandleNewTransaction(cling::Transaction const&) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCling.so root-project#8 0x0000150133a85778 in TCling::UpdateListsOnCommitted(cling::Transaction const&) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCling.so root-project#9 0x0000150133b60018 in cling::MultiplexInterpreterCallbacks::TransactionCommitted(cling::Transaction const&) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCling.so root-project#10 0x0000150133bec069 in cling::IncrementalParser::commitTransaction(llvm::PointerIntPair<cling::Transaction*, 2u, cling::IncrementalParser::EParseResult, llvm::PointerLikeTypeTraits<cling::Transaction*>, llvm::PointerIntPairInfo<cling::Transaction*, 2u, llvm::PointerLikeTypeTraits<cling::Transaction*> > >&, bool) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCling.so root-project#11 0x0000150133b5663a in cling::Interpreter::PushTransactionRAII::~PushTransactionRAII() () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCling.so root-project#12 0x0000150133b6b883 in cling::LookupHelper::findType(llvm::StringRef, cling::LookupHelper::DiagSetting) const () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCling.so root-project#13 0x00001501339b6b06 in ROOT::TMetaUtils::TClingLookupHelper::GetPartiallyDesugaredNameWithScopeHandling(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&, bool) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCling.so root-project#14 0x00001501422f1458 in ResolveTypedefProcessType(char const*, unsigned int, unsigned int, bool, unsigned int, unsigned int, unsigned int, bool&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) [clone .constprop.0] () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCore.so root-project#15 0x00001501422f1a0f in ResolveTypedefImpl(char const*, unsigned int, unsigned int&, bool&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCore.so root-project#16 0x00001501422f2ef3 in TClassEdit::ResolveTypedef[abi:cxx11](char const*, bool) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCore.so root-project#17 0x00001501422f5c20 in TClassEdit::TSplitType::ShortType(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&, int) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCore.so root-project#18 0x00001501422f6888 in TClassEdit::GetNormalizedName(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&, std::basic_string_view<char, std::char_traits<char> >) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCore.so root-project#19 0x000015014232e795 in ROOT::Internal::GetDemangledTypeName[abi:cxx11](std::type_info const&) () from /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02883/el8_amd64_gcc12/cms/cmssw-patch/CMSSW_15_1_RNTUPLE_X_2025-03-31-2300/external/el8_amd64_gcc12/lib/libCore.so ``` [ntuple] Add tests and a warning for projected Real32(Quant|Trunc) Setting a projected field to quantized or truncated works and read proper data but it doesn't do what one would probably expect. The projected field can correctly read values from any fields that can be read as float, but the value range and bits of precision set on the projected field are silently ignored (since they only make sense for physical columns that are actually stored on disk). To reduce confusion, we now emit a warning if a user tries to call AddProjectedField with a quantized or truncated field. Remove current RHist prototype For a number of reasons, it is not what we want for the future generation of histogram classes in ROOT: * It is templated on the DIMENSIONS, the PRECISION, and optional bin statistics. This makes it hard(er) to use for IO. * The RHist interface has a pointer to an abstract, polymorphic RHistImplBase. * The concrete RHistImpl is templated on the processed axis types. * The polymorphic pointer makes the current prototype incompatible with RNTuple (unless using streamer fields). * It also takes minutes to compile for higher dimensions because concrete RHistImpl have to be generated for all combinations of axis types, which grows exponentially. The plan is to remove the current prototype now, in time for the next ROOT release v6.36, and then add a new implementation after. [hist] reset kAxisRange bit when fully unzooming (root-project#18226) When unzoom with mouse wheel, properly reset kAxisRange bit. Adjust documentation [graf] Patch libAsimage: fp.h is now part of math.h Thanks to @ktf (ALICE) for this fix. [runtime-cxxmodules] Adapt the darwin modulemaps to sdk 15.4. (root-project#18231) * [runtime-cxxmodules] Adapt the darwin modulemaps to sdk 15.4. * [runtime-cxmodules] Use separate modulemap for macOS15.2 * [runtime-cxxmodules] Add overlay entry for MacOSX15.2 --------- Co-authored-by: Devajith Valaparambil Sreeramaswamy <devajith.valaparambil.sreeramaswamy@cern.ch> [skip-CI][cling] Refine build and usage guide (root-project#18161) * [cling] Refine build and usage guide Clarified standalone and LLVM-integrated build methods, along with usage instructions. Improved installation steps to ensure successful execution. * [cling] Fix broken link to web page in README Updated the "See also" link in the README to point to the latest version of Cling's web page on GitHack. The previous link referenced a specific commit, which could become outdated. Now, it directly points to the master branch for a more up-to-date reference. * [cling] Improve Cling standalone build instruction Refined the README to provide clearer instructions for building Cling as a standalone project. Updated CMake configuration details, clarified dependency setup, and provided an example command to help users correctly specify LLVM build paths. These improvements aim to make the build process more comprehensible and reduce potential confusion. * [skip-ci] update link to binaries * [skip-ci] fix jupyter link Fixes root-project/cling#527 * [skip-ci] also mention jupyter notebook option see root-project/cling#537 --------- Co-authored-by: ferdymercury <ferdymercury@users.noreply.github.com> [ntuple] Move RNTupleReader out of Experimental [rfile] Remove current RFile prototype It will replaced by a new prototype after 6.36. The new prototype will go in a different direction by removing the shared ownership between RFile and the application and it won't have RFile inherit from RDirectory. [hist] Add function to get sum of weights including under/overflows Fix root-project#8951 Co-authored-by: Stephan Hageboeck <stephan.hageboeck@cern.ch> [cling][JIT] Inject compiler-rt complex division symbols on macOS Clang may emit calls to compiler-rt symbols such as __divdc3 for std::complex<T> division. These may not be resolved during JIT'ing on macOS when using Cling. This patch forward-declares the relevant symbols and explicitly injects them into the JIT symbol table. Similar to: root-project@4b6075b [rbrowser] do not allow to select text in items list [rbrowser] implement items drag and drop Drop handler behaves very similar to drawing with append flag, but does not try to create new widget [rbrowser] let switch between double-click and plain click handler Double-click is ROOT-like style of drawing while single-click is openui5. In last case one also can use keyboard navigation for drawing [rbrowser] Change columns context menu Old functionality deprecated, one need to have something else. Assign to each column headmenu instead [rbrowser] let resize columns via context menu Add 'Resize colums' to context menu [jsroot] dev 4/04/2025 1. Set 'user-select: none' style in drawings to exclude text selection, using `settings.UserSelect` value 2. Better handling of remap of file URL [cmake] Update the version of VDT to 0.4.6 to address the new behaviour of the compiler shipped with XCode 16.3 Properly format the preprocessor if clause following pnggroup/libpng@893b811 [interop] Upgrade to latest, add tag file for diff workflow [ntuple] Mark some CreateEntry const RNTupleReader::CreateEntry() needs to call GetModel(), which is not const. [ntuple] Move RFieldToken out of REntry It will also be used by RRawPtrWriteEntry. [ntuple] Refactor RNTupleFillContext::Fill* ... to avoid code duplication when adding RRawPtrWriteEntry. [ntuple] Implement Detail::RRawPtrWriteEntry It is a container of const raw pointers that can be used to write constant data products in frameworks. Closes root-project#17900 [ntuple] Demonstrate RRawPtrWriteEntry in ntpl014_framework.C [ntuple] Amend documentation on model ID check [tmva][sofie] memory optimization in convK and imcol for ConvTranspose (root-project#18168) * feat: add convK and imcol to memory optimization for ConvTranspose * fix: follow root naming convention for convK and imcol [tmva][sofie] Optimize LayerNormalization operator Do square operation do not use std::pow with base 2 since it is slower than simple multiplication This speeds up evaluation on ATLAS GNN tracking model by 40% Also avoids allocating a small shape vector in LayerNorm and improve indentation of sigmoid operator [TMVA] Disable relevant PyMVA tests if BLAS is not found This commit makes sure that the tests of PyMVA that require BLAS (indirectly via SOFIE) are not run when BLAS is not available. Following up on this PR, which did the same change already for the SOFIE tests: root-project#18065 [CMake] Veto tutorials depending on BLAS if it was not found Closes root-project#16720. Fix a warning when compiling quartz on macos 15.4 with clt 16.3 [tree] Avoid reading from nullptr in TTreeReader. When a TTreeReaderValue is not initialised correctly, it generates an error message. It could, however, inadvertently read a nullptr in doing that. Fix memory leak identified with valgrind in HistFactory histogram reading Apply clang-format [core] improve TDirectory::DecodeNameCycle - avoid scanning the string twice - allow `name` to be null if one only wants the cycle - return the actual name length so the caller can know whether the name was truncated or not - improve function documentation [cppyy] Disable warnings for invalid function casts Some C++ functions defined in CPyCppyy are passed to CPython API, e.g. to build Python methods that use some C++ functionality. This is done by creating a instance of a PyMethodDef struct (https://docs.python.org/3/c-api/structures.html#c.PyMethodDef). One data member of this struct is of type PyCFunction (https://docs.python.org/3/c-api/structures.html#c.PyCFunction), a typedef of a function with signature ``` PyObject *PyCFunction(PyObject *self, PyObject *args); ``` In many cases, the C++ functions that are used as the data member of the PyMethodDef are not directly implemented as PyCFunction, thus need to be cast. This cast is often invalid, since many of such functions do not really respect the function signature prescribed by the API. This pattern is actually encouraged for CPython extension implementations by the official CPython docs (https://docs.python.org/3/extending/extending.html#keyword-parameters-for-extension-functions). As such, the compiler warnings that are generated becausee of the invalid function casts, for example ``` root/bindings/pyroot/cppyy/CPyCppyy/src/Pythonize.cxx:1949:50: warning: cast from 'PyObject *(*)(PyObject *)' (aka '_object *(*)(_object *)') to 'PyCFunction' (aka '_object *(*)(_object *, _object *)') converts to incompatible function type [-Wcast-function-type-mismatch] 1949 | Utility::AddToClass(pyclass, "__repr__", (PyCFunction)ComplexRepr, METH_NOARGS); ``` cannot be avoided and have to be suppressed. This is also done upstream, see wlav/CPyCppyy@33d4a6e Implementing requested changes by @vepadulano [tree] Reset missing proxies list when changing trees. When changing trees while the missing value feature of RDF was in use, the proxies for branches could not get recreated, since TTreeReader refuses the registration when fProxiesSet == true. This lead to the TTreeReaderValue being in a broken state. [tree] Refactor vector of missing proxies to a set. When running over multiple trees, the list of missing proxies in TTreeReader will grow indefinitely. Here, it is refactored to a set, which also shortens the code that looks up the missing proxies. [tree] Include branch name in an error message. Pointed out by Philippe in root-project#18259. [tree] Make test of error messages a bit more rigid. In debugging a UB in TTreeReaderValueBase, it would have been easier to have an explicit note of the expected error messages of that test. Instead of globally suppressing all errors, the three expected message were added here. [tree] Don't dereference a nullptr in TTreeReaderValue test. On mac15 beta, this generated a SIGTRAP, while it was OK on all other platforms, because the reference was never used. But retrieving the pointer instead of dereferencing, the problem seems to go away. [CMake] Restore Python tutorial dependencies. Due to a missing quote, dependencies on Mac15 beta seemed to have gotten deleted. [ntuple] apply I/O rules also to persistent members [ntuple] test I/O rule of persistent member [json] check if map class has dictionary When stream directly std::map object in macro like: ``` std::map<int,string> m; m[1] = "number 1"; m[2] = "number 2"; auto json = TBufferJSON::ToJSON(&m); ``` Real dictionary class for std::map is required. It checked automatically when map is member of user class, but was not checked when map object intself streamed [json] check map dictionary also when read object Same when write object, reading is not possible for std::map<> object without dictionary [json] mention GenerateDictionary for map classes in documentation While std::map uses in TBufferJSON in documentaiton, mention gInterpreter->GenerateDictionary [CMake] Improve handling of RPATH options Some improvements to the handling of RPATH options, which are necessary on my system because I need to add a custom RPATH to make my CUDA package from nix packages findable. 1. Don't set CMake variables to `FALSE` if the default is `FALSE` already. This makes it impossible to change these variables from the CMake user configuration, unnecessarily taking flexibility out of the build system. 2. If `rpath=ON`, ROOT adds the relative path from the binary directory to the lib directory to the RPATH. But it doesn't append it, overriding any `CMAKE_INSTALL_RPATH` that the user might set in the CMake configuration. This commit suggests to append instead. 3. If `rpath=OFF`, ROOT sets `CMAKE_SKIP_INSTALL_RPATH` to `TRUE` without any apparent reason, again, preventing the user from defining the RPATH via `CMAKE_INSTALL_RPATH`. This commit suggests to remove that line. [cppyy] Do not use non-C++ GCC warning Although `bad-function-cast` is present for both Clang and GCC, on the latter it is only used for C and Objective-C code. Using `no-cast-function-type` seems to already disable the PyCFunction-cast-related warnings seen with AppleClang. [ntuple] update RNTupleModel's documentation [math] Ignore vla-cxx-extension in case veccore is active Reintroduce rcanvas_mt.cxx demo with TH1 classes Old macro was used RH1D classes But RCanvas multithreading can be tested without it [rbrowser] reintroduce provider for RCanvas class RBrowser has to know about RCanvas class and which library to load to get access to that class [jsroot] dev 4/04/2025 with RPad fix Fix redraw of RPad primitives Adjust RAxis drawing Adding suggestions from @ferdymercury

Web page does not display this as a link. It appears to be using .md format rather than <a herf> #527

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions