| f9b2882 | Philippe Canal | 19 March 2021, 18:07:02 UTC | Add missing use of ::ROOT to qualify name in dictionary | 22 March 2021, 20:18:35 UTC |
| 66e8a63 | Vassil Vassilev | 19 March 2021, 22:06:59 UTC | [cmake] Do not bind cling's sysroot to a particular version of osx sdk. The CMAKE_OSX_SYSROOT exports the exact version of the sdk which we build against. However, this means that binary releases become sensitive to minor sdk upgrades (eg. MacOSX11.1.sdk -> MacOSX11.2.sdk). Our infrastructure is relatively resilient to such changes. This patch introduces a workaround to address this issue -- it uses the fact that the current sdk's parent directory has a symlink MacOSX.sdk which points to the current sdk. This should resolve root-project/root#7021. | 22 March 2021, 13:21:15 UTC |
| 7b96eef | Jonas Hahnfeld | 17 March 2021, 11:56:02 UTC | Fix unwind info relocation with large code model on AArch64 Backport of D27629, commit 18805ea951. Original commit message: --- Makes sure that the unwind info uses 64bits pcrel relocation if a large code model is specified and handle the corresponding relocation in the ExecutionEngine. This can happen with certain kernel configuration (the same as the one in https://reviews.llvm.org/D27609, found at least on the ArchLinux stock kernel and the one used on https://www.packet.net/) using the builtin JIT memory manager. Co-authored-by: Yichao Yu <yyc1992@gmail.com> Co-authored-by: Valentin Churavy <v.churavy@gmail.com> --- Note: The handling in ExecutionEngine was committed in a different revision and is already part of LLVM 9. We need the part about emitting relocations because eh_frame (allocated in a data section) may be more than 4Gb away from the code section it references. See the discussion in https://github.com/cms-sw/cmssw/issues/31123 for context. (cherry picked from commit f481e8fe6df177ec1a50a10b7d76df5fb0b5c2a2) | 19 March 2021, 19:37:57 UTC |
| 0361339 | Axel Naumann | 17 December 2020, 17:41:51 UTC | [cling] In failed lookup, unload only decls produced, not existing fwd decl: Before, pre-existing fwd decls of specializations got unloaded. OTOH, any decl produced during (failed) template instantiation must also be unloaded; see #6331. Those are easiest identified by unloading the whole (failed) transaction. ``` error: no member named 'value' in 'std::__and_<std::is_copy_assignable<std::__cxx11::basic_string<char> >, std::is_copy_assignable<Inner<int> > >' is_copy_assignable<_T2>>::value, ~~~~~~~~~~~~~~~~~~~~~~~~~~^ ``` which is due to the first decl being invalid (as `Inner<int>` does not have a deinition and a `static_assert` /usr/lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/type_traits:1093 being triggered), then not unloaded, and then picked up again where we *do* have the definition for `Inner<int>`. Add test for templt spec lookup vs unloading. | 19 March 2021, 10:11:55 UTC |
| 3c3de4c | Axel Naumann | 18 December 2020, 16:24:36 UTC | [cling] PushTransactionRAII: do not commit broken transactions: There is no point in emitting a transaction that has been rolled back to CodeGen. To make matters worse, a rolled-back transaction will not show up in the list of transactions. | 19 March 2021, 10:11:55 UTC |
| 95a439a | Axel Naumann | 18 December 2020, 16:02:55 UTC | [cling] Allow unloading of Transactions that have not been committed: If Lookup fails to instantiate, its transaction should not be committed, as we know it's useless and must be unloaded as it might contain invalid Decls. But that means Lookup needs to unload a non-committed transaction. | 19 March 2021, 10:11:55 UTC |
| 8a1ff4a | Axel Naumann | 25 November 2020, 17:52:36 UTC | [metacling] Remember which classes were DeepAutoLoad-ed in this AutoLoad: ROOT-10895 suffers from trying to autoload 220 classes, but causing >80k calls to DeepAutoLoadImpl(). Within one call to AutoLoad(), no new information can arrive that would allow later recursions of DeepAutoLoadImpl() to now autoload a class successfully. So whether autoloading a class successfully or not, any class on which DeepAutoLoadImpl() was called does not need to be re-visited within the same AutoLoad() call. To achieve that, remember the class names that have been visited already, bringing the count to (non-early-return) DeepAutoLoadImpl() to the actual number of classes that we should try to autoload. And a perf-opt: do not determine type names if the type is fundamental. And I'm not a big fan of deep code nesting. | 19 March 2021, 10:11:55 UTC |
| 075c1ad | Axel Naumann | 25 November 2020, 17:52:19 UTC | [metacling] In DeepAutoLoadImpl, use the fact that class name is normalized: Considerably speeds up the deeply nested, hot loops through all members. | 19 March 2021, 10:11:55 UTC |
| 8a36c14 | Axel Naumann | 26 November 2020, 10:45:11 UTC | [clingutils] KeepNParams: break on first template default args. This is a hot function, getMinRequiredArguments() is far from cheap, and instead of taking the last redecl with default args we now take the first. There should only be one anyway. | 19 March 2021, 10:11:55 UTC |
| f976af2 | Axel Naumann | 17 March 2021, 15:06:59 UTC | [cling] Skip non-regular files to find dylibs: To determine the file magic, the file needs to be opened and read. This is done with *each* file in $LD_LIBRARY_PATH, including ./ If one of them is e.g. a FIFO then reading blocks until someone writes into the FIFO, which might cause the process to hang. This was reported a couple of times, such as at https://root-forum.cern.ch/t/compiling-from-source-first-interactive-command-hangs/43997/5 Solution: only check for the file magic of *regular* files. Sadly, llvm::sys::fs::get_file_type never sets file_not_found but returns an unspecific status_error. | 19 March 2021, 10:06:58 UTC |
| 273e0fc | Sergey Linev | 18 March 2021, 16:41:25 UTC | [eve7] fix memory leak in REveCalo | 19 March 2021, 07:20:31 UTC |
| f7e431b | Sergey Linev | 18 March 2021, 07:59:27 UTC | [mathcore] Fix memory leak in GausIntegrator #7556 Do noy clone IntegrandTransform object, which leaked in this case | 18 March 2021, 15:29:21 UTC |
| 5a33ece | Enrico Guiraud | 17 March 2021, 16:52:52 UTC | [cmake] Add ROOTVecOps to ROOT_LIBRARIES It is a dependency of ROOTDataFrame which is already included in ROOT_LIBRARIES, so we want it there. | 18 March 2021, 14:07:38 UTC |
| ae14763 | Enrico Guiraud | 18 March 2021, 10:43:28 UTC | [DF][NFC] Backport major doc improvements to 6.24 | 18 March 2021, 11:29:21 UTC |
| 105e087 | Sergey Linev | 17 March 2021, 07:51:58 UTC | Fix TLeaf::GetLeafCounter function #7507 Lead name with dot was not properly tested | 18 March 2021, 07:44:36 UTC |
| 93b046c | Enrico Guiraud | 16 March 2021, 15:17:56 UTC | [DF] Keep SnapshotHelper args alive until jitting Commit 1db2ea4fd6 introduced a bug in lazy jitted Snapshot calls: the constructor arguments of SnapshotHelper[MT] would go out of scope much before jitting time, eventually causing a nullptr dereference in RJittedAction. This fixes #7547. | 17 March 2021, 19:15:47 UTC |
| 4ebe88c | Enrico Guiraud | 16 March 2021, 16:02:21 UTC | [DF] Add test for lazy jitted Snapshot | 17 March 2021, 19:15:47 UTC |
| 1143bea | Javier Lopez-Gomez | 16 March 2021, 15:56:44 UTC | Updated v6.24 release notes for Cling. | 16 March 2021, 19:08:58 UTC |
| 422855e | Emmanouil Michalainas | 15 March 2021, 12:58:23 UTC | [RF] Create README.md for `roofit/batchcompute`. | 16 March 2021, 18:17:15 UTC |
| e6f4c34 | Emmanouil | 16 March 2021, 18:08:51 UTC | [RF] Update release notes (RooBatchCompute + evaluateSpan()). | 16 March 2021, 18:17:15 UTC |
| ec61446 | Emmanouil | 16 March 2021, 18:05:54 UTC | [RF] Fix typo "evaluatedSpan() -> evaluateSpan()". | 16 March 2021, 18:17:15 UTC |
| daa5df9 | Enrico Guiraud | 15 March 2021, 11:57:28 UTC | [relnotes] Mention new behavior of TTreeReader + friend trees | 16 March 2021, 17:45:56 UTC |
| c8a16ad | Enrico Guiraud | 15 March 2021, 11:53:55 UTC | [relnotes][DF] Update release notes | 16 March 2021, 17:45:56 UTC |
| cd3c91f | Sergey Linev | 16 March 2021, 10:22:30 UTC | Update docu to clearly specify that free() function must be used | 16 March 2021, 16:52:39 UTC |
| 50e34a6 | Sergey Linev | 16 March 2021, 10:18:23 UTC | Fix memory leak when using TASImage::GetImageBuffer Allocated memory should be cleanup with `free` function. In all three places where it was used it was wrong. Either `delete` or `delete []` or just do nothing. | 16 March 2021, 16:52:39 UTC |
| 97d0f00 | Sergey Linev | 16 March 2021, 09:53:37 UTC | Fix clang-tidy issues in TASImage Remove unused values, mark _alphaBlend operation as NOLINT | 16 March 2021, 16:52:39 UTC |
| 61f9ea4 | Sergey Linev | 15 March 2021, 10:51:12 UTC | [tree] Fix memory leak in TTreeCacheUnzip #7429 Memory returned by UnzipBuffer must be released by caller | 16 March 2021, 16:49:39 UTC |
| 809b265 | Sergey Linev | 15 March 2021, 10:43:03 UTC | [tree] use strlcpy in TLeaf | 16 March 2021, 16:49:39 UTC |
| edaadb2 | Sergey Linev | 15 March 2021, 10:23:52 UTC | [tree] return boolean value for boolean result | 16 March 2021, 16:49:39 UTC |
| 0dff927 | Sergey Linev | 15 March 2021, 10:21:33 UTC | [tree] Check if sub-lists exists before use it | 16 March 2021, 16:49:39 UTC |
| a387576 | Sergey Linev | 15 March 2021, 10:13:14 UTC | [tree] check negative entry id in TEntryList::GetEntry() #7429 | 16 March 2021, 16:49:39 UTC |
| 00750d7 | Axel Naumann | 15 March 2021, 10:08:25 UTC | When injecting symbols, do not rely on the last Module: With llvm9, the module ptr gets reset upon emission. Instead, just look for the symbol in the JIT - where it should be found, as Init() is injecting it there, and we want to find *that* symbol, not the one from the binary. Fixes `CodeUnloading/AtExit.C` cling test. | 16 March 2021, 15:14:47 UTC |
| 4a4310b | Axel Naumann | 15 March 2021, 11:36:17 UTC | [cling] Fix expected diags in test/Pragmas/load.C: A failure to load the library will *not* trigger loading the same file as a source file: `#pragma load` will cause zero or one diagnostics, not two. Now that the diagnostic for a missing file to be `#pragma load`ed is "file not found", re-use the existing `expected-error@input_line_12` line!. | 16 March 2021, 15:14:29 UTC |
| e317f01 | Axel Naumann | 15 March 2021, 11:33:59 UTC | [cling] Improve diag on #pragma load failure. | 16 March 2021, 15:14:29 UTC |
| 460b14b | Sergey Linev | 16 March 2021, 14:59:59 UTC | [jsroot] fix external modules loading | 16 March 2021, 14:59:59 UTC |
| 6ec7571 | Axel Naumann | 15 March 2021, 13:27:55 UTC | [cling] Bring `vanilla-cling` lit feature back: it is needed for test/Autoloading/AutoForwarding.C. | 16 March 2021, 14:45:39 UTC |
| c1c219a | Axel Naumann | 15 March 2021, 13:37:56 UTC | [cling] Help test find its parts: was failing before but due to `REQUIRES: vanilla-cling` not run. | 16 March 2021, 14:45:39 UTC |
| 4ebe7a6 | Vincenzo Eduardo Padulano | 15 March 2021, 11:38:16 UTC | Add pyspark to requirements.txt, ignore in dependency_versions.py | 16 March 2021, 14:18:15 UTC |
| 2b84398 | Vincenzo Eduardo Padulano | 21 February 2021, 21:51:03 UTC | Relnotes for ROOT.RDF.Experimental.Distributed | 15 March 2021, 14:52:42 UTC |
| b742d3d | Jonas Rembser | 11 March 2021, 20:47:51 UTC | [RF] Fix the integration of a cloned RooParamHistFunc with range After commit a0fa4fa, the integration of a RooParamHistFunc still doesn't work when the RooParamHistFunc was cloned with `cloneTree`. This was because the ranges were stored in a hash map keyed by argument pointer and then not the correct clones of the arguments were used to look up the ranges afterwards. This commit is fixing that. | 15 March 2021, 13:43:36 UTC |
| 2bffa1d | Sergey Linev | 15 March 2021, 11:53:58 UTC | [jsroot] 6.24 branch | 15 March 2021, 12:23:49 UTC |
| b802a6b | Enrico Guiraud | 12 March 2021, 10:03:50 UTC | [DF] Remove RResultPtr::Release (an unreleased feature) The current implementation is broken (fObjPtr.release() deletes the object thet we then return by pointer) and in general it is extremely awkward to convince a std::shared_ptr to release ownership (and for good reason, it only makes sense if you can be absolutely sure that there is only one shared_ptr alive that owns that object). Instead of using RResultPtr::Release, we recommend moving the result out of the RResultPtr. This effectively reverts commit 03fd58eec04ddf96c1596c02d4269185fef42b9f. | 12 March 2021, 15:16:57 UTC |
| 46ef2a3 | Enrico Guiraud | 12 March 2021, 09:53:03 UTC | [DF][NFC] Simplify BoolArray helper class It's enough to make it move-only, we don't need a generic constructor that always throws, we don't need the CopyArray private method. | 12 March 2021, 15:16:40 UTC |
| 41aa416 | Sergey Linev | 11 March 2021, 15:47:34 UTC | Fix memory leak in TGraphEdge DistanceToPrimitive and ExecuteEvent were leaking lot of memory | 12 March 2021, 09:29:11 UTC |
| 408d523 | Sergey Linev | 11 March 2021, 16:23:28 UTC | Wrong delete operator for array in RooSegmentedIntegrator1D | 12 March 2021, 07:39:00 UTC |
| 429c262 | Vassil Vassilev | 08 October 2019, 08:14:16 UTC | [cxxmodules] Mark the overridden transient file. This patch is a complementary chenge to root-project/root@d2c0929e0d It will turn off the isOutOfDate checks for transient files with different size on disk. This is quite dangerous but we are supposed to control the build environment which prepares the distributable binaries. This should fix the cmssw issue: StdDictionaries/src/DataFormatsStdDictionaries/a/DataFormatsStdDictionaries_all_def.xml input_line_8:1:22: error: file '/usr/include/linux/falloc.h' from the precompiled header has been overridden ^ rootcling: /build/cmsbld/jenkins/workspace/build-any-ib/w/BUILD/slc7_amd64_gcc820/lcg/root/6.17.01/root-6.17.01/interpreter/llvm/src/tools/clang/include/clang/Serialization/Module.h:72: clang::serialization::InputFile::InputFile(const clang::FileEntry*, bool, bool): Assertion `!(isOverridden && isOutOfDate) && "an overridden cannot be out-of-date"' failed. The error tells us that `falloc.h` has different file size on the build machine and on the distribution machine. We should probably rely on an environment variable to turn off this diagnostic selectively and more the reponsibility if something goes wrong to the distribution team. They should have better knowledge what is safe to be ignored anyway. | 11 March 2021, 19:46:51 UTC |
| 3826dbd | Vassil Vassilev | 04 October 2019, 12:49:01 UTC | [cxxmodules][pch] Disable validation of transent files. In clang a transient file is a header file which is stored in the pch/pcm file. If this file does not exist, clang creates a virtual file of the unzipped stored file. However, if the file exists it compares the size of the file on disk and the size of the transent file. If the two sizes are different it does not complain (as validation is off) but it triggers an assert in when constructing and outdated clang::InputFile. The implementation disallows a overridden or transient file to be out of date. This patch extends the DisableValidation abilities to span not only for timestamps but we avoid checking the file sizes if the file is on the disk. And, instead we just create a virtual file of the file that we already have. This patch should fix a relocation issue for cmssw and some of the reported LCG issues. | 11 March 2021, 19:46:51 UTC |
| 74e6ec0 | Sergey Linev | 11 March 2021, 16:29:22 UTC | Fix memory leak in TQCommand Wrong delete operator for array | 11 March 2021, 17:00:14 UTC |
| 52296c9 | Sergey Linev | 11 March 2021, 11:36:27 UTC | [oracle] use SYSTEM PRIVATE specifier for OCCI includes Try to awoid warnings comming from them | 11 March 2021, 16:59:07 UTC |
| a4789b5 | Sergey Linev | 11 March 2021, 11:21:24 UTC | [oracle] remove strange CONST defines, avoid using namespace std Mark all non-implemented constructors as deleted | 11 March 2021, 16:59:07 UTC |
| b9b2121 | Sergey Linev | 11 March 2021, 11:12:02 UTC | [oracle] remove OCCI includes from ROOT headers Use forward declarations. | 11 March 2021, 16:59:07 UTC |
| 3c199ed | Bertrand Bellenot | 11 March 2021, 08:40:52 UTC | v6.24 Silent many warning C4624 on Windows Backport from master: Silent the following zillion of warnings on Windows: ``` C:\Users\sftnight\git\master\interpreter\llvm\src\include\llvm/Support/type_traits.h(97,1): warning C4624: 'llvm::detail::trivial_helper<T>': destructor was implicitly defined as deleted [C:\Users\sftnight\build\relwithdebinfo\interpreter\cling\lib\Interpreter\obj.clingInterpreter.vcxproj] with [ T=clang::Sema::ModuleScope ] C:\Users\sftnight\git\master\interpreter\llvm\src\include\llvm/Support/type_traits.h(150): message : see reference to class template instantiation 'llvm::detail::trivial_helper<T>' being compiled [C:\Users\sftnight\build\relwithdebinfo\interpreter\cling\lib\Interpreter\obj.clingInterpreter.vcxproj] with [ T=clang::Sema::ModuleScope ] C:\Users\sftnight\git\master\interpreter\llvm\src\include\llvm/ADT/SmallVector.h(315): message : see reference to class template instantiation 'llvm::is_trivially_copyable<T>' being compiled [C:\Users\sftnight\build\relwithdebinfo\interpreter\cling\lib\Interpreter\obj.clingInterpreter.vcxproj] with [ T=clang::Sema::ModuleScope ] C:\Users\sftnight\git\master\interpreter\llvm\src\include\llvm/ADT/SmallVector.h(837): message : see reference to class template instantiation 'llvm::SmallVectorImpl<T>' being compiled [C:\Users\sftnight\build\relwithdebinfo\interpreter\cling\lib\Interpreter\obj.clingInterpreter.vcxproj] with [ T=clang::Sema::ModuleScope ] C:\Users\sftnight\git\master\interpreter\llvm\src\tools\clang\include\clang/Sema/Sema.h(1724): message : see reference to class template instantiation 'llvm::SmallVector<clang::Sema::ModuleScope,16>' being compiled [C:\Users\sftnight\build\relwithdebinfo\interpreter\cling\lib\Interpreter\obj.clingInterpreter.vcxproj] ``` | 11 March 2021, 13:03:32 UTC |
| a72ea36 | Stephan Hageboeck | 10 March 2021, 18:56:45 UTC | [RF][Docs] Improve doxygen reference of objects in roofit/batchcompute. (cherry picked from commit 549629aa0815cb3ba5d0f1d2f5863597a712458a) | 11 March 2021, 12:01:30 UTC |
| b020e6e | Stephan Hageboeck | 11 March 2021, 08:40:24 UTC | [RelNotes] Minor formatting of RooDataHist part. (cherry picked from commit 51abeafaa52f6bd7b63df18b5dad276392740c00) | 11 March 2021, 11:49:49 UTC |
| 9ac4624 | Stephan Hageboeck | 11 March 2021, 08:36:28 UTC | [RelNotes] Update RooCrystalBall parts of the release notes. (cherry picked from commit c1969ca7f76bd35aeb4e37cd76e22012fa78316e) | 11 March 2021, 11:49:49 UTC |
| e792196 | Enrico Guiraud | 05 March 2021, 18:37:54 UTC | [DF] Snapshot friend branches also when a branch with same name exists Fixes #7181. | 11 March 2021, 08:15:24 UTC |
| 77bf7d5 | Enrico Guiraud | 05 March 2021, 18:37:19 UTC | [TTree][NFC] Add docs to TFriendElement::GetTreeName. | 11 March 2021, 08:15:24 UTC |
| c4f39ee | Enrico Guiraud | 05 March 2021, 18:10:39 UTC | [DF][NFC] Fix whitespaces | 11 March 2021, 08:15:24 UTC |
| 6d179d6 | Enrico Guiraud | 05 March 2021, 17:51:56 UTC | [DF] Add test for #7181: Snapshotting friend branches | 11 March 2021, 08:15:24 UTC |
| 0d49f6f | Stephan Hageboeck | 10 March 2021, 19:16:56 UTC | [RelNotes] Update release notes for RooFit. Co-authored-by: Axel Naumann <Axel.Naumann@cern.ch> | 11 March 2021, 08:06:59 UTC |
| b83106a | Jonas Rembser | 09 March 2021, 12:29:57 UTC | [RF][NFC] Add release notes about new RooCrystalBall class | 11 March 2021, 07:48:01 UTC |
| 82b4624 | Jonas Rembser | 09 March 2021, 12:27:32 UTC | [RF] Remove RooDSCBShape and RooSDSCBShape classes The `RooDSCBShape` and `RooSDSCBShape` classes are succeeded by the more general `RooCrystalBall` PDF. | 11 March 2021, 07:48:01 UTC |
| 27120b4 | Jonas Rembser | 09 March 2021, 11:59:51 UTC | [RF] Adapt RooCrystalBall to be fully compatible with RooCBShape In the RooCBShape, the sign of the alpha parameter decides on which side the tail is. The new RooCrystalBall now follows the the same convention in single-tail mode. | 11 March 2021, 07:48:01 UTC |
| 5ab5b54 | Jonas Rembser | 09 March 2021, 11:30:23 UTC | [RF] Write documentation for RooCrystalBall Write documentation for RooCrystalBall, adapted from previously existing RooDSCBShape documentation. | 11 March 2021, 07:48:01 UTC |
| b7ebc53 | Jonas Rembser | 18 February 2021, 17:18:04 UTC | [RF] Adds RooCrystalBall class that fully parametrizes crystal ball RooFit currently includes the `RooCBShape` and `RooDSCBShape` to represent the family of crystal ball distributions. The `RooCBShape` only has a left tail, and `RooDSCBShape` has two independently parametrized tails. Some users also want a crystal ball where the width of the Gaussian core is different on the left and the right side (see e.g. Jira ticket [ROOT-10388](https://sft.its.cern.ch/jira/browse/ROOT-10388)). To make the RooFit code more maintainable and also cover the use case of an asymmetric Gaussian core, this commit implements a new `RooCrystalBall` class that should cover all classes of crystal ball shapes, such that `RooCBShape` and `RooDSCBShape` are in principle not necessary anymore. The new features of the `RooCrystalBall` class that make it different from `RooDSCBShape` are: - possibility to set sigma of left side and right side of Gaussian core separately - code modernization and using `std::` math functions - three constructors: fully parametrized, symmetric Gaussian core (to replace `RooDSCBShape`), only left tail (to replace `RooCBShape`) - useage of inlined helper functions to avoid code duplication It is also carefully tested in `testRooCrystalBal`: - test the fully parameterized case - test the constructor for left-tail only case and compare values with `RooCBShape` - test the constructor for right-tail only and symmetric tails - test the constructor with symmetric Gaussian core and compare values with `RooDSCBShape` - test integration with many possible combinations integration ranges to cover all branches of integration code | 11 March 2021, 07:48:01 UTC |
| 789c9c2 | Jonas Rembser | 09 March 2021, 19:55:24 UTC | [RF][NFC] Add release notes about updated RooDataHist::sum function | 10 March 2021, 21:49:28 UTC |
| f31dffc | Jonas Rembser | 09 March 2021, 19:20:11 UTC | [RF] Remove unnesseccary includes in RooDataHist and corollary | 10 March 2021, 21:49:28 UTC |
| 54d7cd4 | Jonas Rembser | 09 March 2021, 19:15:33 UTC | [RF] Cache partial bin volumes in RooDataHist with unordered_map The caching of partial bin volumes turned out to be buggy. If it is enabled, the AnalyticalIntegration test in RooDataHist is failing: if one has a two dimensional histrogram in let's say variables (x,y) and if one first integrates over x and then only over y, the second integration mistakenly reuses the first cached bin volumes. This commit repalces the RooCacheManager that was previously used to cache the partial bin volumes with a std::unordered_map. With this new caching structure, the RooDataHist tests succeed. | 10 March 2021, 21:49:28 UTC |
| 8ab0e8c | Jonas Rembser | 09 March 2021, 14:07:22 UTC | [RF] Some code modernization in RooDataHist | 10 March 2021, 21:49:28 UTC |
| c10f6a1 | Jonas Rembser | 25 February 2021, 16:20:05 UTC | [RF] Expand testRooDataHist to extensively test RooDataHist::sum() The RooDataHist::sum method is used in the analytical integration of RooHistPdf and RooHistFunc. This commit adds several tests of RooDataHist::sum via the RooHistPdf and RooHistFunc interfaces. The test matrix covers the following cases: - integration over full range of 1D or 2D observables - integration over full range of one observable while fixing the value of the other - integration over subranges that align with bin edges - integration over subranges that don't align with bin edges - non-uniform binning | 10 March 2021, 21:49:28 UTC |
| d64ca93 | Jonas Rembser | 28 February 2021, 14:53:21 UTC | [RF] Make RooHistFunc directly use integral code in RooHistPdf After a0fa4fa, this commit fixes the remaining analytical integral problems with the RooHistFunc reported in issue #7182 and Jira ticket ROOT-7413. The starting point of this commit is the following observation: usage of RooHistPdf is far more common than RooHistFunc and its analytical integration code supports more special cases. So one can simply take the hopefully bug free analytical integration code from RooHistPdf and substitute it for the existing code in RooHistFunc. However, there is one change that needs to be made when copy-pasting: the `correctForBinSize` in `RooDataHist::sum()` needs to be enabled because integrating the pdf and the function described by the histogram shape is slightly different: - RooHistPdf: the integral of an empirical PDF is simply the integral of histogram counts divided by a normalization factor - RooHistFunc/RooParamHistFunc: here, the bin counts need to be multiplied by the bin volume Code duplication is avoided by moving the integration code into static functions in `RooHistPdf` that `RooHistFunc` can also use. With all these changes, the problems reported in GitHub issue #7182 and Jira ticket ROOT-7413 are fixed. Furthermore, the bin volume normalization capabilities of `RooDataHist::sum()` are also used in `RooParamHistFunc` now. | 10 March 2021, 21:49:28 UTC |
| 9a4a1ee | Jonas Rembser | 28 February 2021, 15:03:44 UTC | [RF] Correct the handling of bin volumes in RooDataHist::sum() This commit updates `RooDataHist::sum()` such that it could also be used for analytic integration in `RooHistFunc` and `RooParamHistFunc`. The logic of RooDataHist::sum() that is already used for RooHistPdf needs to be changed/corrected as explained in the following log sections. First a little overview on `RooDataHist::sum()`: There are three overloads of `RooDataHist::sum()`: 1. Summing over full range of all parameters 2. Summing over full range of subset of parameters 3. Summing in case at least one parameter is not summed over the full range **First change affecting second and third overload of RooDataHist::sum** The 2nd and 3rd overloads state in the documentation: > If correctForBinSize is specified, the sum of weights is multiplied by the M-dimensional bin volume, (M = N(sumSet)). This was not correctly implemented. Before, the 2nd overload multiplied instead with the bin volume of the slice-only set, and the 3rd overload multiplied with the bin volume integrated over all parameters. Note that there is also another parameter of RooDataHist::sum() called inverseBinCor. If it is enabled, the bin counts should be divided by the bin volume of the slice-only set (used for interpretation as PDF). This was already correctly implemented in the second overload, but not in the third one. It has been fixed for the third one. **Second change affecting only the third overload of RooDataHist::sum** This change affects the third overload if the sumSet variables' ranges don't align with bin borders. In this case, the bin counts need to be corrected by `full bin volume` over `bin volume in range`. Before this commit, all histogram variables were considered for this correction factor. Now, only the sum set is considered. **On the inverseBinCor flag and change in RooAbsPdf** So far, the inverse bin volume correction (dividing over slice-only set bin volume) was enabled in RooHistPdf for only the case where the 2nd overload of `RooDataHist::sum()` was called, not the 3rd. This might have caused bugs that were undetected so far and could be one of the reasons why non-uniform binning was often not handled correctly in RooFit. Now that the commit fixed the 3rd overload of `RooDataHist::sum()`, the inverse bin volume correction was enabled there too. | 10 March 2021, 21:49:28 UTC |
| 7cbc515 | Emmanouil | 08 March 2021, 19:12:21 UTC | [RF] Add `Roofit.ComputationLibraryArch` environment option. | 10 March 2021, 09:45:25 UTC |
| db566da | Jonas Hahnfeld | 09 March 2021, 20:22:34 UTC | [cling] Disable GlobalISel on AArch64 (#7419) In tests on an Apple M1 after the upgrade to LLVM 9, this new instruction selection framework emits branches / calls that expect all code to be reachable in +/- 128 MB. This cannot be guaranteed during JIT, which generates code into allocated pages on the heap and could span the entire address space of the process. (cherry picked from commit fcab0add4a79379b5087fe786261f4ab0cc9776a) | 09 March 2021, 20:24:56 UTC |
| 0442aeb | Vincenzo Eduardo Padulano | 09 March 2021, 11:15:07 UTC | Build option for distributed RDF tests that use pyspark The option `dataframe_distpyspark` is superseded by `test_distrdf_pyspark`, to clarify that finding pyspark and its dependencies (like Java) during configuration is strictly needed only for the tests. Generally speaking, pyspark (and Java) are optional runtime dependencies that interest only users of the distributed RDataFrame python module with the Spark backend. | 09 March 2021, 14:37:46 UTC |
| 302d6c6 | Axel Naumann | 05 March 2021, 17:45:22 UTC | [dictgen] Do not create stray temp files: createUniqueFile actually *creates* the file. Changing their name later will not remove the created file. Instead, create the final file directly. (cherry picked from commit bc9576cb6b87d686c370565253f666dab0a79047) | 09 March 2021, 08:32:47 UTC |
| 280986a | Axel Naumann | 05 March 2021, 20:57:03 UTC | [cling] Put `else` back to handle CastExpr of DeclRefExpr | 09 March 2021, 08:32:36 UTC |
| 43aaf50 | Axel Naumann | 05 March 2021, 17:07:04 UTC | [dictgen,clingutils] Expose fwdDecl errors for --debug. | 09 March 2021, 08:32:36 UTC |
| 5e1e0a8 | Axel Naumann | 05 March 2021, 16:51:02 UTC | [cling] ForwardDeclPrinter: also log skipped types. | 09 March 2021, 08:32:36 UTC |
| 21e7cfd | Axel Naumann | 05 March 2021, 16:49:57 UTC | [cling] skipDecl if template arg cannot be fwd declared: Fixes CMS dictionary build issue with forward declaring a template argument of enum constant type - something we do not forward declare. | 09 March 2021, 08:32:36 UTC |
| 4cfc72e | Axel Naumann | 05 March 2021, 16:48:14 UTC | [cling] A CastExpr cannot be a DeclRefExpr. | 09 March 2021, 08:32:36 UTC |
| 68d0df5 | Jonas Rembser | 08 March 2021, 14:14:08 UTC | [RF] Assume uniform weights in computeBatched if weights vector is empty In the case of unweighted datasets, the current implementation of `RooNLLVar::computeBatched` retreived the weight for every event with `RooDataSet::get()`. This has a lot of overhead and significantly slows down the evaluation. With this commit, it is correctly assumed that the weight is the same for every event. This has been done in the code already before but it was changed accidentally in 0e7c394. Hence, this commit partially reverts the changes made in commit 0e7c394. | 09 March 2021, 07:33:38 UTC |
| 5d85efa | Vincenzo Eduardo Padulano | 04 March 2021, 22:28:57 UTC | Add distributed RDataFrame pyspark build option Add new option dataframe_distpyspark. When the option is ON, the distributed dataframe module will also support distributed execution with pyspark. Enabling the option also triggers checks for pyspark dependencies (e.g. Java, py4j) on the system during configuration. | 08 March 2021, 15:20:46 UTC |
| cf75dd0 | Oksana Shadura | 05 March 2021, 11:17:57 UTC | Update xrootd to v4.12.8 (it contains critical patch fixing compilation on Debian/Arch) | 06 March 2021, 10:01:55 UTC |
| 39cff8a | Sergey Linev | 05 March 2021, 09:57:47 UTC | [rootcling] add extra check for MathCore dependency from Vc Enables incremental builds | 05 March 2021, 17:48:49 UTC |
| 7cbc981 | Enrico Guiraud | 04 March 2021, 15:43:07 UTC | [TChain] Do not try to look up index when loading non-existing entries Fixes ROOT-5464. | 05 March 2021, 10:16:45 UTC |
| 53aa5db | Sergey Linev | 03 March 2021, 15:44:23 UTC | Fix libAfterImage building Remove "l" from "ar clq" command while there is no extra library linked to libAfterImage.a | 05 March 2021, 09:16:47 UTC |
| 6be7af1 | Sergey Linev | 04 March 2021, 13:04:27 UTC | [webgui] mark TASImage as supported class in RBrowser Let show images, stored in ROOT file, in TWebCanvas | 05 March 2021, 07:43:24 UTC |
| 0143b0e | Sergey Linev | 04 March 2021, 15:26:38 UTC | [asimage] use SCREEN_GAMMA as parameter Otherwise image slowly degrade if doing loop: for (int i=0;i<100;++i) { img1->GetImageBuffer(&buf, &sz); img1->SetImageBuffer(&buf); } | 05 March 2021, 07:22:22 UTC |
| 27b0948 | Sergey Linev | 04 March 2021, 13:02:41 UTC | [asimage] set screen gamma to 1 With latest libpng versions value 0 recognized as error and produced results are wrong | 05 March 2021, 07:22:22 UTC |
| 1c758b0 | Philippe Canal | 03 March 2021, 23:59:02 UTC | Remove (protected) enum name to fix TTabComTests on Windows | 04 March 2021, 22:38:06 UTC |
| 69b304a | Axel Naumann | 03 March 2021, 11:03:13 UTC | [cling] Extend lifetime of SectionMemoryManager::DefaultMMapper, again: A function-static does not guarantee the lifetime to be sufficiently extended - at least on CentOS7. Use a unique_ptr which guarantees the reference to stay valid for as long as the referencee is alive. Fixes crash at exit on CentOS7. (cherry picked from commit 80c14bb948df53d54596cbbabe725e6ef0acc8e8) | 03 March 2021, 19:21:57 UTC |
| 3b109bc | Axel Naumann | 03 March 2021, 16:55:02 UTC | "Update ROOT version files to v6.23/99." | 03 March 2021, 16:55:02 UTC |
| ce7772b | Axel Naumann | 03 March 2021, 16:31:09 UTC | "Update ROOT version files to v6.23/02." | 03 March 2021, 16:32:09 UTC |
| 392f621 | Axel Naumann | 03 March 2021, 16:31:59 UTC | [build] Adjust RVersion.h location after 3551b7c1f9f0. | 03 March 2021, 16:32:09 UTC |
| 36b06a4 | Philippe Canal | 02 March 2021, 21:34:36 UTC | 'Hide' kOnlyPrepStep as a protected member of TObject. This enum will be removed shortly (functionality moved from TObject::Write to *::Merge) | 03 March 2021, 15:39:17 UTC |
| c9c70ed | Philippe Canal | 01 March 2021, 20:04:30 UTC | [NFC] Improve documentation (TTreeCloner, TBufferMerger, TFileMerger) | 03 March 2021, 15:39:17 UTC |
| f2ac153 | Philippe Canal | 01 March 2021, 20:03:08 UTC | [NFC] Coding style | 03 March 2021, 15:39:17 UTC |
| 30fd4c7 | Philippe Canal | 27 February 2021, 22:04:10 UTC | TFileMerger delete directory only if we induced its creation/loading | 03 March 2021, 15:39:17 UTC |
| 552e963 | Philippe Canal | 27 February 2021, 00:12:20 UTC | TFileMerger: only delete the directory if the merger created it. This is necessary to support the 'fast' incremental merge where the input will continue to be used. | 03 March 2021, 15:39:17 UTC |
