3b796f8 | Axel Naumann | 01 September 2021, 08:08:19 UTC | "Update ROOT version files to v6.24/06." | 01 September 2021, 08:08:19 UTC |
b7f2351 | Axel Naumann | 01 September 2021, 08:02:09 UTC | [relnotes] v6.24/06. | 01 September 2021, 08:02:09 UTC |
1d23d25 | Lorenzo Moneta | 27 April 2021, 12:57:12 UTC | When the expected number of events is negative, return a NaN instead of 0. This fixes #8011 | 31 August 2021, 10:16:42 UTC |
c06ad10 | Axel Naumann | 26 August 2021, 16:42:01 UTC | [core] TDirectory::RegisterGDirectory must not update other TDirectory without lock: As we cannot even know whether the other TDirectory has been deleted in the meantime, we have to skip the optimization of erasing globalptr from the list of the oldvalue. This is expected to fix MT crashes observed by Alice. (cherry picked from commit 79a669bb1cd0ccd8bcaeb6c8f6b09c5ced761ac7) | 30 August 2021, 14:53:45 UTC |
cd76db4 | Axel Naumann | 26 August 2021, 09:11:30 UTC | [relnotes] Fix version numbers, style. | 26 August 2021, 09:11:30 UTC |
863e15c | Axel Naumann | 25 August 2021, 13:35:58 UTC | "Update ROOT version files to v6.24/05." | 25 August 2021, 13:35:58 UTC |
6017873 | Axel Naumann | 25 August 2021, 12:35:56 UTC | "Update ROOT version files to v6.24/04." | 25 August 2021, 12:35:56 UTC |
e037c7e | Axel Naumann | 25 August 2021, 12:33:44 UTC | [relnotes] 6.24/04 bugs. | 25 August 2021, 12:33:44 UTC |
2399b68 | Enrico Guiraud | 20 August 2021, 09:57:18 UTC | [tree] Also enable pre-fetching if there are no registered clusters Before this commit, when TTree::fAutoFlush == 0 **and fNClusterRange == 0**, the TTree cache size was set to 0, effectively disabling pre-fetching. With this patch we instead set the cache size to a reasonable default also in this case, assuming the default cluster size of ~30MB. This fixes #8713 also for smaller files where fNClusterRange == 0. | 23 August 2021, 08:58:23 UTC |
6fa3cdf | Enrico Guiraud | 19 August 2021, 08:09:11 UTC | [tree] Enable TTreeCache even if fAutoFlush == 0 With the previous logic, if for some reason fAutoFlush was set to 0 in a TTree, the TTreeCache ended up being disabled. That is undesirable: we still can and want to do pre-fetching even if auto-flushing was turned off when writing the TTree. Other more direct methods to turn off the TTreeCache still work, e.g. tree->SetCacheSize(0). This fixes #8713. | 23 August 2021, 08:58:23 UTC |
29e8bac | Enrico Guiraud | 19 August 2021, 08:08:07 UTC | [tree] Add GetMedianClusterSize We need to estimate this value to provides a reasonable cache size default if other heuristics fail. | 23 August 2021, 08:58:23 UTC |
fb5cd36 | Omar Zapata | 09 August 2021, 14:39:44 UTC | ROOT-R:fixed links to the users guide. | 20 August 2021, 07:13:10 UTC |
262df22 | Omar Zapata | 09 August 2021, 14:38:10 UTC | ROOT-R: fixed segfault evaluating R code | 20 August 2021, 07:13:10 UTC |
1fee0aa | Bertrand Bellenot | 17 August 2021, 09:17:09 UTC | Silent a couple of potential CMake Warnings Silent the following warnings when enabling arrow (`-Darrow=ON`) on Ubuntu: ``` CMake Warning at cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package): By not providing "Findre2.cmake" in CMAKE_MODULE_PATH this project has asked CMake to find a package configuration file provided by "re2", but CMake did not find one. Could not find a package configuration file provided by "re2" with any of the following names: re2Config.cmake re2-config.cmake Add the installation prefix of "re2" to CMAKE_PREFIX_PATH or set "re2_DIR" to a directory containing one of the above files. If "re2" provides a separate development package or SDK, be sure it has been installed. Call Stack (most recent call first): /usr/lib/x86_64-linux-gnu/cmake/arrow/Findre2Alt.cmake:25 (find_package) cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package) /usr/share/cmake-3.16/Modules/CMakeFindDependencyMacro.cmake:47 (find_package) /usr/lib/x86_64-linux-gnu/cmake/arrow/ArrowConfig.cmake:96 (find_dependency) cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package) cmake/modules/FindArrow.cmake:206 (find_package) cmake/modules/FindArrow.cmake:313 (arrow_find_package_cmake_package_configuration) cmake/modules/FindArrow.cmake:350 (arrow_find_package) cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package) cmake/modules/SearchInstalledSoftware.cmake:1088 (find_package) CMakeLists.txt:245 (include) ``` and ``` CMake Warning at cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package): By not providing "Findc-ares.cmake" in CMAKE_MODULE_PATH this project has asked CMake to find a package configuration file provided by "c-ares", but CMake did not find one. Could not find a package configuration file provided by "c-ares" with any of the following names: c-aresConfig.cmake c-ares-config.cmake Add the installation prefix of "c-ares" to CMAKE_PREFIX_PATH or set "c-ares_DIR" to a directory containing one of the above files. If "c-ares" provides a separate development package or SDK, be sure it has been installed. Call Stack (most recent call first): /usr/lib/x86_64-linux-gnu/cmake/arrow/Findc-aresAlt.cmake:25 (find_package) cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package) /usr/share/cmake-3.16/Modules/CMakeFindDependencyMacro.cmake:47 (find_package) /usr/lib/x86_64-linux-gnu/cmake/arrow/ArrowConfig.cmake:96 (find_dependency) cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package) cmake/modules/FindArrow.cmake:206 (find_package) cmake/modules/FindArrow.cmake:313 (arrow_find_package_cmake_package_configuration) cmake/modules/FindArrow.cmake:350 (arrow_find_package) cmake/modules/SearchInstalledSoftware.cmake:20 (_find_package) cmake/modules/SearchInstalledSoftware.cmake:1088 (find_package) CMakeLists.txt:245 (include) ``` | 18 August 2021, 07:11:18 UTC |
4729ca9 | Enrico Guiraud | 09 August 2021, 15:11:38 UTC | [TTreeProcessorMT] Use unambiguous syntax when building chains chain->Add("filename/treename") is ambiguous when filename does not end in `.root`: in that case `TChain` interprets its argument as the full path to the file, with no treename specified. We now instead use the unambiguous syntax "filename?query#treename" when building chains in TTreeProcessorMT. This fixes #8739 (reading files with no `.root` extension in RDF's multi-thread event loops). | 10 August 2021, 09:37:36 UTC |
167fa9c | Axel Naumann | 04 August 2021, 11:53:09 UTC | [netxng] Delete XrdFile already at Close(): This prevents the deletion of the TNetXNGFile triggered by static destruction (TROOT::~TROOT() deleting open files) at a point where the Xrootd client library had its static destruction already run, and thus the call to XrdCl::File::~File() fails. Fixes #8767. (cherry picked from commit 3845be7063f4b45b63be6e7641601e7b28ea3652) | 05 August 2021, 12:13:52 UTC |
7d9f681 | Jonas Rembser | 26 July 2021, 12:33:19 UTC | [RF] Fix size of output span from `RooUniform::evaluateSpan()` The size of the output span from the RooUniform in batch mode should be the same as the size of the input spans, not the product of all sizes of the input batches. This fixes a bug reported on the forum: https://root-forum.cern.ch/t/rooaddpdf-evaluatespan-breaks-if-it-contains-a-roouniform | 27 July 2021, 07:33:45 UTC |
a188e3c | Jonas Rembser | 12 July 2021, 13:31:08 UTC | [RF] Fix `if` statement in HypoTestInverterResult::FindIndex() The `if` statement in `HypoTestInverterResult::FindIndex()` didn't consider the `xvalue == 1` case, resulting in no index being found. | 16 July 2021, 10:14:54 UTC |
9ad55e7 | Sergey Linev | 13 July 2021, 15:22:18 UTC | [jsroot] fix update with zooming | 13 July 2021, 16:17:21 UTC |
b1508df | Sergey Linev | 13 July 2021, 12:33:18 UTC | [webcanvas] avoid TPad::Modified() when attributes not changed Set new grid or ticks only when attributes really changed | 13 July 2021, 16:17:21 UTC |
2801b4b | Philippe Canal | 01 July 2021, 19:25:22 UTC | gdml: add support for the official spelling of millimeter | 07 July 2021, 20:06:07 UTC |
6620549 | Jonas Rembser | 30 June 2021, 16:25:02 UTC | [RF] Store a copy of last-used normalization set in RooAddPdf Every RooFit pdf is evaluated with a set of variables to normalize over (normalization set). A pointer to the last used normalization set is stored in the pdf class. sometimes, pdfs are evaluated without a normalization set in RooFit if the normalization doesn't matter. But for a specific class the normalization always matters: the RooAddPdf Because if the components are unnormalized, you get the wrong shape of the sum. That's why commit f6d1543 added some lines to use the last-used normalization set if you evaluate a RooAddPdf without a normalization set. But since the pdf only has a *pointer* to the last-used normalization set, it will have an invalid pointer if the actual previous normalization set gets destroyed. Because of RooFits memory pool for the RooArgSets that only recycles memory every 6000 RooArgSets, the invalid `_normSet` problem only gets visible in large models, but if the model is large enough it causes reproducible crashes, as reported by the ATLAS Higgs combination group. This commit fixes the issue by copying the last-used normalization set into the RooAddPdf, instead of only storing a pointer. Furthermore, the logic to override the empty normalization set with the last-stored normalization set is moved to `RooAbsPdf::getNormAndCache()`, avoiding the overload of `RooAbsPdf::getValV()`. | 01 July 2021, 16:01:21 UTC |
68effbd | Guilherme Amadio | 18 May 2021, 11:38:59 UTC | Do not force enabling of asserts option If the option is disabled on the command line with -Dasserts=OFF, the build system should respect that and not re-enable it behind the user's back. If not specified on the command line, it will still be set if the build type is Debug or dev=ON. | 30 June 2021, 12:04:27 UTC |
901c0f0 | Guilherme Amadio | 23 April 2021, 07:39:06 UTC | Fix TClingTests for when dependency list contains the binary name When TClingTests appears in the list of dependencies, the output was not matching the expectation, so we just omit the name of the binary itself from the list of dependencies to allow the test to pass in this case. It may be caused by using position independent executables (PIE) when compiling (enabled by default for my GCC). | 30 June 2021, 12:04:27 UTC |
3eb3aab | Guilherme Amadio | 22 April 2021, 14:37:35 UTC | Create relative symlinks when installing PyROOT When ROOT is configured for installation in /usr or other system directory, creating the links directly breaks staged installations since unprivileged users cannot modify /usr. Fixes #7964. | 30 June 2021, 12:04:27 UTC |
26464bf | Guilherme Amadio | 22 April 2021, 14:25:14 UTC | Disable LLVM optional components not used by ROOT This avoids auto-detection of optional LLVM features by searching for the packages in the system, which could lead to problems if our binaries are built on a machine where one of these is installed, and run on a machine where it's not. | 30 June 2021, 12:04:27 UTC |
1a82a07 | Guilherme Amadio | 22 April 2021, 14:24:16 UTC | Fix typo in source filename in net/davix/CMakeLists.txt | 30 June 2021, 12:04:27 UTC |
62e3cf7 | Enrico Guiraud | 17 June 2021, 12:11:56 UTC | Disable tutorials/legacy/thread/threadsh1.C It causes randomic failures in our nightlies, and being in legacy code we are happy with disabling the tutorial altogether. | 30 June 2021, 09:05:06 UTC |
e6567f6 | Jonas Hahnfeld | 30 June 2021, 08:18:24 UTC | Avoid undefined behavior in RANLUX++ Left shifting a negative base is undefined behavior until C++20, cast to unsigned where it becomes implementation defined already in earlier standards and works in practice because hardware uses two's complement. Fixes #8503 (cherry picked from commit b7d0cb13f6544cbefadb26a99840fc41dba32891) | 30 June 2021, 08:18:24 UTC |
50039e6 | Axel Naumann | 28 June 2021, 16:14:33 UTC | [relnotes] Typo. | 28 June 2021, 16:14:33 UTC |
5fc30f5 | Axel Naumann | 28 June 2021, 09:32:03 UTC | "Update ROOT version files to v6.24/03." | 28 June 2021, 09:32:03 UTC |
ad1ddb8 | Axel Naumann | 28 June 2021, 09:17:14 UTC | "Update ROOT version files to v6.24/02." | 28 June 2021, 09:17:14 UTC |
01ced27 | Axel Naumann | 28 June 2021, 08:53:59 UTC | [relnotes] v6.24/02 bug fixes. | 28 June 2021, 08:53:59 UTC |
1bdf3a6 | Sergey Linev | 25 June 2021, 15:44:51 UTC | Fix JSROOT error | 25 June 2021, 18:31:27 UTC |
fe73afc | Sergey Linev | 25 June 2021, 09:46:45 UTC | [jupyter] configure also ServerApp to let use static with jupyter lab Also add test for JSROOTSYS to let use special JSROOT local installation as it THttpServer does | 25 June 2021, 18:31:27 UTC |
3c93825 | Sergey Linev | 25 June 2021, 09:19:43 UTC | [jupyter] do not forget remove failed script element before loading next May harm older JSROOT, should be ok with v6.x | 25 June 2021, 18:31:27 UTC |
01ffa25 | Sergey Linev | 25 June 2021, 10:18:50 UTC | [jupyter] use JSROOT v6.1.1 as fallback in notebooks There is important bugfix for TFrame drawing, which affects interactive features | 25 June 2021, 18:31:27 UTC |
2cf563d | Sergey Linev | 24 June 2021, 10:00:49 UTC | [jupyter] use fully qualified function name to avoid conflicts | 25 June 2021, 18:31:27 UTC |
5f21d1b | Sergey Linev | 24 June 2021, 09:48:29 UTC | [jupyter] configure several locations of JSROOT for require.js If local "static" location fails, it fallback to WWW urls. Helps to correctly work with nbview | 25 June 2021, 18:31:27 UTC |
f98bad8 | Sergey Linev | 10 May 2021, 08:09:08 UTC | Use latest JSROOT as fallback in JupyROOT ROOT is now contains v6.x of JSROOT and its API used also in JupyROOT notebooks. Therefore fallback version should be also from v6.x - latest is 6.1.0 | 25 June 2021, 18:31:27 UTC |
4c774cd | Axel Naumann | 25 June 2021, 13:04:44 UTC | [cmake] Prevent inclusion of stdatomic.h in cxxmodules: It causes a clash of a macro defined in stdatomic with an identifier in boost. Fixes https://github.com/root-project/root/issues/6454. (cherry picked from commit 7a3896a0c6c331f5bdbe6a73fd5101982c3a927b) | 25 June 2021, 16:16:33 UTC |
5a3cc90 | Sergey Suvorov | 21 April 2021, 16:57:32 UTC | add missed $ before env var in fish config (cherry picked from commit 0be8bbee9570776d17196b32c82db1d25c6e3530) | 24 June 2021, 13:04:11 UTC |
9a7ea6f | Axel Naumann | 22 June 2021, 13:04:10 UTC | [cling] Also fwd decl underlying type of using decls: Before, only the using decl itself was forward declared, causing undeclared identifiers in forward declaration code, as witnessed in https://github.com/root-project/root/issues/8499 Given the similarity of using and typedef, merge both into a single function, making sure both have the same featureset, and through that fixing this issue as a side-effect. (cherry picked from commit d026b4918a696e6c8a5ef2c15985c8b9d98da2b2) | 23 June 2021, 18:43:13 UTC |
90aca00 | Jonas Rembser | 04 June 2021, 10:33:36 UTC | [RF] Fix warning about sigma parameter shadowing RooCrystalBall member This fixes warnings in the CentOS 7 build in the nightlies. | 23 June 2021, 16:05:31 UTC |
48fabda | Philippe Canal | 22 June 2021, 21:03:46 UTC | Remove stray debug print out. It was inadvertently introduced in 82fad768c9a In TBranchElement::Print use the correct StreamerInfo in case of schema evolution | 22 June 2021, 21:09:59 UTC |
ce5eb56 | Jonas Rembser | 21 June 2021, 09:32:11 UTC | [RF] Fix missing initializer warnings in MemPoolForRooSets.h (2nd try) This PR should finally fix the build warnings in the master nightlies, after the unsuccessful fix attempt in commit 4ed5ea3976743d1206a8f8c7579adeb1818a9695. I have checked that the missing field initializer warnings go away with this change by compiling the following example snippet under gcc48: ```C++ // compile with g++ -Wmissing-field-initializers -std=c++11 -o test test.cc struct A{ A() : arr_{{}} {} std::array<int, 10> arr_ = {{}}; }; int main() { A a{}; std::cout << a.arr_[0] << std::endl; return 0; } ``` The warnings are reproduced, and can successfully be suppressed by replacing `{}` with `{{}}`. | 22 June 2021, 08:11:09 UTC |
20215c7 | Enrico Guiraud | 03 June 2021, 16:38:32 UTC | [graf2d] Remove code that forced a segfault if gDebug == gVirtualX gcc11 warns about the intentionally broken code: ``` ../io/io/src/TMapFile.cxx:1153:54: note: returned from ‘static void* TObject::operator new(size_t)’ [960/2325] Building CXX object graf2d/x11/CMakeFiles/GX11.dir/src/GX11Gui.cxx.o ../graf2d/x11/src/GX11Gui.cxx: In function ‘Int_t RootX11ErrorHandler(Display*, XErrorEvent*)’: ../graf2d/x11/src/GX11Gui.cxx:179:14: warning: ‘void operator delete(void*, std::size_t)’ called on a pointer to an unallocated object ‘1’ [-Wfree-nonheap-object] 179 | delete kil; | ^~~ ``` | 19 June 2021, 11:05:32 UTC |
15576f6 | Sergey Linev | 02 June 2021, 10:03:10 UTC | Provide initializer for RealInfo_t members in TGNumberEntry.cxx gcc11 complains that some values may be used uninitialized | 19 June 2021, 11:05:32 UTC |
cc95e92 | Enrico Guiraud | 04 May 2021, 10:12:10 UTC | [llvm] Silence cmake warnings about policy CMP0116 This patch is a backport of https://reviews.llvm.org/rG638d84b60b7e7b7fa9099939ab4de2ec1e0c52c3 Fixes #8060. | 19 June 2021, 11:05:32 UTC |
7d1c782 | Sergey Linev | 02 June 2021, 09:02:59 UTC | Fix gcc11 warning in TGeoManager destructor | 19 June 2021, 11:05:32 UTC |
db96ed7 | Enrico Guiraud | 03 June 2021, 16:51:07 UTC | [io] Overload `new` as well as `delete` for TMapFile This fixes the following gcc11 warnings: ``` [433/2325] Building CXX object io/io/CMakeFiles/RIO.dir/src/TMapFile.cxx.o ../io/io/src/TMapFile.cxx: In constructor ‘TMapFile::TMapFile(const char*, const char*, Option_t*, Int_t, TMapFile*&)’: ../io/io/src/TMapFile.cxx:424:45: warning: ‘static void TMapFile::operator delete(void*)’ called on pointer returned from a mismatched allocation function [-Wmismatched-new-delete] 424 | TMapFile *mf = new TMapFile(*mapfil); | ^ ``` | 19 June 2021, 11:05:32 UTC |
3cfbb97 | Enrico Guiraud | 04 June 2021, 10:08:23 UTC | [tree] Sync argument name between header and source This also resolves a gcc11 warning due to `destructor` shadowing a global typedef in a Python header. | 19 June 2021, 11:05:32 UTC |
d9a6d9e | Enric Tejedor Saavedra | 27 May 2021, 07:18:20 UTC | [PyROOT][8128] Fixes for Python3.9 Some of the changes from: https://bitbucket.org/wlav/cpycppyy/commits/78a6e67ff1c9944d31a953bea9744f78884343f1 | 19 June 2021, 11:05:32 UTC |
6f3086b | Enrico Guiraud | 03 June 2021, 17:09:00 UTC | [graf3d] Fix definition of gluTessVertex This fixes the mismatch pointed out by this gcc11 warning: ``` [8203/8650] Building C object graf3d/eve7/CMakeFiles/ROOTEve.dir/glu/tess.c.o ../graf3d/eve7/glu/tess.c:425:46: warning: argument 2 of type ‘GLdouble[3]’ {aka ‘double[3]’} with mismatched bound [-Warray-parameter=] 425 | gluTessVertex( GLUtesselator *tess, GLdouble coords[3], void *data ) | ~~~~~~~~~^~~~~~~~~ In file included from ../graf3d/eve7/glu/tess.h:38, from ../graf3d/eve7/glu/tess.c:40: ../graf3d/eve7/glu/GL_glu.h:372:69: note: previously declared as ‘GLdouble *’ {aka ‘double *’} 372 | GLAPI void GLAPIENTRY gluTessVertex (GLUtesselator* tess, GLdouble *location, GLvoid* data); | ~~~~~~~~~~^~~~~~~~ ``` | 19 June 2021, 11:05:32 UTC |
a1f2ea6 | Jonas Rembser | 16 June 2021, 11:53:19 UTC | [RF] Fix missing initializer warnings in MemPoolForRooSets.h | 18 June 2021, 13:18:51 UTC |
494c2b1 | Jonas Rembser | 16 June 2021, 12:08:49 UTC | [RF] Fix warning: ‘arg’ shadows member of RooTemplateProxy | 17 June 2021, 12:27:16 UTC |
b6b9ca4 | Axel Naumann | 20 April 2021, 16:05:40 UTC | [cling] Prevent memory leak of TClingDataMemberInfo::fClassInfo. (#7915) (cherry picked from commit e6d16d66632a3f77b73b03c4547336f49a0cf2f0) | 16 June 2021, 16:38:16 UTC |
8dc85b0 | Advait Dhingra | 27 April 2021, 08:14:08 UTC | Sqrt issue (#7909) * std::sqrt * Added sin cos and tan * std::floor * Vc sqrt error fixä * Added using std:: before method usage * forgot two * Added std::atan2 * Added std::log | 16 June 2021, 15:55:55 UTC |
375f8e1 | moneta | 26 May 2021, 15:14:53 UTC | Fix bug in using the pdf derivative in TUnuranContDist This fixes the ROOT issue #8196 | 16 June 2021, 15:55:24 UTC |
1014375 | moneta | 06 May 2021, 16:52:40 UTC | Avoid self comparison when sorting RooAbsArg on dependency. This avoids a problem observed only on MacOS | 16 June 2021, 07:08:00 UTC |
f146270 | moneta | 28 April 2021, 14:54:10 UTC | Fix printing of warning message when calling without a normalization set but fixCoefNormalization has been called | 16 June 2021, 07:08:00 UTC |
3ddf638 | Philippe Canal | 08 June 2021, 15:19:58 UTC | Mark the 'unset' state of fDeserializeTypeCache.load as the unlikely case | 11 June 2021, 15:30:03 UTC |
6072cfc | Philippe Canal | 08 June 2021, 18:58:32 UTC | Modernize BulkApiMultiple test. And resolve issue with floating point saturation in the input | 11 June 2021, 15:30:03 UTC |
6124818 | Philippe Canal | 02 June 2021, 22:49:27 UTC | Always accept to return the serialized buffer. The user can always find a way to deserialize it. It might be as simple as byte swapping or might requires to call a full StreamerInfoAction sequence. | 11 June 2021, 15:30:03 UTC |
925475e | Philippe Canal | 02 June 2021, 22:40:34 UTC | Extend BulkApiMultiple test to more data type | 11 June 2021, 15:30:03 UTC |
fc0d4d1 | Philippe Canal | 02 June 2021, 22:40:12 UTC | ReadBasketFast no longer required GetDeserializeType to be called explicitly before hand | 11 June 2021, 15:30:03 UTC |
cb9bc8a | Philippe Canal | 02 June 2021, 22:25:08 UTC | Correct value of TLeaf*::GetDeserializeType and usage thereof. Rename DeserializeType::kDestructive into DeserializeType::kExternal as it describe better the situation (the TBufferFile is not large enough to hold the deserialized data). Correct the value of GetDeserializeType for TLeafB, TLeafG, TLeafO, TLeafS in particular several where incorrectly marked as kExternal when part of an array (fLeafCount != nullptr); This fixes #6520. | 11 June 2021, 15:30:03 UTC |
c155224 | Enrico Guiraud | 09 June 2021, 15:45:05 UTC | [DF] Use different filenames in different test cases Co-authored-by: Axel Naumann <Axel.Naumann@cern.ch> | 10 June 2021, 21:25:37 UTC |
f782a09 | Enrico Guiraud | 09 June 2021, 15:42:00 UTC | [DF][NFC] Simplify SnapshotHelper[MT]::GetDataBlockCallback Co-authored-by: Axel Naumann <Axel.Naumann@cern.ch> | 10 June 2021, 21:25:37 UTC |
b8ce28d | Enrico Guiraud | 04 June 2021, 10:47:51 UTC | [DF] Improve Snapshot branch-address-resetting logic Instead of TTree::AddClone + TTree::CopyAddresses, use the data-block callback mechanism to reset addresses when needed. We cannot trust TTree::CopyAddresses to do the right thing because the output branch might be a different kind of branch than the input branch (e.g. TBranch vs TBranchElement), which leads to the problem described at https://github.com/root-project/root/issues/8295 . Note that TBranch::SetAddress (which we now call directly to keep the address of the output branches synced with the location of the input variables) needs the address of a _pointer that points to the object associated to the branch_ instead of just the address of that object unless the output branch is really a simple TBranch. We reuse fBranchAddresses as storage for these pointers, which until now was only used in case the column was of RVec type. This fixes #8295. As a side-effect of moving away from TTree::AddClone + TTree::CopyAddresses, these changes also fix #7727. | 10 June 2021, 21:25:37 UTC |
bfa9b3e | Enrico Guiraud | 10 June 2021, 07:33:48 UTC | [DF] Add test for data-block callbacks | 10 June 2021, 21:25:37 UTC |
9a49c4a | Enrico Guiraud | 03 June 2021, 11:14:59 UTC | [DF] Add support for data-block callbacks in RDF These are callbacks that are run at the beginning of a new "data block": in the case of a TChain, a new data block starts when the TChain switches TTrees or when a new multi-thread task starts. In other cases a new data block starts with each new task. In the future RDataSources might be able to provide their own definition of a data block. For now, only actions are wired to the data-block callback logic. In the future we could also offer a special kind of Defines that only update their values at the beginning of a data block, which would provide a `DefinePerSample` feature. | 10 June 2021, 21:25:37 UTC |
444c396 | Enrico Guiraud | 01 June 2021, 07:33:09 UTC | [DF] Fix gcc 11 warning ../tree/dataframe/test/dataframe_interface.cxx:451:28: warning: loop variable ‘col’ of type ‘const string&’ {aka ‘const std::__cxx11::basic_string<char>&’} binds to a temporary constructed from type ‘const char* const’ [-Wrange-loop-construct] | 10 June 2021, 19:32:21 UTC |
d8557fc | Axel Naumann | 07 June 2021, 12:23:35 UTC | [llvm] Suppress -Wmisleading-indentation note: interpreter/llvm/src/lib/Target/X86/X86GenDAGISel.inc:266881: note: ‘-Wmisleading-indentation’ is disabled from this point onwards, since column-tracking was disabled due to the size of the code/headers We cannot disable the note, we can only disable the warning itself. (cherry picked from commit c203562fc4328db27aec940f44a7af714e23d4b0) | 10 June 2021, 16:19:57 UTC |
6fd928f | Jonas Rembser | 08 June 2021, 00:21:59 UTC | [RF] Add unit test for RooSimultaneous that covers #8307 | 08 June 2021, 19:07:45 UTC |
93f1421 | Jonas Rembser | 08 June 2021, 00:14:25 UTC | [RF] Handle nullptr or empty norm set ambiguity in RooAbsReal::getVal The downsteam code, -- like RooAddPdf::getValV for example -- assume that a nullptr is passed when no normalization is requested. The case of an empty norm set is not handled correctly in RooAddPdf::getValV, leading to wrong results. However, some calling code passes an empty norm set to RooAbsReal::getVal instead of a `nullptr` in an attempt to disable normalization. This commit suggests to solve this ambiguity at the highest possible level: right at the beginning of RooAbsReal::getVal. If the normalization set is empty, the pointer pointing to it will be set to `nullptr`. This fixes issue #8307. | 08 June 2021, 19:07:45 UTC |
d61c344 | Sergey Linev | 07 June 2021, 10:21:57 UTC | [eve] provide OnZeroRefCount() in derived from TEveRefCnt classes Fixing same kind of error as in REve classes, but do not declare method in base class as abstract | 08 June 2021, 07:00:05 UTC |
574566f | Sergey Linev | 07 June 2021, 10:12:12 UTC | [reve] provide also REveFrameBox::OnZeroRefCount Only final class can make proper "delete this" call | 08 June 2021, 07:00:05 UTC |
4103bc8 | Sergey Linev | 07 June 2021, 09:59:55 UTC | [reve] provide REveRGBAPalette::OnZeroRefCount() implementation When using default REveRefCnt implementation, wrong **this** value used with delete operator because of double inheritance | 08 June 2021, 07:00:05 UTC |
0b87802 | Jonas Hahnfeld | 31 May 2021, 10:11:27 UTC | RANLUX++: Extract only 48 bits per number This restores the connection to the theoretical properties when understanding the original subtract-with-borrow recursion as a dynamical system. (cherry picked from commit ee0ba46eadb889c8a05274e1b7f5ea68c590b2ea) | 08 June 2021, 06:33:03 UTC |
0f013f6 | Jonas Hahnfeld | 06 May 2021, 07:50:09 UTC | RANLUX++: Convert LCG state to RANLUX numbers Just using bits from the LCG state introduces a bias because the number is guaranteed to be smaller than m and therefore not equally distributed over the whole range of 2 ** 576. Instead of storing the LCG state, store the RANLUX numbers and the carry bit and convert when needed according to notes by Lüscher. Refactor the computation of r to a common function used by both mod_m and to_ranlux. (cherry picked from commit 55173bd273fad28897c29bdd261b3868a6158de4) | 08 June 2021, 06:33:03 UTC |
366a16b | Jonas Hahnfeld | 05 May 2021, 10:52:13 UTC | [mathcore] Ensure full modulo operation in mod_m Before mod_m only ensured that the result was smaller than 2 ** 576 to fit into the storage, but it could be greater or equal to the modulus. Update the correction term according to Lüscher. (cherry picked from commit 9cbb4c804811e9f67133fce529f7a94c6e81d36e) | 08 June 2021, 06:33:03 UTC |
5936200 | Jonas Hahnfeld | 07 May 2021, 14:03:21 UTC | [mathcore] Move RANLUX++ implementation to directory Split out addition and subtraction related functions to helpers.h, they will be re-used in future changes. Also add header guards and make the functions in mulmod.h static to avoid redefinition errors should the file be included into multiple TUs. (cherry picked from commit 42a224dee93b37a67323233ca51e162ac52f6bab) | 08 June 2021, 06:33:03 UTC |
9a32d7d | Jonas Hahnfeld | 05 May 2021, 08:51:13 UTC | [mathcore] Update tests for modular arithmetic of RANLUX++ * Add a simple test for mulmod, a driver to call multiply9x9 and mod_m, which are tested separately. * Add explanations to test for minimum and maximum values of r in mod_m. (cherry picked from commit 9084eee7ae8b088c2ca6a689d62e279b003b451d) | 08 June 2021, 06:33:03 UTC |
5515131 | Enrico Guiraud | 18 March 2021, 10:36:24 UTC | [DF] Add RJittedAction to LinkDef file After the latest LLVM upgrade, changes in cling's symbol resolution logic cause some unresolved symbol failures in RDF jitting. Adding RJittedAction to the rootmap should help cling autoloading and work around the problem. | 07 June 2021, 17:44:22 UTC |
1bed887 | Oksana Shadura | 04 June 2021, 14:48:00 UTC | Remove not needed fatal message (fixing ##8280) | 07 June 2021, 14:21:47 UTC |
8a0714c | Enrico Guiraud | 01 June 2021, 17:14:32 UTC | [tree] Print error in case of branch kind mismatch in CopyAddresses TTree::CopyAddresses has the built-in pre-condition that the input and output branches are of the same kind. Clones might be added, however, for which the pre-condition is violated. This is currently the case, for example, with certain usages of RDataFrame::Snapshot, which might create an output branch that is a simple TBranch while the input branch is e.g. a TBranchElement. This results in wrong data being written out. With this patch we detect this case and complain. A proper fix will be proposed soon. The issue is tracked as #8295. | 07 June 2021, 12:30:54 UTC |
30016d8 | Enrico Guiraud | 04 June 2021, 10:37:09 UTC | [DF] Use a RAII object to make sure to always run CleanUpTask Before this patch we skipped running CleanUpTask if the status of the TTreeReader after a single-thread event loop over ROOT data encountered an error. | 07 June 2021, 12:30:31 UTC |
069afcc | Enrico Guiraud | 02 June 2021, 08:46:32 UTC | [DF] Fix compilation of Foreach with mutable lambdas This fixes #8317. | 04 June 2021, 08:56:06 UTC |
d1c9706 | Sergey Linev | 02 June 2021, 09:29:35 UTC | Fix bug in TText::Copy fWcsTitle was copied wrongly | 02 June 2021, 12:31:33 UTC |
8679d51 | Vincenzo Eduardo Padulano | 06 May 2021, 12:17:21 UTC | [core] Missing include in string_view header | 31 May 2021, 16:52:55 UTC |
59ae4a6 | Axel Naumann | 26 May 2021, 09:39:46 UTC | [cling] Trust that GCC has `__attribute__` `visibility`: GCC 4.8 (which we support...) does not have __has_attribute. Fixes CentOS7 builds. | 31 May 2021, 07:23:18 UTC |
c503838 | Axel Naumann | 19 May 2021, 10:07:32 UTC | [cling] Remove symbols explicitly exported through CLING_EXPORT. | 31 May 2021, 07:23:18 UTC |
9f7b787 | Axel Naumann | 19 May 2021, 10:06:46 UTC | [cmake] Simplify visibility setting: Rather than string-replacing visibility, use the CMake interface. Do not set visibility in the ROOT case; ROOT can handle that just fine itself. ROOT actually does *not* need visibility=default, now that the few symbols that are required for using cling (apart from clingtest) are explicitly requested through attributes. | 31 May 2021, 07:23:18 UTC |
6cbf2b3 | Axel Naumann | 21 May 2021, 12:32:01 UTC | [cmake] Forbid semantic-interposition for libCling: To prevent non-ROOT LLVM-libraries from replacing the symbols cling needs (with possible version mismatches), tell GCC to just always resolve calls to dynamic symbols to symbols from libCling itself. | 31 May 2021, 07:23:18 UTC |
0e4390b | Axel Naumann | 21 May 2021, 14:35:49 UTC | [cmake] Set visibility in targets depending on cling: Instead of relying on CLING_CXXFLAGS to collect as needed, set the visbility explicitly, and use target_properties instead of CMAKE_CXXFLAGS. | 31 May 2021, 07:23:18 UTC |
2e28550 | Axel Naumann | 21 May 2021, 14:35:35 UTC | [cmake] Remove warning suppression fixed by llvm9. | 31 May 2021, 07:23:18 UTC |
56cc630 | Axel Naumann | 21 May 2021, 12:30:52 UTC | [cmake] Improve cling/llvm visibility setting from ROOT: For ROOT, all of llvm, clang, and cling want CMAKE_VISIBILITY_INLINES_HIDDEN, so move it here, and use the CMake interface instead of passing a CXX_FLAG. | 31 May 2021, 07:23:18 UTC |
a1e9179 | Axel Naumann | 19 May 2021, 09:55:39 UTC | [cling] Make sure required symbols are exported: and use a standard mechanism that plays well with -fvisibility. Given that several symbols now use declspec(dllexport), remove them from the ad-hoc cling_exports in ROOT's CMakeLists.txt. | 31 May 2021, 07:23:18 UTC |
f1c508c | Aaron Ballman | 15 August 2019, 18:35:44 UTC | Allow standards-based attributes to have leading and trailing underscores. This gives library implementers a way to use standards-based attributes that do not conflict with user-defined macros of the same name. Attributes in C2x require this behavior normatively (C2x 6.7.11p4), but there's no reason to not have the same behavior in C++, especially given that such attributes may be used by a C library consumed by a C++ compilation. llvm-svn: 369033 | 31 May 2021, 07:21:32 UTC |