| efe57f3 | Axel Naumann | 13 April 2015, 12:47:49 UTC | Update ROOT version files to v6.02/08. | 13 April 2015, 12:47:49 UTC |
| 7cc2caf | Olivier Couet | 13 April 2015, 09:26:06 UTC | $$$ should be \$\$\$ to be able to generate the PDF. | 13 April 2015, 09:26:06 UTC |
| 76dc390 | Pere Mato | 22 February 2015, 19:17:51 UTC | Fix for ROOT-7070 - ROOT does not compile with latest Pythia 8 version > 8.2 | 13 April 2015, 07:13:49 UTC |
| d6775b6 | Olivier Couet | 02 April 2015, 09:51:57 UTC | When a `TGraph2D` was entirely in the plane `Z=0` the 3D could not be defined. | 02 April 2015, 09:51:57 UTC |
| 9aa6d21 | Pere Mato | 02 April 2015, 07:57:40 UTC | Use another file for python tutorials to avoid clashes while running in parallel. | 02 April 2015, 07:57:40 UTC |
| 0269357 | Axel Naumann | 31 March 2015, 14:40:56 UTC | Prevent unhandled warning about switch cases. (cherry picked from commit ae68602884abd598d1b36d5ad844e9f4741775fa) | 31 March 2015, 15:15:33 UTC |
| ee5e58e | Liza Sakellari | 31 March 2015, 08:38:13 UTC | Fixed parsing of '\'' at the prompt (ROOT-7159). (cherry picked from commit d9e0c08af0e22b3558a54fbe9bc77734009ee83b) | 31 March 2015, 15:15:28 UTC |
| f619bfa | Liza Sakellari | 31 March 2015, 08:30:06 UTC | Fixed escaping strings in xCommand. When calling a macro with arguments like: " \" " or " \\" etc. Macros without arguments are now passing "()"; checking of argument count is now done by Sema. (cherry picked from commit 0fff650d208b12299e6a5582215aa45aef7e7959) | 31 March 2015, 15:15:23 UTC |
| ab68a6c | Liza Sakellari | 31 March 2015, 08:20:05 UTC | Fixed Parser error when calling a macro with arguments. (cherry picked from commit b2fae24ced6293bebad0864d81401a7b1bd14712) | 31 March 2015, 15:15:16 UTC |
| 048a3e7 | Liza Sakellari | 31 March 2015, 08:14:48 UTC | Fixed bug "Parser error when calling a macro with arguments". (cherry picked from commit c4d7d3c1f7f6a53e372276df85d973955258326b) | 31 March 2015, 15:15:11 UTC |
| 9663e86 | Axel Naumann | 25 February 2015, 14:55:29 UTC | Say where the .x function comes from. (cherry picked from commit 314b20bf3f304806ad73f56a1676000804465658) | 31 March 2015, 15:14:24 UTC |
| 149c20d | Axel Naumann | 31 March 2015, 09:03:42 UTC | Support interpreter include paths with whitespace (ROOT-7226). (cherry picked from commit 5c053cf1302d99c927d56fb210330cbb414e8a03) | 31 March 2015, 09:43:31 UTC |
| babb74a | Pere Mato | 29 March 2015, 08:49:17 UTC | The creation and usage of cernstaff.root file does not work when ROOT is installed in a read only volume. Changed to use always the current directory. | 29 March 2015, 08:49:27 UTC |
| 86c54cf | Olivier Couet | 27 March 2015, 15:06:41 UTC | - Implement `TStyle::SetLineScalePS()` to control le global basic line width. | 27 March 2015, 15:06:41 UTC |
| 3559a33 | Olivier Couet | 27 March 2015, 12:09:41 UTC | New method from Jeremi Niedziela <jeremi.niedziela@cern.ch> to return the image in memory. | 27 March 2015, 12:09:41 UTC |
| 5bb205f | Enric Tejedor Saavedra | 26 March 2015, 14:51:31 UTC | Thread safety issue: Protect the removal from the cleanup list with a lock | 26 March 2015, 14:51:31 UTC |
| f7e4f8a | Danilo Piparo | 25 March 2015, 09:57:50 UTC | Adjust genreflex output, also providing new switches Conflicts: core/utils/src/rootcling.cxx | 26 March 2015, 09:15:13 UTC |
| fc92b03 | Axel Naumann | 25 March 2015, 19:18:18 UTC | Update ROOT version files to v6.02/07 to follow new/old versioning. Now, odd patch number are between tags; tags will always have even patch numbers. Thus the next tag will be v6.02/08. This allows users to determine whether they are building against a tag or the ROOT sources after a tag, and it allows us to determine whether a report is from a tagged build or a snapshot. | 25 March 2015, 19:20:53 UTC |
| b9f0c62 | Axel Naumann | 25 March 2015, 10:55:22 UTC | Register $PWD with its full name (ROOT-7114). (cherry picked from commit 9f0daba01af245b531b979eef8001eff6e6db5ae) | 25 March 2015, 19:20:52 UTC |
| 0b51f8d | Pere Mato | 25 March 2015, 18:53:42 UTC | The creation and usage of cernstaff.root file does not work when ROOT is installed in a read only volume. Changed to use always the current directory. | 25 March 2015, 18:57:25 UTC |
| 2836f7b | Olivier Couet | 25 March 2015, 14:04:11 UTC | New option "Rotate scene" in the "Extras" tab of the GL Viewer. It allows to do a real rotation instead of a wobbling when the "Auto Rotator" is launched. (Implemented by Timur, approved by Matevz and used by René). | 25 March 2015, 14:04:11 UTC |
| a79eb8a | Olivier Couet | 25 March 2015, 10:01:35 UTC | Fix ROOT-7206: The axis titles in case of a x:y:z plot with the option COLZ were not correct. | 25 March 2015, 10:01:35 UTC |
| 29f6236 | David Smith | 23 March 2015, 17:05:39 UTC | Fix ROOT-7185 | 23 March 2015, 17:05:39 UTC |
| ba88ac0 | Olivier Couet | 20 March 2015, 16:19:24 UTC | The stroke was a wee bit too thin: it disappeared when the previewer window was reduced. | 20 March 2015, 16:19:24 UTC |
| 27801e8 | Olivier Couet | 20 March 2015, 15:29:10 UTC | make sure the stroke around the solid boxes is thin and not dashed | 20 March 2015, 15:29:10 UTC |
| 7fae2cc | Danilo Piparo | 20 March 2015, 12:59:21 UTC | Lock the Declare interface | 20 March 2015, 13:01:37 UTC |
| 64096cd | Philippe Canal | 20 March 2015, 08:08:08 UTC | Don't try to create an emulated TClass for the global space | 20 March 2015, 08:08:49 UTC |
| 2c9b9ff | David Smith | 19 March 2015, 14:58:24 UTC | ROOT-7185 | 19 March 2015, 14:58:24 UTC |
| 65b8324 | Axel Naumann | 18 March 2015, 16:20:59 UTC | Survive empty sources (thanks, Olivier!). (cherry picked from commit 2d6d9a2467001517f741ccf8400871bf5b6c621f) | 18 March 2015, 16:22:12 UTC |
| e97dfcb | Axel Naumann | 18 March 2015, 16:21:52 UTC | Fix warning (inconsistent override). | 18 March 2015, 16:21:52 UTC |
| 0968b21 | Danilo Piparo | 18 March 2015, 09:56:58 UTC | Properly make SetName thread safe | 18 March 2015, 09:56:58 UTC |
| f905027 | Lorenzo Moneta | 19 November 2014, 16:20:10 UTC | Fixes for ubuntu. Add linking of libgsl and libgslcblas before ROOT libraries (e.g. libMathore) | 18 March 2015, 09:02:13 UTC |
| 465952f | Danilo Piparo | 17 March 2015, 22:46:27 UTC | Revert thread safety in setname there are problems in setting aliases with TTrees. More investigation is needed. | 17 March 2015, 22:46:27 UTC |
| b569218 | Danilo Piparo | 17 March 2015, 21:52:35 UTC | Correct list of special names | 17 March 2015, 21:52:35 UTC |
| 388e5ec | Danilo Piparo | 17 March 2015, 21:06:52 UTC | Protect also the access to the AST necessary to crate the clang::GlobalDecl | 17 March 2015, 21:16:24 UTC |
| b2d41ab | Danilo Piparo | 17 March 2015, 13:08:28 UTC | Fix for ROOT-7188 (ts of creation of TFormula's) | 17 March 2015, 13:12:30 UTC |
| 1167dad | Danilo Piparo | 17 March 2015, 12:50:50 UTC | Build complex dictionary | 17 March 2015, 12:50:50 UTC |
| 771395f | Danilo Piparo | 16 March 2015, 11:16:03 UTC | Remove duplication for routine deciding if name is std::complex type and if yes, which kind of template argument it has. | 16 March 2015, 20:05:02 UTC |
| 460edc0 | Danilo Piparo | 13 March 2015, 19:13:52 UTC | Add support for platform independent std complex This commit allows ROOT6 to perform IO of std::complex<T> classes as in ROOT5. The class provided by the stl implementation of the platform is still used for interactivity, on the other hand, the layout considered for IO (names and sizes of re and im part) is the one of the complex template created for of CINT and not the platform dependent one, which can vary from stl implementation to stl implementation. Summary of the changes: o core/metautils/inc/root_std_complex.h Implementation of the "ROOT complex". A trivial class with two datamembers, simply imported from ROOT5. Some static asserts were added to check the size of this class and the complex one to avoid impatibilities at compile time of (non-ISO) exotic stl implementations. o core/meta/inc/TListOfDataMembers.h / .cxx The datamembers of complex are replaced with the one of the ROOT complex. o core/meta/src/TCling.cxx Inspect members returns the ROOT complex members when asked about complex. This is done to address usages of TMemberInspector's. o core/metautils/CMakeLists.txt / core/metautils/Module.mk All the changes needed to compile the new dictionary. o core/metautils/src/complexLinkdef.h The pragmas needed to select the ROOT complex were added o core/utils/src/rootcling.cxx Add the ROOT complex header to let know the system about it for building the protoclasses in case they involve complexes. Conflicts: core/utils/src/rootcling.cxx | 16 March 2015, 19:56:53 UTC |
| b452a9f | Philippe Canal | 16 March 2015, 16:38:25 UTC | Fix compilation on MacOS (no global thread_local objects) | 16 March 2015, 19:53:10 UTC |
| 01d55ee | Danilo Piparo | 16 March 2015, 16:06:18 UTC | Thread safety related fixes Conflicts: tree/treeplayer/src/TTreePlayer.cxx | 16 March 2015, 19:52:31 UTC |
| 949c56c | Danilo Piparo | 16 March 2015, 16:13:55 UTC | Add missing protection in TH1::SetName | 16 March 2015, 16:13:55 UTC |
| e43be73 | Danilo Piparo | 16 March 2015, 16:00:25 UTC | Adopt better strategy to achieve TS allowing the list of formula to update not only the names but also the hashes related to them. | 16 March 2015, 16:02:38 UTC |
| aeb7b36 | Danilo Piparo | 16 March 2015, 15:59:46 UTC | Remove stray override this keyword must be used consistently inside an interface. | 16 March 2015, 15:59:46 UTC |
| 3b3ff17 | Olivier Couet | 16 March 2015, 14:57:04 UTC | Small update in Preface | 16 March 2015, 14:57:04 UTC |
| c7ae4fa | Olivier Couet | 16 March 2015, 14:45:29 UTC | In case of transparency filled boxes should not be stroked. | 16 March 2015, 14:45:29 UTC |
| 55c50af | Danilo Piparo | 16 March 2015, 14:13:28 UTC | ROOT-7128: TFormula::SetName now is thread safe | 16 March 2015, 14:13:28 UTC |
| 26a6cf4 | Danilo Piparo | 10 March 2015, 11:43:08 UTC | Be able to use multiple TRFIOFile's from multiple threads | 16 March 2015, 14:07:21 UTC |
| b8c383a | Danilo Piparo | 10 March 2015, 10:41:21 UTC | Make documentation about thread safety clear | 16 March 2015, 14:07:21 UTC |
| 05fa9a4 | Danilo Piparo | 10 March 2015, 10:40:44 UTC | Fix a bug (assignment of const char*) and make doc about thread safety clear Conflicts: io/xml/inc/TBufferXML.h io/xml/src/TBufferXML.cxx | 16 March 2015, 14:07:20 UTC |
| 06c8b94 | Philippe Canal | 13 March 2015, 21:03:50 UTC | Protect IncrementalExecutor::m_AtExitFuncs with a spin lock. This prevents: ==4150== Possible data race during write of size 8 at 0xCC136F0 by thread #3 ==4150== Locks held: none ==4150== at 0x146EC214: llvm::SmallVectorTemplateBase<cling::IncrementalExecutor::CXAAtExitElement, false>::grow(unsigned long) (in /afs/cern.ch/cms/sw/ReleaseCandidates/volB/slc6_amd64_gcc491/lcg/root/6.02.00-lnjiaj2/lib/libCling.so) ==4150== by 0x146E9DE3: cling::IncrementalExecutor::AddAtExitFunc(void (*)(void*), void*, cling::Transaction const*) (in /afs/cern.ch/cms/sw/ReleaseCandidates/volB/slc6_amd64_gcc491/lcg/root/6.02.00-lnjiaj2/lib/libCling.so) ==4150== by 0x160A5044: ??? ==4150== by 0x160A67D6: ??? ==4150== by 0x160A6783: ??? ==4150== by 0x145E76D3: TCling::ExecuteWithArgsAndReturn(TMethod*, void*, void const**, int, void*) const (in /afs/cern.ch/cms/sw/ReleaseCandidates/volB/slc6_amd64_gcc491/lcg/root/6.02.00-lnjiaj2/lib/libCling.so) | 14 March 2015, 00:53:26 UTC |
| a15e8a5 | Philippe Canal | 13 March 2015, 21:02:57 UTC | Add cling::internal::SpinLockGuard | 14 March 2015, 00:53:26 UTC |
| 2edda57 | Philippe Canal | 13 March 2015, 19:20:12 UTC | Fix handling base class element. Need to initialize the element (and handle potential base class renaming) before the code needs to use the StreamerInfo representing the base class. | 14 March 2015, 00:53:25 UTC |
| 64722db | Philippe Canal | 13 March 2015, 00:11:49 UTC | Try harder to find the array index/counter in case of schema evolution. This add suppor for renaming a base class which contains the counter or to move the counter within the hierarchy. Make a note that we do not support yet index that are *not* in one of the class or one of its left most base classes. | 14 March 2015, 00:53:25 UTC |
| cb3a73f | Philippe Canal | 13 March 2015, 00:09:41 UTC | Delay the init longer to avoid spurrious error message. With the earlier init, in case of significant schema evolution (including renaming), this was issue error messages including 'missing index for array' | 14 March 2015, 00:53:25 UTC |
| ad4d65a | Philippe Canal | 12 March 2015, 17:10:08 UTC | Add support for custom conversion Streamer in base class streaming | 14 March 2015, 00:53:25 UTC |
| 34608cd | Philippe Canal | 12 March 2015, 17:10:38 UTC | Add support for custom conversion Streamer. Custom conversion streamer are member function with the signature: void Streamer(TBuffer &b, const TClass *onfile_class); which can be told that the object is being read from the data on file that was written under a different class name. The streamer, if present, will be used by the infrastructure instead of the regular custom streamer. | 14 March 2015, 00:53:25 UTC |
| 0c40628 | Philippe Canal | 11 March 2015, 22:23:47 UTC | In BuildOld search when base class is missing, search for a compatible class | 14 March 2015, 00:53:25 UTC |
| 897c4e0 | Lorenzo Moneta | 13 March 2015, 17:00:24 UTC | Add a bit kIsNotW to force an histogram to be not weighted. In that case, even when filling with weights different than one the histogram will consider the weight as the number of entries. Use the bit to fix problems when Fill was used to set the bin contents (for example in the projections to a Profile histogram from a TH2 or a TH3). Remove the setting of defaultsumw2() when testing projections for unweighted histograms. Now stressHistogram tests correctly the above case | 13 March 2015, 17:02:23 UTC |
| fc30d49 | Olivier Couet | 13 March 2015, 12:57:47 UTC | When printing a coloured 2D histograms (with option COLZ) into a PDF or PostScript file, the preview on screen using many standard PDF previewer tools showed very thin white lines between the bins as well as in the color palette. This made very ugly the final output. This problem is due to bad implementation of anti-aliasing in these previewers. A way to bypass this issue was to turn off the anti-aliasing in the previewer bu then the rest of the document does not look nice. This problem is now bypass with this fix in both PDF and PostScript output. | 13 March 2015, 12:57:47 UTC |
| aef539b | Olivier Couet | 12 March 2015, 08:45:00 UTC | Better adjustment of the tilde accent position in case of Cocoa backend. | 12 March 2015, 08:45:00 UTC |
| cd826e2 | Philippe Canal | 11 March 2015, 20:56:15 UTC | Do not overwrite the output file at each iteration | 11 March 2015, 20:57:14 UTC |
| 51420ba | Philippe Canal | 10 March 2015, 18:50:49 UTC | Add nullptr protection | 10 March 2015, 18:50:49 UTC |
| 9691390 | Philippe Canal | 10 March 2015, 17:50:43 UTC | Adapt to v6 | 10 March 2015, 18:04:38 UTC |
| 0605032 | Philippe Canal | 10 March 2015, 17:44:50 UTC | Comment out 'bit' value no longer used | 10 March 2015, 18:04:37 UTC |
| 5123aa9 | Philippe Canal | 10 March 2015, 17:43:44 UTC | C++11 or higher is required | 10 March 2015, 18:04:37 UTC |
| fc3f680 | Philippe Canal | 10 March 2015, 16:22:31 UTC | Fix for C++98 | 10 March 2015, 18:04:37 UTC |
| 507cb71 | Christopher Jones | 10 March 2015, 14:30:57 UTC | Made calling TGenCollectionStreamer::Generate thread safe Although each thread will get its own copy of TGenCollectionStreamer, that copy is made by calling Generate on a global instance of the class. Therefore Generate needs to be thread safe. This translates to requiring TGenCollectionProxy::Initialize be thread safe. This required that TGenCollectionProxy::fValue be atomic and be the last value to be set in TGenCollectionProxy::InitializeEx. Conflicts: io/io/inc/TGenCollectionProxy.h io/io/src/TGenCollectionProxy.cxx Conflicts: io/io/inc/TGenCollectionProxy.h | 10 March 2015, 18:04:34 UTC |
| c78e920 | Danilo Piparo | 10 March 2015, 10:02:35 UTC | Simple ts fixes also aiming not to confuse static analysis. | 10 March 2015, 15:24:53 UTC |
| d72c551 | Philippe Canal | 09 March 2015, 14:28:30 UTC | Add lock making TEnum::GetEnum thread-safe | 09 March 2015, 14:28:30 UTC |
| f26e21d | Philippe Canal | 09 March 2015, 14:24:49 UTC | Properly lock the access to 'raw' list of enums; needed by TEnum::GetEnum | 09 March 2015, 14:24:49 UTC |
| ffb2f6d | Philippe Canal | 09 March 2015, 11:11:13 UTC | Add missing lock in TCling::AutoParse | 09 March 2015, 11:12:52 UTC |
| d08d953 | Danilo Piparo | 09 March 2015, 06:13:00 UTC | Correct fatal typo | 09 March 2015, 06:13:00 UTC |
| a2c8f9b | Danilo Piparo | 09 March 2015, 06:11:15 UTC | Use same string treatment | 09 March 2015, 06:11:15 UTC |
| 6eda37b | Philippe Canal | 08 March 2015, 20:35:23 UTC | Using the override keyword require using ClassDefOverride | 08 March 2015, 20:35:23 UTC |
| a74f27c | Philippe Canal | 08 March 2015, 20:35:04 UTC | Remove unused variable | 08 March 2015, 20:35:04 UTC |
| 0f69f7a | Philippe Canal | 08 March 2015, 19:58:43 UTC | Add TClassEdit::GetUnqualifiedName and use in TEnum::GetEnum and TClingLookupHelper__ExistingTypeCheck | 08 March 2015, 19:58:43 UTC |
| b32ea4e | Philippe Canal | 08 March 2015, 18:07:51 UTC | Use $(OSTHREADLIB) for platform independence | 08 March 2015, 18:07:51 UTC |
| baa143f | Philippe Canal | 08 March 2015, 17:59:50 UTC | In TEnum::GetEnum set lastPos to after the _last_ double colon | 08 March 2015, 17:59:50 UTC |
| fdde4f2 | Philippe Canal | 08 March 2015, 17:31:34 UTC | Don't find the scope in the middle of a template argument ... in TEnum::GetEnum | 08 March 2015, 17:31:34 UTC |
| 5d2f0cf | Philippe Canal | 08 March 2015, 16:58:53 UTC | Allow run-time disable of RooFit Banner via rootrc/gEnv. Set the value of RooFit.Banner either via a rootrc file or a direct call to gEnv to disable (or enable) the printing of the RooFit banner. | 08 March 2015, 17:01:23 UTC |
| 65b25d4 | Philippe Canal | 08 March 2015, 16:41:07 UTC | White space | 08 March 2015, 16:45:02 UTC |
| 8b93440 | Philippe Canal | 08 March 2015, 16:35:20 UTC | Properly create the TListOfEnums in LoadPCM. Since the TListOfEnums for a class is immutable (and thus does not use lock), LoadPCM need to explicitly create it rathen than going through GetListOfEnums(false) which for a class does not really create the list (it is delayed until an actual load and in the meantime a global empty one is returned). | 08 March 2015, 16:45:01 UTC |
| e8de8ae | Philippe Canal | 08 March 2015, 16:41:15 UTC | oups TROOT | 08 March 2015, 16:45:01 UTC |
| da0309b | Philippe Canal | 08 March 2015, 13:08:27 UTC | Add support to 'load' TROOT enum list without exposing TListOfEnums | 08 March 2015, 16:45:01 UTC |
| 430be56 | Philippe Canal | 08 March 2015, 13:03:12 UTC | C++11 is now required | 08 March 2015, 16:45:01 UTC |
| 6884277 | Philippe Canal | 06 March 2015, 22:48:50 UTC | In GetListOfEnums act on load request for namespace even if list already created | 08 March 2015, 16:45:01 UTC |
| bc65316 | Philippe Canal | 06 March 2015, 22:46:50 UTC | Use faster pattern to use/set fProperty | 08 March 2015, 16:45:00 UTC |
| 4401b95 | Philippe Canal | 05 March 2015, 03:52:28 UTC | Coding conventions | 08 March 2015, 16:45:00 UTC |
| 2de30d4 | Christopher Jones | 03 March 2015, 21:05:54 UTC | Fix bug in loading of enums into TClass Now properly handle the case of enums from a namespace | 08 March 2015, 16:45:00 UTC |
| 63af256 | Christopher Jones | 02 March 2015, 22:25:43 UTC | Thread-safe interaction with all enum lists This extends the previous work on enum thread-safety to encompass all the ways the enums can be scoped. | 08 March 2015, 16:45:00 UTC |
| 5eca6a0 | Christopher Jones | 01 March 2015, 17:35:29 UTC | Make list of enums from TClass thread safe The list of enums from TClass are not handled immutably. If the list already is filled then we return that one. If the list is not filled and we are told not to load it then we return an empty list. Else we create a new list and fill it, cache that value and then return it. In this way once the list is handed out to a caller it is not changed. | 08 March 2015, 16:44:59 UTC |
| a662b06 | Christopher Jones | 01 March 2015, 17:33:09 UTC | Convert statics to thread_local for function return values statics were being used to manage memory for buffers which were holding return values from functions. When multiple threads are being used this causes interference. Changing to thread_local avoids the problem. | 08 March 2015, 16:44:59 UTC |
| ffcc720 | Danilo Piparo | 06 March 2015, 07:28:59 UTC | Fix ROOT-7151: link error for rootcling (Makefile infrastructure) | 06 March 2015, 11:01:27 UTC |
| 1dc1e07 | Danilo Piparo | 06 March 2015, 10:36:49 UTC | Adapt the TMVA patch to work on osx the static thread_local class data members are indeed not supported on that platform. | 06 March 2015, 11:01:19 UTC |
| 2b4611c | Danilo Piparo | 06 March 2015, 10:33:28 UTC | Correct name of the mt Reader test into ReaderMT | 06 March 2015, 11:01:03 UTC |
| 7f21098 | Danilo Piparo | 05 March 2015, 13:54:51 UTC | Stress the MT features of the Reader class with a new unit test called utReaderMT. This test runs the code which is in utReader but for 5 times in a row in different threads. Five times in a row because intrinsic non reproducibility is present in multithreaded programs and this makes the test more restrictive as it increases the probability of collisions. | 06 March 2015, 11:00:54 UTC |
| 3115f69 | Christopher Jones | 21 August 2014, 20:41:05 UTC | Make TMVA thread-safe with respect to use of Reader It is now possible to create independent TMVA::Readers and use them simultaneously on different threads. Training of MVAs is still only safe single-threaded. In addition, it is not safe to use multiple instances of MethodCFMlpANN either single or multi-threaded because of a global 'this' pointer. Danilo: Properly take care of the deletion of some atomic pointers Signed-off-by: Danilo Piparo <danilo.piparo@cern.ch> | 06 March 2015, 11:00:36 UTC |
| b00dc08 | Danilo Piparo | 05 March 2015, 08:40:41 UTC | Avoid autoparsing when grabbing non scoped TEnums from the typesystem. | 05 March 2015, 10:10:50 UTC |
| a22b9d1 | Philippe Canal | 05 March 2015, 00:37:46 UTC | Coding conventions | 05 March 2015, 00:37:46 UTC |
