0585d9f | Axel Naumann | 24 June 2015, 12:24:33 UTC | Update ROOT version files to v6.02/12. | 24 June 2015, 12:24:33 UTC |
c258d97 | Axel Naumann | 23 June 2015, 12:20:21 UTC | Fix 6.02 incompatibility in merging from master. | 23 June 2015, 12:20:26 UTC |
3668c5c | Lorenzo Moneta | 23 June 2015, 10:24:07 UTC | Fix a bug setting option "W" for fitting TGraph2DErrors See https://root.cern.ch/phpBB3/viewtopic.php?f=3&t=19863 | 23 June 2015, 10:32:51 UTC |
00f5d6c | Axel Naumann | 22 June 2015, 13:18:05 UTC | Fix fwd-decl detection and loops. (cherry picked from commit 15191e74acb1d8fcb6d0aa5f87bdc4b1cdee075e) | 23 June 2015, 09:07:25 UTC |
93b663e | Axel Naumann | 22 June 2015, 13:17:48 UTC | The RecursiveASTVisitor now calls Visit of bases itself. (cherry picked from commit 1e0298128d114aa9c38df5144219b2165762c6c1) | 23 June 2015, 09:07:18 UTC |
e8b2f90 | Axel Naumann | 19 June 2015, 14:26:08 UTC | Unload fwd decls of enums; redeclaring definitions fails. (cherry picked from commit 135fcc89ae60c0ab3738e075b87556d3f7a1e671) | 23 June 2015, 09:07:11 UTC |
e8a48d7 | Danilo Piparo | 12 June 2015, 22:25:51 UTC | Correct fatal typo in CMakeLists (cherry picked from commit fb193f22a92c8b518b78c7b4a5408e9d4ec613bf) | 22 June 2015, 18:11:06 UTC |
5753b20 | Axel Naumann | 15 June 2015, 08:28:44 UTC | Fix typo. (cherry picked from commit 6dd0c5743d827ef577bc946f740a3fe01d49dcb0) | 22 June 2015, 13:07:45 UTC |
175b8c7 | Axel Naumann | 15 June 2015, 08:20:45 UTC | Capture glibc headers without include guards for CMake. (cherry picked from commit 7c86fff1de42b2b6157e574d649021a35bf284ca) | 22 June 2015, 13:07:40 UTC |
255f017 | Axel Naumann | 02 June 2015, 19:52:51 UTC | Add include guards for some glibc headers to make PCH work; capture a few more. Headers can only be "read" from the PCH if they have a complete include guard. Several glibc headers do not; the PCH cannot guarantee that in the context where they are included at runtime, the parsing would yield the same result as at the time the PCH was generated (with an include guard, the header would simply be skipped). Thus, clang will parse the header; if that is different from the one parsed at PCH generation time, the same header now has inconsistent size / file locations etc - clang goes belly up. To solve this, we surround the original glibc header with an explicit include guard surrounding an include_next (which will then include the original glibc header). wchar.h cannot be surrounded by include guards as it needs to be included several times, with different preprocessor contexts. Instead capture it and embed it in ROOT's runtime environment. For now, this only supports glibc in /usr/include. Let's wait for the first request... (cherry picked from commit b58d3371a84204f372705b46d0da108f7c1e76f6) | 22 June 2015, 13:07:34 UTC |
d909dd3 | Lorenzo Moneta | 21 June 2015, 16:16:21 UTC | After reverting Tim patch in removing name in RooLinkedList we can set again the name in the detailed output list of RooAbsStudy | 21 June 2015, 16:18:07 UTC |
1926bec | Lorenzo Moneta | 21 June 2015, 15:44:32 UTC | Fix copy ctor of RooAbsAnaConvPdf. This problem was noticed by applying the previous patch of T. Adye in the RoolinkedList::Add to checking pointers added to the list. Restore the _name data member of RooLinkedList. This fixes the detailed output functionality of PROOF | 21 June 2015, 15:45:15 UTC |
5e2f1c9 | Lorenzo Moneta | 20 June 2015, 22:06:07 UTC | Fix for enum type returned by Cling | 20 June 2015, 22:16:20 UTC |
7e918fa | Lorenzo Moneta | 19 June 2015, 15:55:31 UTC | RooHashTablke patch from Tim Adye: Restores the speed of RooLinkedList lookups (and some other operations) to that from ROOT 5.34.17 and before. The hash table was used previously, but this was disabled in ROOT 5.34.18+ because it gave incorrect results when list entries were renamed. This patch allows the hash table to be used without risk of failure, improving the speed of operations that make heavy use of RooFit lists, such as combining and importing (eg. factor 3 on an import). Details from Tim Adye: An update (26/06/2014) included in ROOT 5.34.18 fixed name lookups with RooLinkedList::findArg() when a list element had been renamed, by disabling the hash table optimisation. There was still an issue with find(name) and when removing or replacing renamed elements. This patch should mostly restore (or even slightly improve on) the speed we had before that update (eg. in 5.34.17), but with corrected find, remove, and replace operations. For all this to work, I had to make a few changes to how the name registry is used. RooLinkedList now has _useNptr enabled by default, but only uses the name registry when it is safe and efficient to do so. If a RooAbsArg is ever renamed, a flag is set in the name registry (actually a bit in the new TNamed) to indicate that a hash table lookup may give a false negative for that name. That will permanently (for this job) prevent a quick lookup for that name, but that's OK since renaming is relatively rare. To speed up the name registry (which is now used more often), I increased its static hash table size from 31 to 100000. That might seem like a drastic change, but since this is a singleton it only adds 781 kilobytes to the total memory usage. This patch also includes a few minor fixes to RooHashTable: fix when a replaced element had a different name (slot), fix for find(TObject*) which always failed (but wasn't used), and a small optimisation for RooHashTable::findLinkTo(). | 19 June 2015, 16:14:42 UTC |
206eb9c | Lorenzo Moneta | 19 June 2015, 15:53:06 UTC | Apply patch from Tim Adye: Several improvements to reduce the RooFit memory use, especially for the combined CMS Higgs coupling workspaces (eg. reduce the full CMS workspace requirements from 7.3 to 1.8 GB, with a more modest 5.3 to 4.6 GB improvement for an ATLAS workspace). Details from Wouter Verkerke: - For any RooSimultaneous component in which there is a pdf tagged with attribute "MAIN_MEASUREMENT", only that pdf is forwarded to the component calculator, thereby stripping of all subsidiary measurements (these are evaluated elsewhere anyway) This is already done automatically for ATLAS binned likelihood, where the BinnedLikelihood attribute has a similar effect. Since CMS attaches 2305 subsidiary measurements to each of their O(250) channels. this procedure avoids the replication of ~500,000 RooGaussians and 1.5 million associated RooRealVars. - The default buffer size allocation for TTree attached RooVectorDataStores is reduced from 4096 bytes per branch to 1024. - The initially reserved vector size of Roo(Obj)CacheManager for internally cached objects is reduced from 10 to 2. - The TNamed data member of RooLinkedList is removed (was only used in one very obscure place). - The chunk size for memory pools for RooLinkedListElems is reduced from 1M to 256K. To take advantage of this for the CMS workspace, the pdf terms should be tagged with the "MAIN_MEASUREMENT" attribute. If not already included, it can be added with the following code snippet. RooFIter iter = w->components().fwdIterator() ; RooAbsArg* arg ; while(arg=iter.next()) { if (arg->IsA()==RooAddPdf::Class() && TString(arg->GetName()).BeginsWith("pdf_")) { arg->setAttribute("MAIN_MEASUREMENT") ; cout << "component " << arg->GetName() << " is a cms main measurement" << endl ; } } | 19 June 2015, 16:14:21 UTC |
6a05164 | Lorenzo Moneta | 19 June 2015, 15:40:28 UTC | from Wouter: change components to return a RooArgList instead of a RooArgSet to not remove duplicates (e.g. RooProduct of x*x ) | 19 June 2015, 16:10:47 UTC |
938e5f4 | Lorenzo Moneta | 19 June 2015, 15:36:09 UTC | Apply from 5.34 remaining patch of commit e2bbd24558affe62701f06a4ef4bd7e136c18807 https://root.cern.ch/gitweb?p=root.git;a=commit;h=e2bbd24558affe62701f06a4ef4bd7e136c18807 | 19 June 2015, 16:10:43 UTC |
a4c3c80 | Lorenzo Moneta | 24 April 2015, 09:45:09 UTC | - Add Hash method to RooLinkedList otherwise it cannot be used correctly in THashList classes (e.g. when using it with PROOF) - Fix seeding of random numbers when using it with PROOF (use the worker number in the seeding) to be sure different seeds ar generated for each worker | 19 June 2015, 16:10:40 UTC |
6369517 | Lorenzo Moneta | 04 March 2014, 22:24:22 UTC | Fix a bug introduced by the previous changes in RooSharedPropertiesList | 19 June 2015, 16:10:36 UTC |
6f9bb0e | Lorenzo Moneta | 12 February 2014, 09:47:20 UTC | Fix a compilation warning | 19 June 2015, 16:10:32 UTC |
39b9014 | Wouter Verkerke | 23 December 2014, 13:39:58 UTC | o RooSecondMoment - Introduce first moment as offset in calculation of second moment to promote numeric stability of calculation when first moment is large o RooCmdConfig - Increase length of string arguments allowed o RooCustomizer - When replace operations are executed with missing source or target expressions, ignore the offending operation with a warning message, rather than canceling the entire transaction | 19 June 2015, 16:10:25 UTC |
c4003c7 | Lorenzo Moneta | 31 January 2014, 15:40:15 UTC | Speed up RooWorkspace::import by using a has table for the internal Workspace collection if size is > 1000 and by speeding up the removal of the set which is cloned in the import. We don’t delete the set which calls safeDeleteList. It is not needed, the set contains a list of all the nodes, so we simply delete just one by one. Signed-off-by: Wouter Verkerke <verkerke@nikhef.nl> | 19 June 2015, 16:10:21 UTC |
9a09fe6 | Lorenzo Moneta | 31 January 2014, 10:33:29 UTC | Use hash tables for server list sizes >= 1000 in RooAbsArg::addServer Signed-off-by: Wouter Verkerke <verkerke@nikhef.nl> | 19 June 2015, 16:10:17 UTC |
d9b8d18 | Lorenzo Moneta | 30 January 2014, 16:18:57 UTC | Optimize the filling of share properties (done when cloning RooRealVar) by using a std::map Signed-off-by: Wouter Verkerke <verkerke@nikhef.nl> | 19 June 2015, 16:10:12 UTC |
a07c27d | Philippe Canal | 17 June 2015, 17:36:29 UTC | Update to use newer TDirectory::TContext constructors | 17 June 2015, 19:57:52 UTC |
3cfda63 | Philippe Canal | 17 June 2015, 16:45:30 UTC | Have TDirectory::TContext ctxt(nullptr) set gDirectory to nullptr. Also add TDirectory::TContext::TContext() for the case where the user does not want to change the current directory but want to make sure to undo any change to gDirectory that code protected by the TContext may make: { TDirectory::TContext ctxt; f = TFile::Open(....); } | 17 June 2015, 19:57:51 UTC |
0c19f75 | David Smith | 16 June 2015, 07:09:50 UTC | Fix problem where TNetXNGFile could read from the wrong offset within an archive | 16 June 2015, 07:09:50 UTC |
5105ff2 | Danilo Piparo | 12 June 2015, 11:25:41 UTC | Remove compilaion warning | 12 June 2015, 11:26:01 UTC |
e1e374b | Axel Naumann | 12 June 2015, 08:29:11 UTC | Work around ROOT-6966 to unblock CMS and ATLAS: do not validate arch. The "proper" fix will be to use the architecture from the PCH as the architecture of the interpreter. (cherry picked from commit 94dbfd1337db36490c3ff3a65215b677e7873b05) | 12 June 2015, 09:48:39 UTC |
2981a6b | Danilo Piparo | 03 June 2015, 11:14:27 UTC | Refactor emission of warnings during dictionary generation FIXED: Conflicts: core/metautils/src/BaseSelectionRule.cxx Conflicts: core/utils/src/rootcling.cxx | 11 June 2015, 16:48:42 UTC |
e853398 | Danilo Piparo | 02 June 2015, 20:01:14 UTC | Uniform warning style | 11 June 2015, 03:59:46 UTC |
6195088 | Danilo Piparo | 25 February 2015, 17:19:06 UTC | Skip for the moment the templated functions | 11 June 2015, 03:59:22 UTC |
25b6aac | Danilo Piparo | 24 February 2015, 12:39:54 UTC | From V.Innocente: Allow to construct SMatrices w/o initialisation of the matrix elements. This is a performance optimisation. | 11 June 2015, 03:59:10 UTC |
3acea1e | Danilo Piparo | 24 February 2015, 12:36:34 UTC | From V.Innocente: offsets are calculated at compiletime thanks to constexpr functions. | 11 June 2015, 03:59:02 UTC |
259d351 | Pere Mato | 10 June 2015, 11:46:14 UTC | Fix for ROOT-7409 - rootcling does not run on MacOSX | 10 June 2015, 13:37:58 UTC |
a0508a6 | Axel Naumann | 09 June 2015, 13:20:13 UTC | From Patrick Gartung: fix syntax error. (cherry picked from commit 59fa0799d062172f4428964ba2c95c8d97df0858) | 09 June 2015, 15:14:04 UTC |
739eecc | Axel Naumann | 09 June 2015, 06:43:33 UTC | Fix spelling of test &&; re-use generic LLVM_CFLAGS. (cherry picked from commit 550975168ef9bbcb0be61d3f33b363f741210a65) | 09 June 2015, 15:13:59 UTC |
e80ecb4 | Patrick Gartung | 08 June 2015, 19:13:58 UTC | add -fno-omit-frame-pointer for gcc on osx (cherry picked from commit bc51e791def28a6a414cc3db002c8f0a58d40090) | 09 June 2015, 15:13:28 UTC |
495e513 | Lorenzo Moneta | 08 June 2015, 09:24:33 UTC | Fix contour computations in RooMinimizer (fix ROOT-7290) | 08 June 2015, 10:00:05 UTC |
029813b | Pere Mato | 05 June 2015, 14:52:15 UTC | Replace PATHS by HINTS to avoid getting the system installed ROOT (e.g. lxplus) | 05 June 2015, 15:01:20 UTC |
de03e19 | Pere Mato | 04 June 2015, 19:37:34 UTC | Fix for ROOT-7401 - Define ROOT_XXX_FLAGS in ROOTConfig.cmake for use by client packages | 05 June 2015, 15:01:10 UTC |
7a528b5 | Lorenzo Moneta | 04 June 2015, 14:56:26 UTC | Fix calling 2 times BufferEmpty in case of histograms filled with weights different than one. This is similar to ROOT-7328 | 04 June 2015, 16:01:40 UTC |
c9d09ff | Gerardo Ganis | 07 February 2015, 11:07:15 UTC | XRootD-support: fix for out of source builds | 03 June 2015, 17:11:07 UTC |
c09a4a1 | Gerardo Ganis | 03 June 2015, 16:55:34 UTC | Proof: fix issue with trees in subdirectories This patch make sure that the treename is passed to TDirectoryFile::GetKey . For some reason in ROOT 5 the search was more 'tolerant', but changes/optimizations occured in this part of the code in the last two years require this patch. This fixes the residual problem with stressProof in ROOT 6 . | 03 June 2015, 16:56:15 UTC |
eafc067 | Gerardo Ganis | 03 June 2015, 14:25:42 UTC | Proof: add support for aliases in Draw actions | 03 June 2015, 14:53:54 UTC |
fcfbc1e | Lorenzo Moneta | 01 June 2015, 10:45:34 UTC | Fix bug in extension of axis for TProfile2D and 3D. (ROOT-7389) Add tests for axis extension | 01 June 2015, 11:52:41 UTC |
c791802 | Gerardo Ganis | 06 February 2015, 17:02:00 UTC | XProofd: remove launcher class Not used and already disabled, but bringing in a unwanted dependency on a private xrootd header | 29 May 2015, 22:26:53 UTC |
f9e97f6 | Pere Mato | 10 December 2014, 17:59:52 UTC | Implementation of XROOTD cross-protocol redirects in ROOT (Lukasz Janyst). Copied stressIOPlugins test from master | 29 May 2015, 22:20:46 UTC |
23066bd | Gerardo Ganis | 06 February 2015, 19:12:29 UTC | XRootD-Support: workaround for the problem with 'private' headers During a discussion with the XRootD main developers we agreed on the steps to gradually remove the dependency from headers distributed in the 'private' RPM - not installed, for example, on lxplus. In the meantime, the workaround consists in importing under 'proof/xrdinc' the headers incriminated, so that builds can be done also when include/xrootd/private is not installed. The relevant legacy libraies are always installed, for the time being; the time scale for their removal (probably end of this year) determines the time scale for the removal of private header dependency. This patch creates 'proof/xrdinc' with the relavant legacy headers and makes the relevant changes in the configuration and (c)make files to deal with it. | 29 May 2015, 22:13:38 UTC |
0078591 | Gerardo Ganis | 29 May 2015, 20:44:19 UTC | Proof: remove from cache files created during load This patch explicitely removes the files '<name>_*' created during macro loading. This prevents a nasty problem on second load badly affecting PROOF for versions >= 6 . | 29 May 2015, 21:06:35 UTC |
3916114 | Gerardo Ganis | 21 May 2015, 15:39:10 UTC | Proof: no longer use temporary files for SETUP (2nd) | 29 May 2015, 21:06:26 UTC |
f35ab52 | Gerardo Ganis | 18 May 2015, 08:34:13 UTC | stressProof: include dialog headers | 29 May 2015, 21:06:17 UTC |
cb8e8d6 | Philippe Canal | 20 May 2015, 17:44:23 UTC | In Proof no longer use temporary file for SETUP routines. Use TMacro::Load instead which pass the string directoy to cling | 29 May 2015, 21:06:07 UTC |
eee95c8 | Philippe Canal | 20 May 2015, 17:43:28 UTC | TMacro::Exec no longer use a temporary file. Switch TMacro::Exec to use TMacro::Load and thus to no longer go via a temporary file | 29 May 2015, 21:05:57 UTC |
b73c90d | Philippe Canal | 20 May 2015, 17:42:37 UTC | Add TMacro::Load. This loads the content of the Macro in the interpreter without executing it. It does not use a temporary file. | 29 May 2015, 21:05:48 UTC |
4172a1e | Pere Mato | 29 May 2015, 12:03:06 UTC | From master: change the name of the produced tarfile by CPack following current conventions | 29 May 2015, 12:03:06 UTC |
876018f | Axel Naumann | 29 May 2015, 09:11:40 UTC | Update ROOT version files to v6.02/11. | 29 May 2015, 09:11:40 UTC |
4e5f909 | Axel Naumann | 29 May 2015, 08:57:08 UTC | Start next patch dev cycle. | 29 May 2015, 08:57:08 UTC |
54c9343 | Axel Naumann | 29 May 2015, 08:33:42 UTC | Update ROOT version files to v6.02/10. | 29 May 2015, 08:33:42 UTC |
b68c097 | Philippe Canal | 28 May 2015, 22:35:31 UTC | Protect TTree ctor against null gDirectory | 28 May 2015, 22:38:50 UTC |
15543bd | Philippe Canal | 28 May 2015, 02:10:33 UTC | Add THashTable::AddBefore to fix the ordering of the keys in a TDirectoryFile Using THashList::AddBefore you could order the object in the list but not within the buckets of the hash table resulting in a different ordering depending how the objects were searched. Recently TDirectoryFile was enhanced by replacing a search through the whole list by a search through the bucket resulting (without this patch) in a change of the ordering. | 28 May 2015, 22:38:39 UTC |
3f66c3b | Philippe Canal | 22 May 2015, 20:21:18 UTC | Extend hadd -f to support compression setting. Previsously this was limited to zlib compression level. Now, we can also specify the compression algorithm. See TFile::SetCompressionSettings for the range of value supported. The patch was provided by Christopher Rob Jones in ROOT-7327. | 28 May 2015, 22:38:39 UTC |
e7fc182 | Philippe Canal | 22 May 2015, 16:26:19 UTC | Properly reset the cached 'current basket' info when reusing/reset that basket | 28 May 2015, 22:38:39 UTC |
4d5eb15 | Lorenzo Moneta | 28 May 2015, 13:19:13 UTC | Add a tolerance comparison check to fix a failure seen in 32 bits arch. | 28 May 2015, 15:29:36 UTC |
cf12bba | Axel Naumann | 27 May 2015, 15:35:24 UTC | Survive empty arrays (ROOT-7362). Thanks, Maximilian Knut Kiesel! (cherry picked from commit 28ecf496f3d2b6fe6d46f4e2cb8c2e1ffcb275ab) | 28 May 2015, 07:42:30 UTC |
6e77294 | Olivier Couet | 27 May 2015, 13:41:04 UTC | Fix https://sft.its.cern.ch/jira/browse/ROOT-6703. | 27 May 2015, 13:41:04 UTC |
b1ed1ec | Lorenzo Moneta | 27 May 2015, 09:40:56 UTC | Improve TH1::Add by trying to use TH1::Merge when the two histograms are not consistent and the coefficients are equal to 1 | 27 May 2015, 09:42:28 UTC |
100de6a | David Abdurachmanov | 26 May 2015, 08:41:00 UTC | TClass,TCling: resolve use-after-free bugs Found by ASan + GCC 4.9.1 while compiling CMSSW. In both cases c_str() or data() is called on a temporary string. Temporary is destructer/freed after statement thus afterwards pointers are not valid. Signed-off-by: David Abdurachmanov <David.Abdurachmanov@cern.ch> | 26 May 2015, 09:20:12 UTC |
28d5d9e | David Abdurachmanov | 26 May 2015, 08:59:11 UTC | TClass: resolve use-after-free bug in TClass::GetClass Found by ASan + GCC 4.9.1. Data() is called on a temporary string. Signed-off-by: David Abdurachmanov <davidlt@cern.ch> | 26 May 2015, 09:17:46 UTC |
292f220 | Philippe Canal | 21 May 2015, 19:34:06 UTC | Finish commit 8b9875f360 for c++14 (cmake flavor) Conflicts: cmake/modules/RootConfiguration.cmake | 21 May 2015, 19:37:47 UTC |
b378842 | Philippe Canal | 21 May 2015, 19:31:27 UTC | Finish commit 8b9875f360 for c++14 | 21 May 2015, 19:35:32 UTC |
8665a4a | Philippe Canal | 21 May 2015, 19:30:48 UTC | White space | 21 May 2015, 19:35:26 UTC |
42837dd | Pere Mato | 03 March 2015, 10:53:16 UTC | Implemented configuration option cxx14 to enable compiling with this c++14 version of the standard. Conflicts: cmake/modules/RootConfiguration.cmake | 21 May 2015, 18:55:25 UTC |
2b4f9de | Philippe Canal | 30 October 2014, 18:12:01 UTC | Add --enable-cxx14 and --cflags to the configure script. For now --enable-cxx14 uses -std=c++1y --cflags should be used when custom configuring ROOT with compiler flags that affects binary compatibility (like -std=c++1y for example) This prevents this kind of important customization, historically done in MyRules.mk, from being 'forgotten' from root-config. Conflicts: config/Makefile.ios config/Makefile.ios64 config/Makefile.iossim config/Makefile.linux config/Makefile.linuxarm64 config/Makefile.linuxx8664gcc config/Makefile.linuxx8664icc config/Makefile.macosx config/Makefile.macosx64 config/root-config.in configure | 21 May 2015, 18:55:25 UTC |
bbed670 | Olivier Couet | 21 May 2015, 08:47:07 UTC | Undo the "fix" forcing pdf option when "Title:" was present. It destroyed the table of content for pdf. | 21 May 2015, 08:47:07 UTC |
2ffd586 | Olivier Couet | 20 May 2015, 14:22:23 UTC | fix typo reported here https://root.cern.ch/phpBB3/viewtopic.php?f=10&t=19724 | 20 May 2015, 14:23:22 UTC |
ad66ada | Philippe Canal | 19 May 2015, 14:35:29 UTC | Avoid (unncessary) race condition in TMVA::MethodCategory::InitCircularTree | 19 May 2015, 14:35:38 UTC |
6921edd | David Smith | 13 May 2015, 17:25:47 UTC | Backport of fix for ROOT-7015 This is a backport of the fix on master for ROOT-7015, commit 5fc68e7a161370008aa97cac9c5cdda6ad326ea5 to avoid the reported segmentation violation. This port leaves out some aspects of the master change: * the previous non-automatic cache creation behaviour of SetCacheSize is kept, i.e. first call or different cache size resets the cache, including any user set branches and learning status. * calls to the TTree cache control methods, like TTree:AddBranchToCache will not implicitly create a cache if the autocache feature is not enabled. * return type kept as void for various cache control methods * some new warning or error messages are not added | 13 May 2015, 17:25:47 UTC |
3f52ab4 | Axel Naumann | 12 May 2015, 13:29:40 UTC | Only emit Value return placement new for CallInit (ROOT-7310). (cherry picked from commit 3816ee6f9fba93704e0e3ee6cdf24952c6f37f02) | 13 May 2015, 10:50:33 UTC |
5ad1b57 | David Smith | 12 May 2015, 15:36:58 UTC | Move noisy message in xrootd plugin during GetPathInfo. ROOT-7185 | 12 May 2015, 15:36:58 UTC |
d604a8d | Lorenzo Moneta | 11 May 2015, 09:31:45 UTC | Fix calling BufferEmpty in several TH1 functions. This is needed since we are using now in ROOT 6 RetrieveBinContent instead of GetBinContent that is calling automatically BufferEmpty. Add a test in stresshistogram | 11 May 2015, 13:47:50 UTC |
59fe9c6 | Philippe Canal | 08 May 2015, 18:27:27 UTC | Support TTree with more than 32766 branch in TTreeFormula. The limit is now the same as the location (index need to fit in a signed integer). This is since in a real use case where the TTree has a total of 34879 (sub)branches. | 08 May 2015, 18:34:37 UTC |
4185c36 | Philippe Canal | 08 May 2015, 16:54:22 UTC | Properly handle const seen before and after a std::basic_string | 08 May 2015, 17:38:29 UTC |
96ad847 | Philippe Canal | 08 May 2015, 16:53:42 UTC | In GetPartiallyDesugaredTypeImpl fix type replacement. When looking up in the list of type to be replaced by a typedef (basic_string -> string), use the canonical type to make sure we have the right search (any sugar will make the search fail). | 08 May 2015, 17:38:23 UTC |
6fc9c4f | Philippe Canal | 08 May 2015, 15:16:59 UTC | More extensions of const keyword support in name normalization. Support the cases: vector<const cl::inner<const long>*const> vector<const cl_t::inner<const long>*const> but break pair<const basic_string<char,char_traits<char>,allocator<char> >,list<int> > | 08 May 2015, 17:38:16 UTC |
c40ee59 | Lorenzo Moneta | 08 May 2015, 12:57:33 UTC | Fix check for future GSL versions (fix for ROOT-7325) | 08 May 2015, 13:00:16 UTC |
bb49d53 | Philippe Canal | 08 May 2015, 04:45:09 UTC | Handle properly also mix of const,* and & as type suffix. This completes/replaces/finishes commit 87b16b5. | 08 May 2015, 12:04:24 UTC |
8230408 | David Smith | 08 May 2015, 09:46:02 UTC | ROOT-7296 | 08 May 2015, 09:46:02 UTC |
3f277e9 | Philippe Canal | 07 May 2015, 20:09:26 UTC | Don't force the StreamerInfo of transient member. This is seen by CMS in the case where a transient member is an std::atomic and is the target of an I/O rule. | 07 May 2015, 20:09:26 UTC |
4cff540 | Philippe Canal | 07 May 2015, 20:08:42 UTC | Typo | 07 May 2015, 20:08:42 UTC |
3c57be9 | Philippe Canal | 07 May 2015, 20:08:24 UTC | Add TStreamerElement::IsTransient | 07 May 2015, 20:08:24 UTC |
4d98a51 | Philippe Canal | 07 May 2015, 15:36:03 UTC | Properly support normalization when there a trailing const. This was the cause of ROOT-7321. | 07 May 2015, 15:50:04 UTC |
a1e1256 | Philippe Canal | 07 May 2015, 15:33:58 UTC | Better default value for fClassProperty. This avoids emulated classes from being marked with their properties 'on'. | 07 May 2015, 15:50:04 UTC |
c32fe93 | Lorenzo Moneta | 07 May 2015, 14:17:34 UTC | Fix check on GSL version | 07 May 2015, 14:18:57 UTC |
a7e42b3 | Lorenzo Moneta | 07 May 2015, 13:50:04 UTC | Fix the model to be fitted in the test. A variable which is not part of the model would be set constant to avoid using in the fit with a derivative=0. | 07 May 2015, 14:18:48 UTC |
c5c3a15 | Philippe Canal | 06 May 2015, 19:27:17 UTC | Alleviate the symptoms of ROOT-7181. The underlying problem is still there and shown below. The cause is that when GetNormalizedName is run with autoloading disabled, it can lead to the premature instantiation of partially specialized templates. .x /local2/pcanal/root_working/test/2015-04-attila/root6DVDictLoading/RootCore/scripts/load_packages.C TFile *_file0 = TFile::Open("../AOD.01597980._000098.pool.root.1") std::string normalizedName; int oldAutoloadVal = gInterpreter->SetClassAutoloading(false); TClassEdit::GetNormalizedName(normalizedName, "SG::AuxTypeVectorFactory<ElementLink<DataVector<xAOD::Photon_v1> > >") gInterpreter->SetClassAutoloading(oldAutoloadVal) TClass::GetClass("ElementLink<DataVector<xAOD::Photon_v1> >")->GetClassInfo() ... In file included from input_line_143:60: In file included from .../RootCore/include/xAODEgamma/PhotonContainer.h:8: /local2/pcanal/root_working/test/2015-04-attila/root6DVDictLoading/RootCore/include/xAODEgamma/versions/PhotonContainer_v1.h:14:1: error: explicit specialization of 'DataVectorBase<xAOD::Photon_v1>' after instantiation DATAVECTOR_BASE( xAOD::Photon_v1, xAOD::Egamma_v1 ); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .../RootCore/include/AthContainers/DataVector.h:612:43: note: expanded from macro 'DATAVECTOR_BASE' #define DATAVECTOR_BASE(T, BASE) \ ^ .../RootCore/include/AthContainers/DataVector.h:622:20: note: expanded from macro '\ DATAVECTOR_BASE_FWD' template <> struct DataVectorBase<T> \ ^~~~~~~~~~~~~~~~~ .../RootCore/include/AthContainers/DataVector.h:718:42: note: implicit instantiation first required here template <class T, class BASE = typename DataVectorBase<T>::Base> ^ Error in <TInterpreter::AutoParse>: Error parsing payload code for class DataVector<xAOD::Photon_v1> with content: | 06 May 2015, 19:28:23 UTC |
09b7416 | Wim Lavrijsen | 05 May 2015, 17:52:28 UTC | fix for ROOT-6634: binary+ for Math::LorentzVector | 05 May 2015, 17:52:28 UTC |
fa6ef63 | Enric Tejedor Saavedra | 05 May 2015, 15:07:17 UTC | Protect removal from global list of functions | 05 May 2015, 15:07:41 UTC |
eb21ce1 | Brian Bockelman | 01 May 2015, 02:54:47 UTC | Quiet unnecessary warning from plugin manager. | 01 May 2015, 16:47:50 UTC |
64a7c79 | Lorenzo Moneta | 29 April 2015, 14:29:59 UTC | Merge form master fix for calling right method of TH2 in RuleFit | 29 April 2015, 15:49:52 UTC |