Hi David, Sorry I didn't answer you earlier, I was busy concluding my term. Finally finished yesterday, I now have 100% of my time for GCC.
I build yesterday the analyzer with -fanalyzer enabled. Globally there are not many coherent warnings, and a common issue are the thesis-long warnings. Below are the headers of these warnings, the first two were emitted dozen of times across the build. I'll look into these two particularly as it would clean up a lot of noise easily enough. ../../gcc/gcc/wide-int.h:1338:30: warning: use of uninitialized value ‘‘result_decl’ not supported by dump_expr<expression error>’ [CWE-457] [-Wanalyzer-use-of-uninitialized-value] ../../gcc/gcc/analyzer/bounds-checking.cc:658:44: warning: use of uninitialized value ‘<unknown>’ [CWE-457] [-Wanalyzer-use-of-uninitialized-value] ../../gcc/gcc/make-unique.h:41:30: warning: use of possibly-NULL ‘operator new(120)’ where non-null expected [CWE-690] [-Wanalyzer-possible-null-argument] I'm also adding new tests for c++. Specifically I'm writing tests for the operators [placement] new and delete, as a number of false-positives occurred repeatedly when building the analyzer. I'm also reediting some of the c tests but with their c++ counterparts, such as using the standard libraries containers and see how the analyzer behaves. Should I put them under c-c++-common or c++ ? With a new analyzer folder and analyzer.exp I guess. For the warnings diving too deep into the standard library, I considered adding a flag to the analyzer to control the maximum depth of the warnings. Have a nice day, Benjamin.
