|
From: <oli...@on...> - 2005-01-31 16:12:05
|
When running UnRAR 3.4.3 (http://www.rarlab.com/rar/unrarsrc-3.4.3.tar.gz) in valgrind (latest CVS source), I get a lot of warning like this: ==9188== Thread 3: ==9188== Conditional jump or move depends on uninitialised value(s) ==9188== at 0x1C6660AE: Unpack::DecodeNumber(Decode*) (../../extlib/unrar/unpack.cpp:107) ==9188== by 0x1C664C27: Unpack::Unpack20(bool) (../../extlib/unrar/unpack20.cpp:82) ==9188== by 0x1C665F41: Unpack::DoUnpack(int, bool) (../../extlib/unrar/unpack.cpp:59) ==9188== by 0x1C65A341: CmdExtract::ExtractCurrentFile(CommandData*, Archive&, int, bool&) (../../extlib/unrar/extract.cpp:709) ==9188== by 0x1C657B8E: ProcessFile(void*, int, char*, char*, wchar_t*, wchar_t*) (../../extlib/unrar/dll.cpp:268) ==9188== by 0x1C657CC6: RARProcessFile (../../extlib/unrar/dll.cpp:288) ==9188== by 0x1C611EC6: cal_archive::CRarExtractor::ProcessFile(RARHeaderDataEx const&, int) I looked through the code and debugged into it and I could see where an uninitialised could be. I also messaged the author of the code about this and he didn't see something either. Could this warning be an invalid one? |
|
From: Nicholas N. <nj...@ca...> - 2005-01-31 21:03:45
|
On Mon, 31 Jan 2005 oli...@on... wrote: > ==9188== Conditional jump or move depends on uninitialised value(s) > > I looked through the code and debugged into it and I could see where an > uninitialised could be. I also messaged the author of the code about > this and he didn't see something either. Could this warning be an > invalid one? It's possible; Memcheck can produce false positives. But generally not very many. Without more information, it's impossible to say more. If you're confident your code is ok, you can write a suppression (use --gen-suppressions=yes to auto-generate it). N |
|
From: Julian S. <js...@ac...> - 2005-02-01 00:44:26
|
> > I looked through the code and debugged into it and I could see where an > > uninitialised could be. I also messaged the author of the code about > > this and he didn't see something either. Could this warning be an > > invalid one? > > It's possible; Memcheck can produce false positives. But generally not > very many. Without more information, it's impossible to say more. Does this code use MMX/SSE ? Memcheck is not so good on vector code (although that will improve later this year). J |