On Fri, Nov 28, 2014 at 12:26 PM, wuning shi <wuning....@gmail.com> wrote: > > hello > > 1. can you tell me where is the memory scan in source code?(which file? > which class? which function?) > i read the most lsan source code, but i don't find that scan code : ( ScanRangeForPointers() in projects/compiler-rt/lib/lsan/lsan_common.cc is what you're looking for. > 2. why 64bit platform is ok? a random 8 byte sequences don't like a pointer? Please refer to https://code.google.com/p/valgrind-variant/wiki/LeakCheckingOn32bits The example code there demonstrates the difference between leak checking on 32- and 64-bit platforms. > 3. is it possible that we ignore some large malloc memory scan? then we can > reduce missing memory leak percent. If there are pointers in the ignored range, you're going to get false positives (the tool will report leaks of memory that didn't leak). This is even worse, because developers soon start to ignore reports from a tool with many false positives. > i know it can't make sure all leak will be detect, but sometimes it will > be usefull, right? > after all there is no better tools to detect memory leak on 32bit > platform. There are tools like Valgrind and Dr.Memory, which are slightly better, because they only scan initialized memory for pointers. However 32-bit Valgrind is also prone to false negatives (see the link above), I suppose the same applies to Dr.Memory as well. > > Thanks so much for response : ) > > 在 2014年11月28日星期五UTC+8下午2时06分56秒，Glider写道： >> >> Hello. >> >> The main problem with 32-bit leak checking is that too many random 4-byte >> sequences look like pointers. >> In order to find live objects in the heap LSan scans the memory that's >> already known to be live and looks for pointers in it. Because LSan doesn't >> know anything about data types, it has to aggressively treat everything that >> looks like a pointer as such. This approach may lead to false negatives >> (missing memory leaks), and on 32-bit platforms it actually does. >> >> HTH, >> Alex >> >> On Nov 28, 2014 5:15 AM, "wuning.shi" <wunin...@gmail.com> wrote: >>> >>> hello >>> >>> we want to use lsan on 32 bit arm platform, and i see your explain of the >>> main problem. >>> https://code.google.com/p/address-sanitizer/issues/detail?id=294 >>> >>> i have read some source code. >>> but i still can't understand why 64 bit platform has no problem. why lsan >>> need check a 32bit number? >>> >>> can you explain more about lsan design principle please? >>> >>> thanks very much : ) >>> > -- > You received this message because you are subscribed to the Google Groups > "address-sanitizer" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to address-sanitizer+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/d/optout.
-- You received this message because you are subscribed to the Google Groups "address-sanitizer" group. To unsubscribe from this group and stop receiving emails from it, send an email to address-sanitizer+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
