[cfe-dev] using a faster libunwind on linux for LSAN_OPTIONS=fast_unwind_on_malloc=false ?

[Milian Wolff via cfe-dev]

Hey there,

I'm regularly using LSAN on user applications where some third party 
dependencies are compiled without framepointers (and without sanitizers). In 
such cases, the leak reports from LSAN are usually garbage, and the fix is to 
use `LSAN_OPTIONS=fast_unwind_on_malloc=false`. But that comes at a hefty 
performance price, which makes it infeasible for me to have it enabled all the 
time.

I wonder - would it be possible to use libunwind [1] (not the one from LLVM), 
or alternatively improve the LLVM-libunwind to apply similar caching 
techniques?

[1]: https://github.com/libunwind/libunwind/

I'm using libunwind in heaptrack, where I also record backtraces for every 
single heap allocation. I have the feeling that the performance impact of that 
is much lower compared to using `LSAN_OPTIONS=fast_unwind_on_malloc=false`.

Has this ever been discussed?

Thanks

-- 
Milian Wolff
mail at milianw.de
http://milianw.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20201017/0c9dfaad/attachment.sig>