[Rcpp-devel] AddressSanitizer: stack-use-after-scope Error

[Kevin Ushey]

Are you able to share a reproducible example?

The CRAN check report from
https://www.stats.ox.ac.uk/pub/bdr/memtests/clang-ASAN/MatchIt/00check.log
points at MatchIt/src/nnm.cpp:182:24. You could probably distill the code
used in that location into a smaller reproducible example.

Best,
Kevin

On Sun, Nov 15, 2020 at 3:31 PM Noah Greifer <noah.greifer at gmail.com> wrote:

> Hello,
>
> I use Rcpp in the package I maintain, MatchIt. I recently upload it to
> CRAN with no problems, but then received a message saying there were
> "additional issues" I needed to address, which were titled clang-ASAN and
> gcc-ASAN. Looking at the log, it seems that the error was triggered when I
> used an R function (mahalanobis()) inside the Rcpp code. I realized that I
> had not properly initialized the function in my Rcpp script. Although the
> package functions ran correctly, perhaps something about the way the memory
> was allocated caused this issue. I think I fixed the issue, but I want to
> check it without uploading it to CRAN again. This issue was not picked up
> by R cmd check or by the check in winbuilder, so I'm not sure how to check
> if I have solved the issue. CRAN even accepted my package before letting me
> know about this issue.
>
> In searching for solutions to the problem of being able to check whether
> the issue in my package remains, I came across this
> <http://dirk.eddelbuettel.com/code/sanitizers.html> article posted by
> Dirk ("sanitizers: Testing R packages for ASAN and UBSAN errors"). It seems
> like this article is about using sanitizers, which helps you figure out if
> your build of R is able to detect the memory allocation errors I was
> facing, but doesn't give clear instructions on how to set up R to do this.
> As someone for whom using Rcpp was my first use of programming outside R,
> the instructions given in the article and in section 4.3 of Writing R
> Extensions are completely foreign to me and seem to require a deep
> knowledge of C++.
>
> I'm wondering if someone can provide me with a clear procedure on how to
> check if the memory allocation problem still remains in my package. Thank
> you.
>
> Noah Greifer
>
> PS What seems to be the relevant text in the error message is pasted below:
>
> ==1503820==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7ffc220ca7a0 at pc 0x7fc2fb4668ad bp 0x7ffc220ca5b0 sp 0x7ffc220ca5a8
> READ of size 8 at 0x7ffc220ca7a0 thread T0
>     #0 0x7fc2fb4668ac in Rcpp::PreserveStorage<Rcpp::Vector<14, Rcpp::PreserveStorage> >::get__() const /data/gannet/ripley/R/test-clang/Rcpp/include/Rcpp/storage/PreserveStorage.h:26:20
>     #1 0x7fc2fb4668ac in Rcpp::Vector<14, Rcpp::PreserveStorage>::size() const /data/gannet/ripley/R/test-clang/Rcpp/include/Rcpp/vector/Vector.h:277:39
>     #2 0x7fc2fb4668ac in Rcpp::sugar::Vectorized<&(sqrt), true, Rcpp::Vector<14, Rcpp::PreserveStorage> >::size() const /data/gannet/ripley/R/test-clang/Rcpp/include/Rcpp/sugar/block/Vectorized_Math.h:40:50
>
> ...
>
> Address 0x7ffc220ca7a0 is located in stack of thread T0 at offset 448 in frame
>     #0 0x7fc2fb46644f in void Rcpp::Vector<14, Rcpp::PreserveStorage>::assign_sugar_expression<Rcpp::sugar::Vectorized<&(sqrt), true, Rcpp::Vector<14, Rcpp::PreserveStorage> > >(Rcpp::sugar::Vectorized<&(sqrt), true, Rcpp::Vector<14, Rcpp::PreserveStorage> > const&) /data/gannet/ripley/R/test-clang/Rcpp/include/Rcpp/vector/Vector.h:1038
>
>   This frame has 1 object(s):
>     [32, 48) 'ref.tmp.i.i.i.i.i' <== Memory access at offset 448 overflows this variable
> HINT: this may be a false positive if your program uses some custom stack unwind mechanism, swapcontext or vfork
>       (longjmp and C++ exceptions *are* supported)
> SUMMARY: AddressSanitizer: stack-use-after-scope /data/gannet/ripley/R/test-clang/Rcpp/include/Rcpp/storage/PreserveStorage.h:26:20 in Rcpp::PreserveStorage<Rcpp::Vector<14, Rcpp::PreserveStorage> >::get__() const
> Shadow bytes around the buggy address:
>   0x1000044114a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>
> _______________________________________________
> Rcpp-devel mailing list
> Rcpp-devel at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20201116/32c0cca4/attachment.html>