--- Begin Message ---
|
Subject: |
Memory leaks in test suite make check after building with build-asan |
|
Date: |
Mon, 10 Sep 2018 07:31:14 +0000 |
Hi everyone
I think I may have found 3 memory leak bugs in sed.
When I built in the following way.
git clone git://git.sv.gnu.org/sed sed-build-asan
cd sed-build-asan
./bootstrap
./configure
make build-asan
make check
Make check told me that gnulib-tests/test-suite.log would contain details.
You can find this file attached as test-suite.log.
The version of sed which I am using, according to make check, is GNU sed
4.5.31-261c7-dirty.
My system is an up-to-date arch linux
uname -a
Linux archlinux 4.18.5-arch1-1-ARCH #1 SMP PREEMPT Fri Aug 24 12:48:58 UTC 2018
x86_64 GNU/Linux
Please forgive me if there is anything incomplete or wrong about this report.
This is my first time submitting a bug. I am happy to help out by running some
patches or something but I am no more than a novice to C programming.
- deltatau
test-suite.log
Description: Text Data
--- End Message ---
--- Begin Message ---
|
Subject: |
Re: bug#32685: Memory leaks in test suite make check after building with build-asan |
|
Date: |
Tue, 9 Oct 2018 07:46:16 -0600 |
|
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.9.1 |
Hello,
On 10/09/18 01:31 AM, deltatau wrote:
[...] This is my first time submitting a bug. I am happy to help
out by running some patches or something but I am no more than a
novice to C programming.
First and foremost, thank you for taking the time to run the tests
and reporting this bug in details.
This is helpful and encouraged.
As for the bugs:
I think I may have found 3 memory leak bugs in sed.
Make check told me that gnulib-tests/test-suite.log would contain
details. You can find this file attached as test-suite.log.
FAIL: test-copy-acl.sh
======================
[...]
Direct leak of 17 byte(s) in 1 object(s) allocated from:
#0 0x7fa615dbe019 in __interceptor_malloc
/build/gcc/src/gcc/libsanitizer/asan/asan_malloc_linux.cc:86
#1 0x563664a51531 in fread_file
/home/ksk/delete/sed-build-asan/gnulib-tests/read-file.c:73
#2 0x563664a51908 in internal_read_file
/home/ksk/delete/sed-build-asan/gnulib-tests/read-file.c:147
#3 0x563664a51a36 in read_file
/home/ksk/delete/sed-build-asan/gnulib-tests/read-file.c:174
#4 0x563664a50545 in main
/home/ksk/delete/sed-build-asan/gnulib-tests/test-sameacls.c:58
#5 0x7fa615924222 in __libc_start_main (/usr/lib/libc.so.6+0x24222)
[...]
So, these are indeed memory leaks (and it's nice that using the recent
'build-asan' addition it was easy to find).
There are two subtleties here:
First,
These leaks happen in the gnulib tests - gnulib is a separate project
from sed. gnulib is used by many gnu project for common code.
I will report this to their mailing list (and CC you).
Second,
In C programming, it is sometimes customary to skip freeing memory
when it is known that the program will terminate soon.
When a program terminates, the operating system automatically
frees all the program's memory. So freeing the memory explicitly is
redundant.
This is even more so in small tests - which are more concerned with
the correctness of the code, and less with memory issues.
However,
With automated testing (like ASAN) it could be useful to
free the allocated memory to reduce noise and false positives.
In sed, we recently added additional code to free memory *if*
we are running under such memory checkers like ASAN:
https://git.savannah.gnu.org/cgit/sed.git/commit/?id=35b30b124c6f05fcfdd09a2dadabcfe570856c79
As such, I'm marking this as "not a bug" (for sed)
and closing it.
Thanks again and keep up the good work,
regards,
- assaf
--- End Message ---