[ardovm]

The address sanitizer can be a useful tool for debugging.

Many of the tools built for the purpose of building AOO are leaking memory by design (they can afford that: they have very short execution time and are not designed to be distributed) so the leak detection is disabled during the AOO build.

This PR was tested under Linux only. I had to disable unit tests. Some modules do not seem to build with the sanitizer enabled... but this could be a good start for further debugging.

[DamjanJovanovic]

I had to disable unit tests. Some modules do not seem to build with the sanitizer enabled... but this could be a good start for further debugging.

I tried this before, and in a 2017-11-26 email entitled "How to evaluate OSS-Fuzz issues", I said that:

"The whole of AOO can't be built with ASAN for many reasons. Some of the linker options we use conflict with ASAN. Even when that's fixed I got other errors. Worst of all, some internal build tools (can't remember which) rely on buffer overflows to work properly 🙈, and ASAN causes them to crash and break the build instead..."

[ardovm]

My attempt was more successful than yours, but it's not my merit. ;-)
For example, c56f8ff and 8ebeb3b were not there when you tried. I just discovered them while trying to enable the sanitizer for AOO41X.
I could successfuly build trunk on my Linux system with the sanitizer enabled just adding 776d9c2.

[ardovm]

Sorry, I forgot an important note: the build with the sanitizer fails on testools module, bridgetest test.
I had to disable it for the build to succeed.