diff options
| author | Dan Liew <dan@su-root.co.uk> | 2020-04-20 17:22:35 -0700 |
|---|---|---|
| committer | Dan Liew <dan@su-root.co.uk> | 2020-04-20 19:08:08 -0700 |
| commit | 7039773b240b6eee00b5be5bc325c5c57501788a (patch) | |
| tree | cce416e5368485f35ef89d743d9e9e3e77ff88b9 /lldb/source/Plugins/ScriptInterpreter/Python/PythonDataObjects.cpp | |
| parent | f0019d4ff29c31ba12e5ecd145552a266d1ffe37 (diff) | |
Unbreak ASan runtime in the simulators.
Summary:
861b69faee5df8d4e13ef316c7474a10e4069e81 (rdar://problem/58789439) while
fixing symbolization for TSan completely broke ASan's runtime for the
simulators.
The problem with the previous patch is that the memory passed to
`putenv()` was poisoned and when passed to `putenv()` it tripped
an interceptor for `strchr()` which saw the memory was poisoned and
raised an ASan issue.
The memory was poisoned because `AtosSymbolizerProcess` objects
are created using ASan's internal allocator. Memory from this
allocator gets poisoned with `kAsanInternalHeapMagic`.
To workaround this, this patch makes the memory for the environment
variable entry a global variable that isn't poisoned.
This pass also adds a `DCHECK(getenv(K_ATOS_ENV_VAR))` because the
following DCHECK would crash because `internal_strcmp()` doesn't
work on nullptr.
rdar://problem/62067724
Reviewers: kubamracek, yln
Subscribers: #sanitizers, llvm-commits
Tags: #sanitizers
Differential Revision: https://reviews.llvm.org/D78525
Diffstat (limited to 'lldb/source/Plugins/ScriptInterpreter/Python/PythonDataObjects.cpp')
0 files changed, 0 insertions, 0 deletions