Conditional allocations shouldn't fail if cond=false, size=0 -> nullptr result #7255
Comments
|
This may be an issue unique to the C backend, if that's where you observed this. I see code that handles this in CodeGen_Posix.cpp:307 |
|
Yep. That logic looks simpler than what I have for a fix, I'll try replicating that instead. Fix on the way either way. |
steven-johnson
added a commit
that referenced
this issue
Dec 28, 2022
Allocations can be conditional; if the condition evaluates to false, we end up calling `halide_malloc(0)` (or `halide_tcm_malloc(0)` in the xtensa branch). Since it's legal via spec for `malloc(0)` to return nullptr, we need to be cautious here: if we are compiling with assertions enabled, *and* have a malloc() (etc) implementation that returns nullptr for alloc(0), we need to skip the assertion check, since we know the result won't be used. Note: a similar check will be inserted in the xtensa branch separately. Note 2: LLVM backend already has this check via Codegen_Posix.cpp
|
FYI: via inserting a crash at Codegen_Posix.cpp:298 and rebuilding the universe, I can't find a single AOT case in our tests that actually triggers the |
steven-johnson
added a commit
that referenced
this issue
Dec 28, 2022
…) (#7256) * Conditional allocations shouldn't fail for size=0 in C++ backend (#7255) Allocations can be conditional; if the condition evaluates to false, we end up calling `halide_malloc(0)` (or `halide_tcm_malloc(0)` in the xtensa branch). Since it's legal via spec for `malloc(0)` to return nullptr, we need to be cautious here: if we are compiling with assertions enabled, *and* have a malloc() (etc) implementation that returns nullptr for alloc(0), we need to skip the assertion check, since we know the result won't be used. Note: a similar check will be inserted in the xtensa branch separately. Note 2: LLVM backend already has this check via Codegen_Posix.cpp * Update CodeGen_C.cpp
ardier
pushed a commit
to ardier/Halide-mutation
that referenced
this issue
Mar 3, 2024
…ide#7255) (halide#7256) * Conditional allocations shouldn't fail for size=0 in C++ backend (halide#7255) Allocations can be conditional; if the condition evaluates to false, we end up calling `halide_malloc(0)` (or `halide_tcm_malloc(0)` in the xtensa branch). Since it's legal via spec for `malloc(0)` to return nullptr, we need to be cautious here: if we are compiling with assertions enabled, *and* have a malloc() (etc) implementation that returns nullptr for alloc(0), we need to skip the assertion check, since we know the result won't be used. Note: a similar check will be inserted in the xtensa branch separately. Note 2: LLVM backend already has this check via Codegen_Posix.cpp * Update CodeGen_C.cpp
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Allocations can be conditional; if the condition evaluates to false, we end up calling
halide_malloc(0). Since it's legal via spec formalloc(0)to return nullptr, we need to be cautious here: if we are compiling with assertions enabled, and have a malloc() (etc) implementation that returns nullptr for alloc(0), we need to skip the assertion check, since we know the result won't be used... but the assertion would fail because the (ignored) result is null.NOTE: I've only seen this in the C++ backend; so far, I can't replicate it with LLVM backends.
The text was updated successfully, but these errors were encountered: