Error due to reading of structs after call to free about hyperscan-java HOT 7 CLOSED

Here is a thread that discusses this issue.

from hyperscan-java.

Not that the problem is not Expression.validate per se, but calls to hs_free_compile_error, which also occur in Database.handleErrors.

from hyperscan-java.

Yeah, thats true.

I was able to reproduce it.

Basically it boils down to:
a) disabling autoread on the structs and freeing the structs using hs_free_compile_error
or
b) not freeing the struct and leaving it to the garbage collector, since the JNA implementation should take care of it

from hyperscan-java.

hs_error_t HS_CDECL hs_free_compile_error(hs_compile_error_t *error) {
#if defined(FAT_RUNTIME)
    if (!check_ssse3()) {
        return HS_ARCH_ERROR;
    }
#endif
    freeCompileError(error);
    return HS_SUCCESS;
}

That's the implementation of the hs_free_compile_error. In case it's a "fat library" (supporting multiple instruction sets), it makes sure at least sse3 is available. After that it calls freeCompileError(error), which is implemented as:

void freeCompileError(hs_compile_error_t *error) {
    if (!error) {
        return;
    }
    if (error == &hs_enomem || error == &hs_einternal ||
        error == &hs_badalloc) {
        // These are not allocated.
        return;
    }

    hs_misc_free(error->message);
    hs_misc_free(error);
}

hs_misc_free then just calls the default deallocator for the platform (free) or in case a user provided hyperscan with their own deallocator by using hs_set_misc_allocator() it then uses that.

from hyperscan-java.

Since the memory is not allocated by JNA, it seems like it won't free the memory on garbage collection, so "disabling autoread on the structs and freeing the structs using hs_free_compile_error" is the proper solution.

I'm testing it right now and will push it later.

from hyperscan-java.

Fixed in v0.4.11

from hyperscan-java.

👍

from hyperscan-java.