Hey, Sorry if this is the wrong list to be subscribed to for this
question.
I'm working on a haiku os x86_64 port. While working on the port I made
a new gcc 4.3.3 target for x86_64 haiku elf files.
However in practice I've noticed some weird behavior in the compiled
binaries.
While compiling with the -m32 option. Which should output an elf_i386
binary it compiles fine, but executing the code has some weird effects.
For example. in this code segment:
if (address < KERNEL_BASE || address + size > sNextVirtualAddress) {
panic("mmu_free: asked to unmap out of range region (%p, size %
lx)\n",
(void *)address, size);
}
(address + size > sNextVirtualAddress) always returns true... no matter
what size (address + size) actually is.
This code segment works perfectly with the normal x86 gcc 4.3.3 target
we've been using.
If I typecast size to be a void* the code works perfectly with the
x86_64 target(compiled as 32bit code). However, this isn't needed with
the normal x86 target. I'm assuming since it needs to be typecasted this
is a sign that there is a bigger issue in the x86_64 target. Is that a
correct assumption?
Secondly, While building gcc it outputs a x86_64 gcclib, but it doesn't
output the x86 gcclib. Is there an option or setting in the build script
that will output the gcclib for both architectures?
Here's some code for the haiku x86_64 target.
http://dev.haiku-os.org/attachment/ticket/1141/x86_64-buildtools-buildtoolstrunk.patch
I'm sorry that I'm diving into gcc without truly understanding the code
base, but I'm on a schedule to complete my project. So any help would be
extremely appreciated.
