Move allocators to separate files #151
Conversation
carenas
force-pushed
the
mem_allocator
branch
from
366ecd6
to
ade1a6a
Compare
carenas
force-pushed
the
mem_allocator
branch
2 times, most recently
from
3a730eb
to
6364ecb
Compare
|
Are you ok with the move patch? Do you want to merge it with this patch before landing (I can close my pr and let you do the whole thing)? |
|
I don't think the original patch is ready for merging, even after squashing this change on top, which is needed to fix building in macOS (only test it with Intel) and avoid crashes in 32bit NetBSD with ProtExec (should likely be also broken in Linux and other 32bit *NIX but didn't test it yet). |
|
It is very likely there will be issues, so it will take time to fix everything. Hopefully we will get feedback from users as well. I never thought the allocators will be this complex, and we need to do something with the growing complexity at some point. |
|
The jit tests works on 32 bit linux with prot exec allocator. Btw CHUNK_SIZE is configurable now, so Prot allocators could use bigger files than 64KB, something like 1MB should be better. |
|
Why do we want to combine netbsd and linux? The point is separating OS-es for easier maintainability. |
carenas
force-pushed
the
mem_allocator
branch
2 times, most recently
from
2344d8e
to
9a758ae
Compare
|
I tried to get a branch rebased and integrated with all changes, but it is not working. Will keep looking at it and retarget in this PR then when it is ready, feel free to close your original PR |
carenas
force-pushed
the
mem_allocator
branch
from
9a758ae
to
0c6e2aa
Compare
|
Ok. You can make a single patch from it (combining the original and your work). I hope you can do it soon. |
carenas
changed the title
Consolidated changes to refactor the allocator files into mostly independent files per supported platform