536568 - nanojit::Allocator should return 8-aligned blocks

Status: RESOLVED WONTFIX

(Core Graveyard :: Nanojit, defect)

Description

[Steven Johnson]

Attachment: Demonstrate the issue

Various bits of code (notably patches for https://bugzilla.mozilla.org/show_bug.cgi?id=533546) assume that the nanojit Allocator returns 8-aligned blocks, but it doesn't seem to enforce this. We should either guarantee 8-aligned pointers, or (at least) add a new API for allocating such pointers for the cases that really matter.

See enclosed patch with assertion that fails.

Comment 1

[Steven Johnson]

BTW: the nature of Allocator is that it relies on the underlying system allocator to return 8-aligned chunks. The failure mode I see is on MacTel32 Debug avmshell builds; other builds may or may not be smarter, but IMHO we should require all builds to provide a predictable alignment.

Comment 2

[Lars T Hansen]

(In reply to comment #1)
> BTW: the nature of Allocator is that it relies on the underlying system
> allocator to return 8-aligned chunks. The failure mode I see is on MacTel32
> Debug avmshell builds; other builds may or may not be smarter, but IMHO we
> should require all builds to provide a predictable alignment.

Allocator uses mmfx_alloc which is not guaranteed to return 8-aligned memory - it only guarantees memory aligned as strictly as the platform requires.  An important reason for that fairly soft requirement is that mmfx_alloc may map directly to malloc, and malloc does not guarantee 8-alignment.

Comment 3

[Steven Johnson]

Eek. Good to know. I'm guessing that only a subset of allocations need to be 8-aligned but those are pretty important. Perhaps we should add a separate API to Allocator to specifically request 8-alignment?

Comment 4

[Nicholas Nethercote [inactive]]

Nanojit has been dead for several years. Its Bugzilla component has been moved to the graveyard (bug 984276).

I checked all the open bugs. They're all uninteresting, so I'm WONTFIXing them all. Apologies for the bugspam.