Subject:
[ruby-ffi] Re: Possible issues with Ruby FFI on ARM processor (Raspberry PI)
From:
Wayne Meissner
Date:
2/27/13 4:46 PM
To:
ruby-ffi@googlegroups.com

The glibc error does tend to indicate memory corruption issues - given you eliminated #1 and #3 as reasons, then its down to memory ownership, or inadequate allocation size.

You did miss one autorelease - config.pointer.autorelease=false

Also, when you allocate a MemoryPointer with a block, it is always freed at the end of the block, regardless of the autorelease setting - so you'll need to convert them to the non-block form.

It is also possible there are bugs in libspotify itself - you could try converting your C example to use malloc() allocated memory instead of stack/global memory (i.e. emulate what FFI does), and see if it fails in the same way.


On Thursday, 28 February 2013 07:12:29 UTC+10, Kim Burgestrand wrote:
On Monday, February 25, 2013 10:49:54 PM UTC+1, Wayne Meissner wrote:

Off the top of my head, a couple of issues could be:

1) data sizes/layout.  Verify that the sizes of all your FFI structs match the C ones, and that the sizes & offsets of members match.
That’s a good idea. I tried it, and they do match (was kind of hoping they didn’t — updated the gist with the offset-checking code).
 
2) Given it only crashes on exit, it could be some sort of GC/finalizer interaction.  Set all the MemoryPointer instances, including the ones backing the structs to autorelease=false - this causes FFI to disown the memory, so it won't try to free it.
Hm, I’ve updated the gist with these changes. Unfortunately the crashes still occur. I know how vague crash messages can be, but in case it might ring any bells: `*** glibc detected *** ruby: free(): invalid next size (fast): 0x01a3acf0 ***`

3) Endian-ness issues in ruby-ffi itself.  Is that arm arch big-endian or little-endian?
From what I could gather from googling is that it’s primarily little-endian. By primarily is that it apparently has some big-endian support if you do a little bit of hoop-jumping.

Thanks for your time Wayne, I appreciate it.

--
 
---
You received this message because you are subscribed to the Google Groups "ruby-ffi" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ruby-ffi+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.