Thank you for your re-post, @kyab. I'm interested in your hack. And there needs some additional hacks, I think.
I'll send pull request to you later (maybe tomorrow). Could you review my additional patch?

Yes, I will review your additional patch.

Thanks for your review and merge, @kyab.
In addition, I think it's better to reduce function call arguments.
I have a plan to do. Please stay tuned.

I just noticed our patch crashes in some phase on loading *.mrb.

[koji@macbookpro:~/work/mruby]$ cat foo.rb
puts "Hello"
[koji@macbookpro:~/work/mruby]$ mrbc foo.rb
[koji@macbookpro:~/work/mruby]$ mruby -b foo.mrb
failed to load mrb file: foo.mrb
mruby(8113) malloc: *** error for object 0x6000000000000000: pointer being freed was not allocated
*** set a breakpoint in malloc_error_break to debug
Abort trap: 6

Maybe more hack required..

It seems a free twice bug that can find out easily, right?
Anyway. I'll send pull request to your branch whether you close this request for now, or not.

I did a double check with foo.rb. I could run without crashes.
My environment is OSX 10.6.8, Apple genuine gcc4.2.1

@monaka It's strange, it should crash or at least fail to load foo.mrb.
The root problem is crc mismatch. In our patch, dump side calculate crc with little endian, but load side calculate crc with bin(big endian), in case of hex dump(dump to *.mrb. not *.c).

I've fixed, but might be ugly hack.. Could you please retry to reproduce crash(or fail) and review my fix?
https://github.com/kyab/mruby_min/commit/67f6859156bfd0a9c33a2ef2ce2af400092e102e

I reproduced the crash. And I checked it is stopped by the patch 67f6859.
The patch is effective ... but it's better you rearrange your patches, I feel, as you might feel.

I put my comment on #964

IMO, this issue can be closed as the second suggestion is on #1338. @kyab WDYT?

I think this issue can be closed.