[09:12] <mjg59> doko: Our elilo problems are binutils-related
[09:12] <mjg59> Any ideas what might have broken efi-ia32 building?
[10:47] <doko> mjg59: no, is the source package in the archive?
[10:47] <mjg59> Yes
[10:48] <mjg59> I haven't traced whether it's gnu-efi that ends up broken (and hence all efi apps are), or if it's ok and elilo is broken
[10:50] <doko> hmm, how could I test if it's broken or not?
[10:51] <mjg59> doko: Good question. 
[10:51] <mjg59> Unfortunately the answer is probably "Run the binaries", but...
[10:52] <mjg59> In theory it's possible to run an EFI environment under qemu or something, but I've never done it
[10:52] <doko> me neither :-/
[10:54] <doko> is it supposed to work with vmware?
[10:55] <mjg59> I'd assume so
[10:56] <mjg59> But I don't know where to get hold of the test environment
[12:31] <mjg59> doko: Ok, the breakage appears to purely be in elilo
[12:33] <mjg59> I can build gnu-efi with the new binutils, and elilo still works as long as I build it with the old binutils
[12:35] <doko> mjg59: ok, so I assume we have to do a binary search with different binutils.
[12:37] <mjg59> doko: The obvious difference (looking at the headers) is that the .dynsym and .reloc sections are in the reverse order
[12:39] <mjg59> The only difference in the disassembly is that some of the addresses are slightly different
[12:39] <mjg59> In the broken one, they're all 4 bytes later
[12:39] <mjg59> So my suspicion is in the section ordering...
[12:40] <mjg59> Or the fact that the .dynsym section is slightly larger in the working one
[12:41] <doko> mjg59: I'm looking at it today
[12:41] <mjg59> doko: Ok, thanks
[12:41] <mjg59> doko: Would you like a copy of a working and a broken binary?
[12:46] <doko> mjg59: sure, might help. most likely I'll have to ask you anyway ...
[12:47] <mjg59> doko: www.codon.org.uk/~mjg59/tmp/efi
[12:50] <doko> got it
[01:06] <mjg59> Ok, if I edit the linker script to force the section ordering to be the same, it still fails
[01:06] <mjg59> The only obvious difference now is that the .dynsym section is smaller
[01:06] <mjg59> (In the broken one)
[04:26] <doko> mjg59: could you check http://people.ubuntu.com/~doko/binutils_2.16.1cvs20060314-0ubuntu1_i386.deb (current upstream)
[06:04] <mjg59> doko: Will do once I get home (30 minutes or so)
[08:07] <mjg59> doko: Still broken
[08:09] <doko> mjg59: hmm, ok, going backward then tomorrow
[08:37] <mjg59> doko: Are the earlier cvs packages available anywhere?