Hello Kurt, Can you then give me an account to the machine, which is _exactly_ equivalent to the buildd? I tested it on pergolesi, and it worked great (of course I couldn't do a full build, because of missing build dependencies).
And I need _real_ information about why amd64 hackers thought, that this code supports amd64. What I see about this amd64 thing is the following anyway: 1. report arrives, that the code supports amd64, so build for it 2. it build on a developer machine, which is opened for me 3. you say, that that the assembly is for i386 only. I also need test results about booting with the built code (if we can build it once). I ask for _real_ help from the amd64 team, or I will simply go back for i386 only with etherboot. Of course I think that it is not a problem at all for Debian while upstream doesn't support amd64. If the porter team has no time or ability to produce a buildable and workable patch, I don't think that I have a real chance to success. Thanks for your help, Gergely On Tue, 22 Mar 2005 22:12:21 +0100, Kurt Roeckx <[EMAIL PROTECTED]> writes: > reopen 265919 > thanks > > Hi, > > The latest upload did not build properly at all. > > From the build log: > gcc -DCONFIG_PCI -DCONFIG_ISA -DASK_BOOT=3 -DBOOT_FIRST=BOOT_NIC > -DALLOW_ONLY_ENCAPSULATED -DBACKOFF_LIMIT=7 -DCONGESTED -DTAGGED_IMAGE > -DELF_IMAGE -DDOWNLOAD_PROTO_TFTP -DPXE_IMAGE -DPXE_EXPORT -Os -ffreestanding > -Wall -W -Wno-format -DPCBIOS -fstrength-reduce -fomit-frame-pointer > -march=i386 -falign-jumps=1 -falign-loops=1 -falign-functions=1 -mcpu=i386 > -DVERSION_MAJOR=5 -DVERSION_MINOR=3 -DVERSION=\"5.3.14\" -I include -I > arch/i386/include -DARCH=i386 -DCONFIG_X86_64 -o bin/undi.o -c > arch/i386/drivers/net/undi.c > /tmp/ccLhp0Px.s: Assembler messages: > /tmp/ccLhp0Px.s:699: Error: `rm_undi_call(%rip)' is not a valid 32 bit > base/index expression > /tmp/ccLhp0Px.s:700: Error: `rm_undi_call_end(%rip)' is not a valid 32 bit > base/index expression > /tmp/ccLhp0Px.s:701: Warning: `movq' is not supported on `i386' > /tmp/ccLhp0Px.s:701: Error: suffix or operands invalid for `movq' > /tmp/ccLhp0Px.s:702: Warning: `movq' is not supported on `i386' > /tmp/ccLhp0Px.s:702: Error: bad register name `%r8' > /tmp/ccLhp0Px.s:704: Error: bad register name `%r9d' > [...] > > This seems to be caused by the real mode / 16 bit > assembler in it. I think it needs to return to the > sledgehammer arch / 64 bit mode. > > Then we get: > arch/i386/core/start32.S: Assembler messages: > arch/i386/core/start32.S:473: Error: suffix or operands invalid for `and' > arch/i386/core/start32.S:476: Error: suffix or operands invalid for `and' > arch/i386/core/start32.S:480: Error: suffix or operands invalid for `and' > > Those lines contain something like: > andq $0xffffffff, %rbx > > I think those should be changed to: > andl $0xffffffff, %ebx > > I think they can even be removed. > > Next, in arch/i386/core/realmode.c you have some pushl's > and a popl. They're not supported on x86_64, you only > have pushq/popq. > > There are probably a few more problems with it to get it to > build. > > > Kurt -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
