Skip to contentSkip to conversation

gcc-4.6.3, was Re: Debian on mac68k

DISCUSSION Finn Thain 7 participants 19 messages Nov 6, 2016 – Nov 22, 2016 11 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

Finn Thain

author
#11 lore raw

Has anyone else tried the m68k cross-compiler from kernel.org and observed strange behaviour from the binaries it produces?

I used that cross-compiler (the i686 one) to build the kernel seen booting below. The resulting binary consistently crashes two different systems (LC III and LC 475), both during kernel startup. These two machines use different drivers and have different failure modes, but the failures are completely repeatable. The failures went away when I reverted to my usual (patched) gcc-4.4.6.

Perhaps gcc-4.6.3 is more strict than gcc-4.4, in such a way that would affect code generation, when presented with incorrect or undefined code? Are there any warnings that could be enabled to discover that code?

Is there some way to run the GCC test suite against this cross-compiler?

Quoted 125 lines · Mac User

Please see attached for console log of LC III booting "vmlinux-4.1.35-mac_scsi+". [...]

ABCFGHIJK
Linux version 4.1.35-mac_scsi+ (fthain@nippy) (gcc version 4.6.3 (GCC) ) #1 Tue6 Saving 402 bytes of bootinfo
Detected Macintosh model: 27
VIA1 at 50f00000 is a 6522 or clone
VIA2 at 50f26000 is an RBV
Apple Macintosh LC III
Built 1 zonelists in Zone order, mobility grouping on. Total pages: 9135 Kernel command line: root=/dev/sda6 console=ttyS0 nologo
PID hash table entries: 256 (order: -2, 1024 bytes)
Dentry cache hash table entries: 8192 (order: 3, 32768 bytes) Inode-cache hash table entries: 4096 (order: 2, 16384 bytes) Sorting __ex_table...
Memory: 32572K/36864K available (2825K kernel code, 247K rwdata, 664K rodata, 1) Virtual kernel memory layout:

    vector  : 0x003879e8 - 0x00387de8   (   1 KiB)
    kmap    : 0xd0000000 - 0xf0000000   ( 512 MiB)
    vmalloc : 0x02800000 - 0xd0000000   (3288 MiB)
    lowmem  : 0x00000000 - 0x02400000   (  36 MiB)
      .init : 0x003ab000 - 0x003c7000   ( 112 KiB)
      .text : 0x00001000 - 0x002c3784   (2826 KiB)
      .data : 0x002c6bf0 - 0x003aaa64   ( 912 KiB)
      .bss  : 0x00387900 - 0x003aaa64   ( 141 KiB)

SLUB: HWalign=16, Order=0-3, MinObjects=0, CPUs=1, Nodes=8 NR_IRQS:72
Killing onboard sonic... Done.
WARNING: Persistent clock returned invalid value!
Check your CMOS/BIOS settings.
Console: colour dummy device 80x25
console [ttyS0] enabled
Calibrating delay loop... 5.81 BogoMIPS (lpj=29056)
pid_max: default: 32768 minimum: 301
Mount-cache hash table entries: 1024 (order: 0, 4096 bytes) Mountpoint-cache hash table entries: 1024 (order: 0, 4096 bytes)
devtmpfs: initialized
clocksource jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 1911s
NET: Registered protocol family 16
NuBus: Scanning NuBus slots.
Slot E:
Board resource:

    type: [cat 0x1 type 0x0 hw 0x0 sw 0x0]
    name: Farallon EtherMac LC-TP
    board id: 0x8
    vendor info:
    ID: Farallon
    revision: 001
    part: PN 592a-TP
  Function 0x80:
    type: [cat 0x4 type 0x1 hw 0x10c sw 0x100]
    name: Farallon EtherMac LC-TP
    memory offset: 0x000d0000
    MAC address: 00:00:c5:43:36:11

SCSI subsystem initialized
NET: Registered protocol family 2
TCP established hash table entries: 1024 (order: 0, 4096 bytes)
TCP bind hash table entries: 1024 (order: 0, 4096 bytes)
TCP: Hash tables configured (established 1024 bind 1024)
UDP hash table entries: 256 (order: 0, 4096 bytes)
UDP-Lite hash table entries: 256 (order: 0, 4096 bytes)
NET: Registered protocol family 1
RPC: Registered named UNIX socket transport module.
RPC: Registered udp transport module.
RPC: Registered tcp transport module.
RPC: Registered tcp NFSv4.1 backchannel transport module.
futex hash table entries: 256 (order: -1, 3072 bytes)
NFS: Registering the id_resolver key type
Key type id_resolver registered
Key type id_legacy registered
Block layer SCSI generic (bsg) driver version 0.4 loaded (major 253) io scheduler noop registered
io scheduler deadline registered (default)
macfb: framebuffer at 0x60b00000, mapped to 0xd0000000, size 507k
macfb: mode is 832x624x8, linelength=832
Console: switching to colour frame buffer device 104x39
fb0: Sonora frame buffer device
pmac_zilog: 0.6 (Benjamin Herrenschmidt <[email protected]>) scc.0: ttyS0 at MMIO 0x50f04002 (irq = 4, base_baud = 230400) is a Z85c30 ESCC t scc.1: ttyS1 at MMIO 0x50f04000 (irq = 4, base_baud = 230400) is a Z85c30 ESCC t
brd: module loaded
Warning: no ADB interface detected
Uniform Multi-Platform E-IDE driver
ide-gd driver 1.18
scsi host0: Macintosh NCR5380 SCSI, irq 19, io_port 0x0, base 0x50f10000, can_q
scsi 0:0:0:0: Direct-Access SEAGATE ST318416N 0010 PQ: 0 ANSI: 3
kernel BUG at block/deadline-iosched.c:332!
*** TRAP #7 *** FORMAT=0
Current process id is 1
BAD KERNEL TRAP: 00000000

Modules linked in:
PC: [<001882f2>] deadline_merge+0x0/0x8a
SR: 2000  SP: 02031ba0  a2: 0202c000
d0: 0000002b    d1: 00000059    d2: 00000000    d3: 00086c86
d4: 020cb8a8    d5: 00000001    a0: 00373d8e    a1: 00373d92
Process swapper (pid: 1, task=0202c000)
Frame format=0 
Stack from 02031bd4:
        0032302f 00337762 0000014c 0017392a 0213a200 0211d2c4 0211d0f4 00178d2c
        0213a200 0211d2e0 0037ba98 0018a144 0211d2c4 0213b014 020cba50 020cb800
        020cb8a8 001f0dc2 0211d2e0 00000000 00000003 00000002 001f0ccc 020cb8d0
        020cb8d8 001915c0 0002881e 020cba50 020cb8d8 001f0cc8 001f0ccc 020cba50
        001cc0f6 020cb8d0 020efcc0 020cb8f4 0037fe28 0018a144 020cb8d8 00000000
        020cb800 02031cd2 0213b300 001ee63c 020cb8f4 020ce120 ffffffff ffffffff
Call Trace: [<0017392a>] elevator_exit+0x24/0x40
 [<00178d2c>] blk_release_queue+0x44/0xce
 [<0018a144>] kobject_release+0x3a/0x7c
 [<001f0dc2>] scsi_device_dev_release_usercontext+0xf6/0x134
 [<001f0ccc>] scsi_device_dev_release_usercontext+0x0/0x134
 [<001915c0>] memset+0x0/0x8c
 [<0002881e>] execute_in_process_context+0x52/0x5a
 [<001f0cc8>] scsi_device_dev_release+0x16/0x1a
 [<001f0ccc>] scsi_device_dev_release_usercontext+0x0/0x134
 [<001cc0f6>] device_release+0x26/0xa6
 [<0018a144>] kobject_release+0x3a/0x7c
 [<001ee63c>] scsi_probe_and_probe_and_add_lun+0x0/0xb92
 [<001e5608>] scsi_is_host_device+0x0/0x14
 [<001cbf80>] put_device+0x0/0x14
 [<001d32ee>] transx4c4
 [<002c0920>] mutex_lock+0x0/0x22
 [<001d119e>] platform_drv_probe+0x16/0x44
 [<001cff9c>] driver_probe_device+0x17e/0x2
#12 lore raw
Quoted 2 lines · Finn Thain

Has anyone else tried the m68k cross-compiler from kernel.org and observed strange behaviour from the binaries it produces?

No, not really. Is there any particular reason why you are using these old compilers? In Debian, we have up-to-date versions of the gcc cross-toolchain for all target architectures we have in Debian, so I think there is little incentive to use these old compilers.

Adrian

author
#13 lore raw
Quoted 7 lines · John Paul Adrian Glaubitz
Quoted 2 lines · Finn Thain

Has anyone else tried the m68k cross-compiler from kernel.org and observed strange behaviour from the binaries it produces?

No, not really. Is there any particular reason why you are using these old compilers?

Mature, proven tools are important to upstream developers. Developers of distros needn't have the same expectations, since they take it upon themselves to put together reliable tools.

Please forgive my cross-posting.

Quoted 2 lines

In Debian, we have up-to-date versions of the gcc cross-toolchain for all target architectures we have in Debian,

Some years ago, for Debian users, I added links to those tools here:
http://www.mac.linux-m68k.org/docs/cross-dev.php

But I suspect that this information is no longer current. It may give the impression that Debian m68k cross-compilers are unavailable. Do you have any better sources of information that I could link to?

Quoted 1 line

so I think there is little incentive to use these old compilers.

It is useful for upstream developers to have distro-neutral tools. Ideally, we could use the kernel.org compiler as a "reference compiler". For automated builds, it seems to be that already. But no-one boots those binaries AFAIK.

Quoted 4 lines

Adrian

#14 lore raw
Quoted 6 lines · Finn Thain
Quoted 2 lines

No, not really. Is there any particular reason why you are using these old compilers?

Mature, proven tools are important to upstream developers. Developers of distros needn't have the same expectations, since they take it upon themselves to put together reliable tools.

To be honest, my personal experience is different. The toolchain has often bugs and lots of bugs are constantly being fixed upstream which is why Debian is always following the respective gcc SVN branches instead of using release versions.

Thus, gcc-6 in Debian isn't based on a release version but rather on the current gcc-6 SVN branch with all recent fixes and updates.

Quoted 9 lines
Quoted 2 lines

In Debian, we have up-to-date versions of the gcc cross-toolchain for all target architectures we have in Debian,

Some years ago, for Debian users, I added links to those tools here:
http://www.mac.linux-m68k.org/docs/cross-dev.php

But I suspect that this information is no longer current. It may give the impression that Debian m68k cross-compilers are unavailable. Do you have any better sources of information that I could link to?

If you're on Debian Stretch or newer, you can just install everything directly from the regular package archives, very convenient:

$ apt install gcc-m68k-linux-gnu g++-m68k-linux-gnu binutils-m68k-linux-gnu
Quoted 6 lines
Quoted 1 line

so I think there is little incentive to use these old compilers.

It is useful for upstream developers to have distro-neutral tools. Ideally, we could use the kernel.org compiler as a "reference compiler". For automated builds, it seems to be that already. But no-one boots those binaries AFAIK.

The problem with the toolchain on kernel.org seems that it's not updated very often if at all. I mean, gcc-4.6.3 is ancient, that was released over four years ago [1]. gcc has seen lots of improvements and fixes, particularly the SH has had so many bugs fixed that I'd highly discourage people using any of the older versions.

Adrian

Quoted 1 line

[1] https://gcc.gnu.org/gcc-4.6/

Signature · 4 lines
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - [email protected]
`. `'   Freie Universitaet Berlin - [email protected]
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913
author
#15 lore raw
Quoted 8 lines · John Paul Adrian Glaubitz
Quoted 6 lines · Finn Thain
Quoted 2 lines

In Debian, we have up-to-date versions of the gcc cross-toolchain for all target architectures we have in Debian,

Some years ago, for Debian users, I added links to those tools here:
http://www.mac.linux-m68k.org/docs/cross-dev.php

Quoted 8 lines
Quoted 3 lines

But I suspect that this information is no longer current. It may give the impression that Debian m68k cross-compilers are unavailable. Do you have any better sources of information that I could link to?

If you're on Debian Stretch or newer, you can just install everything directly from the regular package archives, very convenient:

$ apt install gcc-m68k-linux-gnu g++-m68k-linux-gnu binutils-m68k-linux-gnu

Nice. More widely useful might be a docker image or some other portable app container offering these packages. That's not a criticism of Debian, just a reflection of upstream development practices.

Quoted 10 lines
Quoted 7 lines
Quoted 2 lines

so I think there is little incentive to use these old compilers.

It is useful for upstream developers to have distro-neutral tools. Ideally, we could use the kernel.org compiler as a "reference compiler". For automated builds, it seems to be that already. But no-one boots those binaries AFAIK.

The problem with the toolchain on kernel.org seems that it's not updated very often if at all.

Updating gcc in this case may or may not fix the crash. I fixed it by downgrading gcc, but even that result can't be taken to imply a compiler "bug", just a change in behaviour in gcc-4.6.3 which may even benefit someone somewhere.

Updating needlessly makes fault-finding more difficult, as witnessed by the effort required to isolate the "kernel BUG" regression earlier in this thread.

Dependable tools are more important to me than new tools. YMMV.

Quoted 10 lines

I mean, gcc-4.6.3 is ancient, that was released over four years ago [1]. gcc has seen lots of improvements and fixes, particularly the SH has had so many bugs fixed that I'd highly discourage people using any of the older versions.

Adrian

Quoted 1 line

[1] https://gcc.gnu.org/gcc-4.6/

#16 lore raw
In reply to Finn Thain

Hi Finn,

Quoted 6 lines
Quoted 1 line

so I think there is little incentive to use these old compilers.

It is useful for upstream developers to have distro-neutral tools. Ideally, we could use the kernel.org compiler as a "reference compiler". For automated builds, it seems to be that already. But no-one boots those binaries AFAIK.

Not for Mac, but Geert usually boots kernels using ARAnyM. Pretty basic, for sure, but it's been good enough so have something to work from.

Having something like ARAnyM for Mac or Amiga would increase coverage, but the varied and quirky Mac hardware out there makes that almost pointless again.

Cheers,

Michael

author
#17 lore raw
Quoted 2 lines · Michael Schmitz

Not for Mac, but Geert usually boots kernels using ARAnyM. Pretty basic, for sure, but it's been good enough so have something to work from.

I thought Geert still used gcc-4.1 (?)

Quoted 4 lines

Having something like ARAnyM for Mac or Amiga would increase coverage, but the varied and quirky Mac hardware out there makes that almost pointless again.

AFAIK, MAME/MESS is the only thing that would be accurate enough.

For the Mac models that I was interested in (LC III for Egret and LC 520 for Cuda), I was able to patch some of the known bugs in MAME/MESS to get the ROM to boot to a flashing question mark, but I didn't try to boot any further. I will try loading EMILE.

Quoted 6 lines

Cheers,

Michael

#18 lore raw
Quoted 5 lines · Finn Thain
Quoted 2 lines · Michael Schmitz

Not for Mac, but Geert usually boots kernels using ARAnyM. Pretty basic, for sure, but it's been good enough so have something to work from.

I thought Geert still used gcc-4.1 (?)

Indeed.

Gcc-4.1 has issues with some bitfields, e.g. to compile gfs2 I need workarounds like:

-       struct gfs2_rbm rbm = { .rgd = rgd, .bii = 0, .offset = 0 };
+       // FIXME Avoid gcc 4.1.2 20061115 (prerelease) (Ubuntu 4.1.1-21) ICE
+       // struct gfs2_rbm rbm = { .rgd = rgd, .bii = 0, .offset = 0 };
+       struct gfs2_rbm rbm = { .bii = 0, .offset = 0 };
+       rbm.rgd = rgd;

I also have a 4.9.0 from kernel.org, if needed.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author
#19 lore raw

Re: Mac emulation, was Re: gcc-4.6.3

In reply to Finn Thain
Quoted 7 lines · On Wed, 9 Nov 2016, I wrote:
Quoted 4 lines · Michael Schmitz

Having something like ARAnyM for Mac or Amiga would increase coverage, but the varied and quirky Mac hardware out there makes that almost pointless again.

Quoted 7 lines

AFAIK, MAME/MESS is the only thing that would be accurate enough.

For the Mac models that I was interested in (LC III for Egret and LC 520 for Cuda), I was able to patch some of the known bugs in MAME/MESS to get the ROM to boot to a flashing question mark, but I didn't try to boot any further. I will try loading EMILE.

Unfortunately EMILE did not boot the emulated Mac IIx or Mac IIci in MAME/MESS. This resulted in a sad mac (0000000F 00000002):

$ pramutil -f ~/.sdlmess/nvram/maciix/rtc --boot-slot e --boot-slotrsrc 254
$ sdlmess -ramsize 8M maciix -nbe image -disk boot-2.6.18-mac_0.12.bin
$ pramutil -f ~/.sdlmess/nvram/maciici/rtc --boot-slot e --boot-slotrsrc 254
$ sdlmess -ramsize 16M maciici -nbe image -disk boot-2.6.18-mac_0.12.bin

Both of these emulated machines were able to boot to the desktop using a System 7 disk image.

I didn't test EMILE on the emulated LC III as I can't make that machine load any media at all, so I can't get past the flashing question mark.

#20 lore raw
In reply to Michael Schmitz
Quoted 3 lines · Michael Schmitz

Having something like ARAnyM for Mac or Amiga would increase coverage, but the varied and quirky Mac hardware out there makes that almost pointless again.

Current versions of WinUAE or fs-uae (available in Debian) should actually be able to boot Linux now as they include MMU support now. Haven't tried it yet though.

Adrian

#22 lore raw
Quoted 7 lines · Kaj-Michael Lang
Quoted 3 lines · Michael Schmitz

Having something like ARAnyM for Mac or Amiga would increase coverage, but the varied and quirky Mac hardware out there makes that almost pointless again.

Has anyone tried booting Linux/sun3 on tme,
http://people.csail.mit.edu/fredette/tme/ ?

No, not yet. But that would indeed be worth a shot.

Adrian

Signature · 4 lines
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - [email protected]
`. `'   Freie Universitaet Berlin - [email protected]
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913
author
#23 lore raw

Re: sun3 emulation, was Re: gcc-4.6.3

In reply to Kaj-Michael Lang
Quoted 8 lines · Kaj-Michael Lang
Quoted 3 lines · Michael Schmitz

Having something like ARAnyM for Mac or Amiga would increase coverage, but the varied and quirky Mac hardware out there makes that almost pointless again.

Has anyone tried booting Linux/sun3 on tme,
http://people.csail.mit.edu/fredette/tme/ ?

Good question (I've Cc'd Sam).

#24 lore raw

Re: sun3 emulation, was Re: gcc-4.6.3

Quoted 8 lines · Finn Thain
Quoted 5 lines · Kaj-Michael Lang
Quoted 3 lines · Michael Schmitz

Having something like ARAnyM for Mac or Amiga would increase coverage, but the varied and quirky Mac hardware out there makes that almost pointless again.

Quoted 5 lines
Quoted 3 lines

Has anyone tried booting Linux/sun3 on tme,
http://people.csail.mit.edu/fredette/tme/ ?

Good question (I've Cc'd Sam).

I've thought about trying it a few times, never actually done it. Though now that I've remembered it exists, I'm kinda tempted again.

IIRC physical 3/150 hardware worked at one point in time (at least for the specific set of boards in my 3/150).

-- Sam

#25 lore raw
In reply to Finn Thain

Hi Finn,

Quoted 11 lines · Finn Thain

Has anyone else tried the m68k cross-compiler from kernel.org and observed strange behaviour from the binaries it produces?

I used that cross-compiler (the i686 one) to build the kernel seen booting below. The resulting binary consistently crashes two different systems (LC III and LC 475), both during kernel startup. These two machines use different drivers and have different failure modes, but the failures are completely repeatable. The failures went away when I reverted to my usual (patched) gcc-4.4.6.

Quoted 6 lines

Perhaps gcc-4.6.3 is more strict than gcc-4.4, in such a way that would affect code generation, when presented with incorrect or undefined code? Are there any warnings that could be enabled to discover that code?

Is there some way to run the GCC test suite against this cross-compiler?

I also observed issues with some kernel.org toolchains, at least for coldfire the 4.9.x was producing some wrong opcode, due to incorrect gcc libs used over linking.

Maybe you can try some toolchains i prepared, they are the kernel.org with some packaging adjustments done by me. Let me know in case if they work for you.

http://sysam.it/toolchains.html

Regards,
Angelo Durgehello

author
#26 lore raw
Quoted 11 lines · Angelo Dureghello

I also observed issues with some kernel.org toolchains, at least for coldfire the 4.9.x was producing some wrong opcode, due to incorrect gcc libs used over linking.

Maybe you can try some toolchains i prepared, they are the kernel.org with some packaging adjustments done by me. Let me know in case if they work for you.

http://sysam.it/toolchains.html

Thanks for that.

The only problem is that my fast Linux host is i686 and your compiler runs on x86_64. But it is a convenient way to get hold of a gcc-5.2 compiler to target m68k/coldfire.

I may have to build a recent cross-compiler to run on my i686 host if and when my gcc-4.4 build lets me down, or I get time to chase down bugs like the issue I have with the kernel.org compiler.

Quoted 4 lines

Regards,
Angelo Durgehello

#27 lore raw
Quoted 3 lines · Finn Thain

The only problem is that my fast Linux host is i686 and your compiler runs on x86_64. But it is a convenient way to get hold of a gcc-5.2 compiler to target m68k/coldfire.

Installing Debian testing and running

apt install gcc-5-m68k-linux-gnu

is not an option?

Adrian

Signature · 4 lines
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - [email protected]
`. `'   Freie Universitaet Berlin - [email protected]
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913
author
#28 lore raw
Quoted 10 lines · John Paul Adrian Glaubitz
Quoted 3 lines · Finn Thain

The only problem is that my fast Linux host is i686 and your compiler runs on x86_64. But it is a convenient way to get hold of a gcc-5.2 compiler to target m68k/coldfire.

Installing Debian testing and running

apt install gcc-5-m68k-linux-gnu

is not an option?

As a way to get a compiler, it's more difficult than untarring the one from kernel.org, which was the effort I've been willing to invest so far.

I would consider installing the Debian Testing compiler in the form of a container, if one was available.

Or I could just update my cross-compiler build scripts again, but that solution doesn't get the upstream project much closer to the ideal solution, i.e. a convenient, reliable "reference" compiler.

That would require supported compiler and distro releases. Certainly Debian's gcc-5-m68k-linux-gnu or gcc-6-m68k-linux-gnu packages are good candidates for building a container, and time permitting, I would willingly test them and isolate and report any bugs I found.

Quoted 4 lines

Adrian

#29 lore raw
Quoted 2 lines · Finn Thain

As a way to get a compiler, it's more difficult than untarring the one from kernel.org, which was the effort I've been willing to invest so far.

If you're not on a Debian-based distribution, that is :).

Quoted 2 lines

I would consider installing the Debian Testing compiler in the form of a container, if one was available.

You can just install Debian into a systemd-nspawn chroot with the help of debootstrap. No need to use Docker or anything else. I have used this method to install and boot Fedora Rawhide on my Debian unstable system. Pretty nifty and simple:

# debootstrap --no-check-gpg --arch=amd64 --variant=buildd unstable /srv/unstable/ ftp://ftp.debian.org/debian
# systemd-nspawn -bD /srv/unstable

That's all you need to boot a Debian container on any machine running systemd.

Quoted 3 lines

Or I could just update my cross-compiler build scripts again, but that solution doesn't get the upstream project much closer to the ideal solution, i.e. a convenient, reliable "reference" compiler.

Does that really make such a difference? I have always used the gcc versions shipped in Debian and never had any issues. Matthias Klose also updates them very regularly, so I don't have to worry about that.

Quoted 4 lines

That would require supported compiler and distro releases. Certainly Debian's gcc-5-m68k-linux-gnu or gcc-6-m68k-linux-gnu packages are good candidates for building a container, and time permitting, I would willingly test them and isolate and report any bugs I found.

We're thoroughly testing the toolchain in Debian on the buildds. By actively building packages, we have found many bugs in gcc and binutils, most of them were in the SH and SPARC targets though. I don't remember running into any m68k-related toolchain issue.

Btw, there is one additional advantage using Debian for cross-platform
work: Multi-Arch. You can easily install build dependencies when cross-building for other targets. I have used this extensively to bootstrap ghc [1].

Adrian

Quoted 1 line

[1] https://wiki.debian.org/PortsDocs/BootstrappingGHC

Signature · 4 lines
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - [email protected]
`. `'   Freie Universitaet Berlin - [email protected]
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913