Re: [hatari-devel] Bug in Hatari 2.0 |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/hatari-devel Archives
]
- To: hatari-devel@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [hatari-devel] Bug in Hatari 2.0
- From: Thomas Huth <th.huth@xxxxxxxxx>
- Date: Mon, 29 Jan 2018 07:37:22 +0100
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/simple; d=posteo.de; s=2017; t=1517207844; bh=28W4pwBX7Qpd4EqZjxpeFW8H7aeB35NYAp3WYJ6u2Co=; h=Subject:To:From:Date:From; b=MkFd4ft+54K8wiim8EG7CVRvQ8AHZOaRoB0LI4MHss16cFAFyXXa+VvWIvMTpO/PR 16kJhJmWvuf8c0sRPHiACb0YJGkHeq2SGKQp01Rtxu0vm5rBkGayvnGsxT+xiWeYDS +1HJPLrHGAG3xxisSamOsmIgTUIUcPCpEdf1nRIOCdqzccL3/jbg6bcqUJ6rXoci8f xDL+lATh07GgXR83D8ptQfE4p0GnK7dO3r22oVYATQxw/BfRIBIGewWeJtaSEjoVbv 1LnCfFboDgy9OjGfhQV7q4ssiykpx77xWMt2o7qQE80GQNVnXpSk++YdrfST78MZa1 OzRjn9PXrnUuQ==
On 28.01.2018 23:26, Eero Tamminen wrote:
> Hi,
>
> On 01/28/2018 08:11 PM, Nicolas Pomarède wrote:
>> Le 28/01/2018 à 17:23, Thomas Huth a écrit :
>>> Thanks! With that config file, I can reproduce the crash even with the
>>> latest version from the Mercurial repo. Seems like the problem is when
>>> you specify an FPU together with the GEMDOS hard disk. I was able to
>>> reproduce it with this minimal config file, too:
>>>
>>> [HardDisk]
>>> bUseHardDiskDirectory = TRUE
>>> szHardDiskDirectory = /tmp/hatari/gemdos.drv
>>>
>>> [ROM]
>>> szTosImageFileName = /tmp/hatari/etos512k.img
>>>
>>> [System]
>>> n_FPUType = 68882
>>>
>>> The "n_FPUType = 68882" seems to trigger a re-initialization of the CPU
>>> tables (even though it likely should not do that) as soon as you press
>>> the OK button in the GUI. I guess that somehow kills the special
>>> GEMDOS HD opcodes of Hatari, so that you finally run into a crash.
>>> Unfortunately, I currently don't have much time for further debugging
>>> (for the next 1.5 weeks) ... so either somebody else needs to have a
>>> closer look or this has to wait a little bit.
>>
>> when I try this with my config, it just prints :
>> FPU is not supported in 68000/010 configurations.
>
> After this message, our WinUAE CPU core code forces FPU model to zero.
>
> This looks like long standing bug in our WinUAE CPU core, which has
> been there at least since WinUAE 2.3 code import.
>
> One can install FPU to ST and I'm pretty sure Hatari has supported
> ST+FPU configurations earlier, at least with oldUAE CPU core. Thomas
No, FPU can not work in 68000 mode, since it needs Line-F. We don't (and
never did) support the MMIO FPUs in Hatari.
With the old UAE core, your only option was to select 68020+FPU, i.e.
there even was not a separate FPU configuration available.
>> Does this need a specific CPU or machine type to see the problem ?
>
> Yes, the one with which you got that warning (68000/010).
>
> As Roger stated, you need to press F12 & Enter, after which
> you will see just illegal instructions. Happens also with
> normal TOS.
>
> Attached patch makes that issue go away, but I didn't test
> whether FPU emulation actually works properly with ST emulation.
That's the wrong way. We should rather tell the user with a dialog that
FPU is not possible, and force the ConfigureParams to FPU_NONE.
Thomas