OpenEuphoria

2. Re: bug report - loop broken

• Posted by Hayden McKay <hmck1 at dodo.com.au> May 04, 2007
• 570 views
• Last edited May 05, 2007

oops sorry... i noticed that i broke

cache =  allocate_low(S * 512 + 16)
cache += 4-remainder(cache,16)        -- align(4)

replace with...

cache =  allocate_low(S * 512 + 4)
cache += 4-remainder(cache,4)         -- align(4)

Hayden McKay wrote:
> 
> I wrote this script while playing arround with bochs, if you scroll down
> to the main loop and uncomment the line
> }}}
<eucode>
> 
> --  dpmi[REG_DX] = NULL            -- broke inner loop
> 
> </eucode>
{{{

> you will see the inner loop gets broken
> 
> Another issue is that sometimes dos_interrupt() returns bogus results as
> in sighned words & unsighned dwords, the return sequence should be words
> eh? if you scroll down and comment out the line
> 
> }}}
<eucode>
> 
> dpmi = and_bits(dpmi,#FFFF) -- patch bogus result
> 
> </eucode>
{{{

> 
> loop is broken completly cause 'dos_interrupt()' return a sighned dword
> note that causeway v3.45 supports dword dos interrupts although the old
> causeway wrapped with eu v2.4 sometimes return theese bogus result too?
> 
> }}}
<eucode>
> include euphoria.e
> 
> sequence dpmi     -- dpmi structure
> 
>     dpmi = repeat(-1,REG_LIST_SIZE)
> 
>     -- check for the drive a: geometry
> 
>     dpmi[REG_ES] = NULL
>     dpmi[REG_DI] = NULL
>     dpmi[REG_DX] = NULL
>     dpmi[REG_AX] = 8 * 256
>     dpmi = dos_interrupt(#13,dpmi)
> 
>     dpmi = and_bits(dpmi,#FFFF) -- patch bogus result
> 
> constant fn = open("fdd.img","wb")
> 
> constant S = and_bits(dpmi[REG_CX],#3F)
> 
>     atom cache
> 
>     cache =  allocate_low(S * 512 + 16)
>     cache += 4-remainder(cache,16)        -- align(4)
> 
>     atom t t = time()
> 
>     for  C = NULL to floor(dpmi[REG_CX]/256) by 1 do
>     for  H = NULL to floor(dpmi[REG_DX]/256) by 1 do
> 
>     -- docs state that you should have at least three
>     -- io attempts if failed, but i assume the bios's
>     -- fdd handler will do that...?
> 
>     mem_set(cache,#4E,S*512)
> 
>     dpmi[REG_DX] = NULL
>     dpmi[REG_AX] = NULL
>     dpmi = dos_interrupt(#13,dpmi) -- recalibrate fdd
> 
>     dpmi[REG_ES] = floor   (cache/16)
>     dpmi[REG_BX] = and_bits(cache,15)
>     dpmi[REG_DX] = or_bits(H * 256,0) -- {head,drive}
>     dpmi[REG_CX] = or_bits(C * 256,1) -- {cyld,start}
>     dpmi[REG_AX] = or_bits(2 * 256,S) -- {func,count}
> 
>     dpmi = dos_interrupt(#13,dpmi) -- get fdd sectors
> 
> --  dpmi[REG_DX] = NULL            -- broke inner loop
> --  dpmi[REG_AX] = 1 * 256
> --  dpmi = dos_interrupt(#13,dpmi)
> 
> --  dpmi[REG_CX] = NULL    -- but not brake outer loop
> 
>     ? {C,H}
> 
>     puts(fn,peek({cache,S*512}))   -- flush the cache
> 
>     end for
>     end for
> 
>     t = time() - t
>     printf(1,"%d secs\n",floor(t))
> 
>     close(fn)
>     abort(NULL)
> </eucode>
{{{

> 
> <OT> i'm currently messing arround with a customized bootdisk ( kernel )
>      created in assembler, hence this is my bootdisk imager for da bochs
>      i'll probably end up useing the lba style sector addressing so then
>      i'll only need one loop variable.
> 
>      everyone is welcome to use this code...public domain...
> 
>      nb. latest causeway is v3.49 and there is a 4.02 patch file cw32.asm
>          that allows smid intructions in dos. although i doubt people use
>          smid in dos since mmx is usualy good enough for the old dos code