Adding data to proper SECTION make strange behaviour...

[fstarred]

I know it sounds silly, but.. I have a simple scrolling text with 3 bitplanes (standard, no interleaved).


Everything works OK until I add a module (MT_DATA) on the proper SECTION (see below)

SECTION    Data,DATA_C


FONT:
    incdir  "dh1:own/demo/repository/resources/fonts/"
    incbin  "32x32-FL.raw"


; added module

MT_DATA:
    incdir    "dh1:resources/mod/"    
    incbin    "mod.song"

When I launch the assembly command everything is ok, but when I launch the demo, the FONT is displayed with less bitplane..


I've attached two examples, one as it should be and one which is the screwed.

I'm going mad, anyone experienced something similar ???


I'm using AsmOne with FS-UAE

[ross]

If you upload somewhere the two executables i'll take a look.

[fstarred]

https://github.com/fstarred/amiga_pl.../demo/2/demo.s

Try to assemble both with binary under MT_DATA and without.


Then launch the demo on A500 with KS1.3 and see the results

[Don_Adan]

Originally Posted by fstarred

I know it sounds silly, but.. I have a simple scrolling text with 3 bitplanes (standard, no interleaved).


Everything works OK until I add a module (MT_DATA) on the proper SECTION (see below)

SECTION    Data,DATA_C


FONT:
    incdir  "dh1:own/demo/repository/resources/fonts/"
    incbin  "32x32-FL.raw"


; added module

MT_DATA:
    incdir    "dh1:resources/mod/"    
    incbin    "mod.song"

When I launch the assembly command everything is ok, but when I launch the demo, the FONT is displayed with less bitplane..


I've attached two examples, one as it should be and one which is the screwed.

I'm going mad, anyone experienced something similar ???


I'm using AsmOne with FS-UAE

Maybe your code used memory from MT_DATA area? You can place empty area "ds.b 100000" between both labels.

[majikeyric]

maybe you need your data to be word aligned somewhere in you section? (using 'even').

[Don_Adan]

Perhaps the easiest way to psedo fix is exchanging place for both incbins. First mod, second fonts.

[roondar]

I wonder if this is fixed if you change the Screen section to a non-BSS one.
If I recall correctly, BSS sections are not correctly supported by Amiga OS 1.3.

Though you would expect that lead to trouble with data on the screen buffer and not really with the FONT data.

[ross]

There is a bug in code

Hint: adding words is not always good

[fstarred]

Ross do you mean about EVEN keyword after TEXT? Uhm I'm very curious now!

[ross]

Originally Posted by roondar

If I recall correctly, BSS sections are not correctly supported by Amiga OS 1.3.

BSS sections are properly supported.

I've to check, but is the BSS part of DATA (and CODE) sections that are not zeroed.

[ross]

Originally Posted by fstarred

Ross do you mean about EVEN keyword after TEXT? Uhm I'm very curious now!

No.

Think about what differs between sum a word value on Dx and Ax registers.
However it's a pretty sneaky mistake and manifests itself, rightly, if you 'move' the data in memory.


(scusa, dovrei dirtelo subito, mi sento un po' ..)

[fstarred]

Gosh this is not my first time on this kind of error.. Unfortunately right now I can't try because I'm on on a PC however.. challenge accepted!

(Figurati anzi che ti sei preso il disturbo di darci un occhiata)

[fstarred]

Originally Posted by ross

No.

Think about what differs between sum a word value on Dx and Ax registers.
However it's a pretty sneaky mistake and manifests itself, rightly, if you 'move' the data in memory.


(scusa, dovrei dirtelo subito, mi sento un po' ..)

add.l #ScrBpl*h,d0

[ross]

Originally Posted by fstarred

add.l #ScrBpl*h,d0

Bingo!

[roondar]

Originally Posted by ross

BSS sections are properly supported.

I've to check, but is the BSS part of DATA (and CODE) sections that are not zeroed.

Well, I clearly misremembered then. I'm sure I read something about BSS sections and Kickstart 1.3 though. No idea what then, perhaps it's that they're not cleared?

Originally Posted by fstarred

add.l #ScrBpl*h,d0

Yeah, that's a sneaky one. Part of the reason I try to keep addresses in address registers

[meynaf]

Originally Posted by roondar

Well, I clearly misremembered then. I'm sure I read something about BSS sections and Kickstart 1.3 though. No idea what then, perhaps it's that they're not cleared?

It is "code_bss" and "data_bss", i.e. code or data sections with empty region at the end, that require V37+.

[Leffmann]

BSS space after code and data has been supported since 1.2 and will be uninitialized. On 2.0 and later it will be zero-filled.

[borchen]

I see you have two chipdata-sections:

...

SECTION    GRAPHIC,DATA_C
...
SECTION    Data,DATA_C
...

This ofcourse is not wrong, but I think it's safe to remove the second one.

[phx]

Doesn't hurt at all. As long as there are no PC-relative references between those sections (which you rarely have in data).

It even makes it more likely to be able to load the executable in a heavily fragmented low-memory situation.