Re: [freedos-32-dev] malloc

[Hanzac C. <ha...@ho...>]

Hi, Salvo

>From: Salvo Isaja
>If I get it correctly, this is used to load it at the offset it is
>statically linked at. When properly loaded through the wrapper, that
>would be not necessary any longer (and IMHO, wrong, as you too said).
>Furthermore, I don't understand why the wrapper allocates memory with
>mem_get_regions by try-and-error incrementing the base value instead of
>using the regular mem_get: if this is for alignment reasons, I think
>the proper solution would be using a better memory manager.

Using mem_get is O.K. I once tested it. But doing this currently is just
for precaution, see Luca's Message, he once said,"
>>>
The idea is that I wanted to be able to chose where to load the
executable: when a DPMI program tries to allocate some memory (at least,
when a DJGPP program tries to do so), it wants to be able to allocate it
contiguously to the executable image...
For example:
- Let's assume that the program is loaded at 0x400000, and has size
0x10000
- Let's assume that the stack has size 0x1000
- So 0x400000 --> 0x410000: program image; 0x410000 --> 0x411000: stack
- When it allocates (for example) 0x2000 bytes of memory, it wants them
between 0x411000 and 0x413000.
If we use mem_get(), maybe there will not be enough memory for
allocation after the image end... This is why I used mem_get_region(),
so that we can decide where to allocate the program, to have a better
choice to find enough free space at its end...

But since I recently changed the way the PMM allocates memory, maybe
this problem is not so important anymore (previously, mem_get()
allocated memory at the end of the free space).
<<<

_________________________________________________________________
FREE pop-up blocking with the new MSN Toolbar - get it now! 
http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/