[SUG] free memory in W32 console to support batch operation

Index > Compiler Internals > [SUG] free memory in W32 console to support batch operation

Author

Thread

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20451
Location: In your JS exploiting you and your system

revolution 13 Apr 2010, 05:40

I have encountered a problem when running the console version. I keep running out of memory. Or at least it appears as though I have run out of memory, I actually have plenty of memory to compile with but I can't use it.

The reason is because the win32 console app does not free the memory when exiting. Of course Windows will free the memory eventually so there is no memory leak but Windows doesn't free the memory straight away.

I made a batch file to assemble all of my modules in with one command. It has this line: "for %%i in (*.asm) do fasmarm -m200000 %%~nxi" But the problem is every second file cannot be assembled due to lack of memory. The first file assembles fine and quits, the second file fails with a memory error, the third file assembles fine and quits, the fourth file fails with a memory error, an so on.

I fixed it with the following small addition to "SYSTEM.INC":

Code:

;...
exit_program:
      movzx   eax,al
      push    eax
 push    MEM_RELEASE             ;<----- add
      push    0                       ;<----- add
      push    [memory_start]          ;<----- add
      call    [VirtualFree]           ;<----- add
      call    [ExitProcess]
;...

And naturally I had to add VirtualFree to the import table in FASM.ASM

After adding this additional code to free the memory I can do batch assembly without any problem.

13 Apr 2010, 05:40

a115433

Joined: 05 Mar 2010
Posts: 144

a115433 13 Apr 2010, 06:33

not freeing memory before exit is a serious flaw.

13 Apr 2010, 06:33

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8359
Location: Kraków, Poland

Tomasz Grysztar 13 Apr 2010, 08:20

That's another DOS legacy in work - since there memory was naturally freed on exiting, the interface (which was the base for all later ports) didn't have anything like that. Though, perhaps, if I knew that Win32 console works this way, I would have done something with this earlier.

13 Apr 2010, 08:20

TmX

Joined: 02 Mar 2006
Posts: 843
Location: Jakarta, Indonesia

TmX 14 Apr 2010, 12:32

I applied this fix on FASM 1.69.13

Now every time I compile a program, FASM will crash.
But the program is compiled.

I'm on Windows 7, BTW.

Description:		Download
Filename:	fasm.revolution.fix.zip
Filesize:	4.13 KB
Downloaded:	569 Time(s)

14 Apr 2010, 12:32

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20451
Location: In your JS exploiting you and your system

revolution 14 Apr 2010, 13:14

TmX: You should use "push" not "call" -

Code:

  call [memory_start]  ;<----use PUSH

14 Apr 2010, 13:14

TmX

Joined: 02 Mar 2006
Posts: 843
Location: Jakarta, Indonesia

TmX 14 Apr 2010, 14:50

Ah silly me.
Now FASM runs fine.

Thanks Very Happy

14 Apr 2010, 14:50

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20451
Location: In your JS exploiting you and your system

revolution 27 Apr 2010, 01:59

This patch is still needed for fasm 1.69.14

I have fixed it the the fasmarm v1.16 release.

27 Apr 2010, 01:59

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8359
Location: Kraków, Poland

Tomasz Grysztar 27 Apr 2010, 11:57

Since this is part of interface, not core, they fix doesn't require core version change. Thus I'm just updating the 1.69.14 package.

27 Apr 2010, 11:57

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20451
Location: In your JS exploiting you and your system

revolution 27 Apr 2010, 12:03

Thanks for fixing. I don't mind making a few patches when each new version comes out, but when those patches get integrated into the official package it makes my life easier.

27 Apr 2010, 12:03

revolution
When all else fails, read the source

Joined: 24 Aug 2004
Posts: 20451
Location: In your JS exploiting you and your system

revolution 27 Apr 2010, 12:12

Tomasz: I think that is the first time you have ever included one of my patches without any modification. Cool

I was expecting something like this:

Code:

 movzx   ebx,al
      push    MEM_RELEASE 0 [memory_start]
        call    [VirtualFree]
       push    ebx
 call    [ExitProcess]

Because the 'push eax' above is rather hacky coding since it is not used until after the VirtualFree. I usually just makes these patches hastily, knowing you will want to "make them your own" anyway. So I don't go much further than making it just work and leave it there for the master to make the proper adjustments.

27 Apr 2010, 12:12

Tomasz Grysztar

Joined: 16 Jun 2003
Posts: 8359
Location: Kraków, Poland

Tomasz Grysztar 27 Apr 2010, 12:17

Well, this time I really liked it the way it was. Smile

27 Apr 2010, 12:17

< Last Thread | Next Thread >

Forum Rules:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum