www.digitalmars.com         C & C++   DMDScript  

c++.chat - Memory model options in ctools Makefiles

reply Cesar Rabak <csrabak uol.com.br> writes:
I'm posting this in this NG because I cannot see any other where this
would have better fitš.

After reading Altaf's post in c++.dos.32.bits, I remembered that some
time ago I was hand holding a friend (by the phone, he lives at some 40
km from me), and after directing him to d/l the dmc package, to check
everything was OK, he asked for some sources to compile, after the
famous, but not very robust 'hello world' test.

I remember a that occasion not all targets compiled. Eventually my
friend and I gave out and now he is doing his work, reviving some old
Zortech code for his dissertation work.

So I went back to the ctools sources and found the following in
TOOLKIT.MAK:

CFLAGS= -DMSDOS -o+space -I..\include 

.c.obj:
	sc -c $(CFLAGS) $*
.
.
.
makedep.com : makedep.obj mem.obj list.obj filespec.obj filesear.obj
	sc -mti makedep.obj mem.obj list.obj filespec.obj filesear.obj \

chmod.com : chmod.c
       sc -mti -I..\include chmod.c

.
.
.
menu.exe : ..\include\menu.h menu.c
	sc -mi -I..\include -DTEST menu


Some interesting questions:

Is the 'i', as used in the mti option, an old memory specifier no longer
user in newer versions of the compiler? See, for example, the
makedep.com target.

Does the generic rule for objects w/o a memory model specifier presently
clash with the target's spec?

What was supposed to be the memory model for 'menu.exe'? The 'i' is not
recognized nor listed as a memory model.

I understand these makefiles are 12 to 10 years old, and the ctools
package is not 'actively' supported, but I think we can bring these
makefiles straight easily!

Regards

-- 
Cesar Rabak
GNU/Linux User 52247.
Get counted: http://counter.li.org/

[1] Perhaps if we arrive at a actionable bottom line, we can find other
to cross post ;-)
Feb 18 2002
parent reply "Walter" <walter digitalmars.com> writes:
The default memory model is now -mn, which is the win32 memory model. It is
no longer -ms. If you need to generate 16 bit executables, you now need to
specify a 16 bit memory model.

"Cesar Rabak" <csrabak uol.com.br> wrote in message
news:3C711968.7770AB9B uol.com.br...

 I'm posting this in this NG because I cannot see any other where this
 would have better fitš.

 After reading Altaf's post in c++.dos.32.bits, I remembered that some
 time ago I was hand holding a friend (by the phone, he lives at some 40
 km from me), and after directing him to d/l the dmc package, to check
 everything was OK, he asked for some sources to compile, after the
 famous, but not very robust 'hello world' test.

 I remember a that occasion not all targets compiled. Eventually my
 friend and I gave out and now he is doing his work, reviving some old
 Zortech code for his dissertation work.

 So I went back to the ctools sources and found the following in
 TOOLKIT.MAK:

 CFLAGS= -DMSDOS -o+space -I..\include

 .c.obj:
 sc -c $(CFLAGS) $*
 .
 .
 .
 makedep.com : makedep.obj mem.obj list.obj filespec.obj filesear.obj
 sc -mti makedep.obj mem.obj list.obj filespec.obj filesear.obj \

 chmod.com : chmod.c
        sc -mti -I..\include chmod.c

 .
 .
 .
 menu.exe : ..\include\menu.h menu.c
 sc -mi -I..\include -DTEST menu


 Some interesting questions:

 Is the 'i', as used in the mti option, an old memory specifier no longer
 user in newer versions of the compiler? See, for example, the
 makedep.com target.

 Does the generic rule for objects w/o a memory model specifier presently
 clash with the target's spec?

 What was supposed to be the memory model for 'menu.exe'? The 'i' is not
 recognized nor listed as a memory model.

 I understand these makefiles are 12 to 10 years old, and the ctools
 package is not 'actively' supported, but I think we can bring these
 makefiles straight easily!

 Regards

 --
 Cesar Rabak
 GNU/Linux User 52247.
 Get counted: http://counter.li.org/

 [1] Perhaps if we arrive at a actionable bottom line, we can find other
 to cross post ;-)
Feb 18 2002
parent reply Cesar Rabak <csrabak uol.com.br> writes:
Walter wrote:

 
 The default memory model is now -mn, which is the win32 memory model. It is
 no longer -ms. If you need to generate 16 bit executables, you now need to
 specify a 16 bit memory model.
 


OK. Thanks.

So this bring the first 'correction' or 'update' to present technology
would be to have the line

sc -c $(CFLAGS) $*

changed to

sc -c -ms $(CFLAGS) $*

in TOOLKIT.MAK.

Now the mystery remains for the possible original intent of the 'mi' and
'mti' memory models... I surmise that changing 'mti' to 'mt' is OK (it
seems clear to me that the original intent is the generation of .COM
programs.

So, what would be the correct way of fixing the '-mi' option?

TIA



 "Cesar Rabak" <csrabak uol.com.br> wrote in message
 news:3C711968.7770AB9B uol.com.br...

 I'm posting this in this NG because I cannot see any other where this
 would have better fitš.

 After reading Altaf's post in c++.dos.32.bits, I remembered that some
 time ago I was hand holding a friend (by the phone, he lives at some 40
 km from me), and after directing him to d/l the dmc package, to check
 everything was OK, he asked for some sources to compile, after the
 famous, but not very robust 'hello world' test.

 I remember a that occasion not all targets compiled. Eventually my
 friend and I gave out and now he is doing his work, reviving some old
 Zortech code for his dissertation work.

 So I went back to the ctools sources and found the following in
 TOOLKIT.MAK:

 CFLAGS= -DMSDOS -o+space -I..\include

 .c.obj:
 sc -c $(CFLAGS) $*
 .
 .
 .
 makedep.com : makedep.obj mem.obj list.obj filespec.obj filesear.obj
 sc -mti makedep.obj mem.obj list.obj filespec.obj filesear.obj \

 chmod.com : chmod.c
        sc -mti -I..\include chmod.c

 .
 .
 .
 menu.exe : ..\include\menu.h menu.c
 sc -mi -I..\include -DTEST menu


 Some interesting questions:

 Is the 'i', as used in the mti option, an old memory specifier no longer
 user in newer versions of the compiler? See, for example, the
 makedep.com target.

 Does the generic rule for objects w/o a memory model specifier presently
 clash with the target's spec?

 What was supposed to be the memory model for 'menu.exe'? The 'i' is not
 recognized nor listed as a memory model.

 I understand these makefiles are 12 to 10 years old, and the ctools
 package is not 'actively' supported, but I think we can bring these
 makefiles straight easily!

 Regards

 --
 Cesar Rabak
 GNU/Linux User 52247.
 Get counted: http://counter.li.org/

 [1] Perhaps if we arrive at a actionable bottom line, we can find other
 to cross post ;-)





-- 
Cesar Rabak
GNU/Linux User 52247.
Get counted: http://counter.li.org/
Feb 18 2002
parent "Walter" <walter digitalmars.com> writes:
I don't know what -mi was for. My suggestion is just remove it. -Walter

"Cesar Rabak" <csrabak uol.com.br> wrote in message
news:3C71B760.78887071 uol.com.br...

 Walter wrote:

 The default memory model is now -mn, which is the win32 memory model. It




is

 no longer -ms. If you need to generate 16 bit executables, you now need




to

 specify a 16 bit memory model.


 OK. Thanks.

 So this bring the first 'correction' or 'update' to present technology
 would be to have the line

 sc -c $(CFLAGS) $*

 changed to

 sc -c -ms $(CFLAGS) $*

 in TOOLKIT.MAK.

 Now the mystery remains for the possible original intent of the 'mi' and
 'mti' memory models... I surmise that changing 'mti' to 'mt' is OK (it
 seems clear to me that the original intent is the generation of .COM
 programs.

 So, what would be the correct way of fixing the '-mi' option?

 TIA



 "Cesar Rabak" <csrabak uol.com.br> wrote in message
 news:3C711968.7770AB9B uol.com.br...

 I'm posting this in this NG because I cannot see any other where this
 would have better fitš.

 After reading Altaf's post in c++.dos.32.bits, I remembered that some
 time ago I was hand holding a friend (by the phone, he lives at some






40

 km from me), and after directing him to d/l the dmc package, to check
 everything was OK, he asked for some sources to compile, after the
 famous, but not very robust 'hello world' test.

 I remember a that occasion not all targets compiled. Eventually my
 friend and I gave out and now he is doing his work, reviving some old
 Zortech code for his dissertation work.

 So I went back to the ctools sources and found the following in
 TOOLKIT.MAK:

 CFLAGS= -DMSDOS -o+space -I..\include

 .c.obj:
 sc -c $(CFLAGS) $*
 .
 .
 .
 makedep.com : makedep.obj mem.obj list.obj filespec.obj filesear.obj
 sc -mti makedep.obj mem.obj list.obj filespec.obj filesear.obj \

 chmod.com : chmod.c
        sc -mti -I..\include chmod.c

 .
 .
 .
 menu.exe : ..\include\menu.h menu.c
 sc -mi -I..\include -DTEST menu


 Some interesting questions:

 Is the 'i', as used in the mti option, an old memory specifier no






longer

 user in newer versions of the compiler? See, for example, the
 makedep.com target.

 Does the generic rule for objects w/o a memory model specifier






presently

 clash with the target's spec?

 What was supposed to be the memory model for 'menu.exe'? The 'i' is






not

 recognized nor listed as a memory model.

 I understand these makefiles are 12 to 10 years old, and the ctools
 package is not 'actively' supported, but I think we can bring these
 makefiles straight easily!

 Regards

 --
 Cesar Rabak
 GNU/Linux User 52247.
 Get counted: http://counter.li.org/

 [1] Perhaps if we arrive at a actionable bottom line, we can find






other

 to cross post ;-)





 --
 Cesar Rabak
 GNU/Linux User 52247.
 Get counted: http://counter.li.org/
Feb 18 2002