• Charles Banas (12/12) Mar 17 2003 i noticed in the DMD documentation that D can be used in systems
• Walter (6/13) Mar 19 2003 That's a feature of the linker. See www.digitalmars.com/ctg/optlink.html
• Charles Banas (9/16) Mar 19 2003 GCC's LD refers to it as "bin" format. i honestly don't want to have to...
• Walter (7/18) Mar 20 2003 go
• Charles Banas (12/23) Mar 20 2003 not so. it's a raw binary without headers. like a 32-bit com file (if
• Roberto Mariottini (9/21) Mar 21 2003 Also Microsoft's should support that, at least for backward
• Charles Banas (8/16) Mar 22 2003 i don't have the experience or skill to write a compiler backend myself....
• Sean L. Palmer (4/8) Mar 21 2003 It's because Microsoft wants to own all the standards.
• Charles Banas (5/13) Mar 22 2003 which is something i'd like to prevent. :)
• Walter (8/17) Mar 21 2003 OMF was the only one used on the PC by probably 50 compilers for about 1...
• Charles Banas (12/34) Mar 22 2003 the thing is, not only is it well-worn, but it's worn-out. those 50
• Mark Evans (4/7) Mar 22 2003 I second the motion. Object formats are a problem with Digital Mars too...

Charles Banas <greywolf greyfade.net> writes:

i noticed in the DMD documentation that D can be used in systems 
programming, but it is only mentioned in the introduction, and never 
expounded on.

for example, i can't find information on setting the start point (as i can 
with GNU GCC's ld) of an executable, outputting a raw binary at link time, 
or linking to oject files compiled with a separate assembler.

stuff like that.

am i just not looking hard enough, forgetting something, or has no one ever 
tried using DMD in a systems context?

if anyone has tried using DMD this way, i'd appreciate knowing about it. :)

-- 
Charles "grey wolf" Banas

"Walter" <walter digitalmars.com> writes:

"Charles Banas" <greywolf greyfade.net> wrote in message
news:oprl7p2ki78ctebf news.digitalmars.com...
 i noticed in the DMD documentation that D can be used in systems
 programming, but it is only mentioned in the introduction, and never
 expounded on.

 for example, i can't find information on setting the start point (as i can
 with GNU GCC's ld) of an executable,

That's a feature of the linker. See www.digitalmars.com/ctg/optlink.html

 outputting a raw binary at link time,

I don't know what you mean by this. The linker outputs an executable in PE
format.

 or linking to oject files compiled with a separate assembler.

Easy, just specify the .obj files in the command to DMD.

Charles Banas <greywolf greyfade.net> writes:

On Wed, 19 Mar 2003 01:21:06 -0800, Walter <walter digitalmars.com> wrote:

 That's a feature of the linker. See www.digitalmars.com/ctg/optlink.html

my bad.  i forgot DMD used that linker.

 outputting a raw binary at link time,

 I don't know what you mean by this. The linker outputs an executable in 
 PE
 format.

GCC's LD refers to it as "bin" format.  i honestly don't want to have to go 
the extra step of stripping a PE.

 or linking to oject files compiled with a separate assembler.

 Easy, just specify the .obj files in the command to DMD.

i'm just not exactly sure what format(s) it expects for the obj files.  
that, and i've never heard of OMF.  (or, more accurately, i've never seen 
anything that supports it.)

-- 
Charles "grey wolf" Banas

"Walter" <walter digitalmars.com> writes:

"Charles Banas" <greywolf greyfade.net> wrote in message
news:oprma8t3z58ctebf news.digitalmars.com...
 outputting a raw binary at link time,

 I don't know what you mean by this. The linker outputs an executable in
 PE
 format.

 GCC's LD refers to it as "bin" format.  i honestly don't want to have to

go
 the extra step of stripping a PE.

The LD "bin" format, if for windows32, is likely a PE (portable executable)
file.

 or linking to oject files compiled with a separate assembler.

 Easy, just specify the .obj files in the command to DMD.

 i'm just not exactly sure what format(s) it expects for the obj files.
 that, and i've never heard of OMF.  (or, more accurately, i've never seen
 anything that supports it.)

The object file format used by D and the linker is OMF (Object Module
Format). It's a standard put out by Intel.

Charles Banas <greywolf greyfade.net> writes:

On Thu, 20 Mar 2003 15:27:50 -0800, Walter <walter digitalmars.com> wrote:


 The LD "bin" format, if for windows32, is likely a PE (portable 
 executable)
 file.

not so.  it's a raw binary without headers.  like a 32-bit com file (if 
such a thing exists) without the extra 100-byte header.

 or linking to oject files compiled with a separate assembler.

 Easy, just specify the .obj files in the command to DMD.

 i'm just not exactly sure what format(s) it expects for the obj files.
 that, and i've never heard of OMF.  (or, more accurately, i've never 
 seen
 anything that supports it.)

 The object file format used by D and the linker is OMF (Object Module
 Format). It's a standard put out by Intel.

and the only other two linkers that seem to support it are Intel's and 
Borland's.  i don't have access to Intel's.  i never ceases to amaze me 
when some standards are never used by everyone.

since i work on multiple platforms, i need the ability to work with more 
common object file formats (like elf and coff) and to output a raw 32-bit 
binary file (with no header, relocation, or segment data) for my project.  
i guess i'll have to wait for the GCC frontend before i can use D.  suck.

-- 
Charles "grey wolf" Banas

Roberto Mariottini <rmariottini lycosmail.com> writes:

Charles Banas wrote:
 On Thu, 20 Mar 2003 15:27:50 -0800, Walter <walter digitalmars.com> wrote:

[...]
 The object file format used by D and the linker is OMF (Object Module
 Format). It's a standard put out by Intel.

 and the only other two linkers that seem to support it are Intel's and 
 Borland's.  i don't have access to Intel's.  i never ceases to amaze me 
 when some standards are never used by everyone.

Also Microsoft's should support that, at least for backward 
compatibility, since MASM can output also OMF object files.
So we have at least 3 multinational companies that uses this standard.

 since i work on multiple platforms, i need the ability to work with more 
 common object file formats (like elf and coff) and to output a raw 
 32-bit binary file (with no header, relocation, or segment data) for my 
 project.  i guess i'll have to wait for the GCC frontend before i can 
 use D.  suck.

I think you'll need a _backend_ not a _frontend_.
Have you tried the linux D compiler?
Or you can write your own backend, since the frontend sources are public.

Ciao

Charles Banas <greywolf greyfade.net> writes:

On Fri, 21 Mar 2003 09:13:10 +0100, Roberto Mariottini 
<rmariottini lycosmail.com> wrote:

 since i work on multiple platforms, i need the ability to work with more 
 common object file formats (like elf and coff) and to output a raw 32- 
 bit binary file (with no header, relocation, or segment data) for my 
 project.  i guess i'll have to wait for the GCC frontend before i can 
 use D.  suck.

 I think you'll need a _backend_ not a _frontend_.
 Have you tried the linux D compiler?
 Or you can write your own backend, since the frontend sources are public.

i don't have the experience or skill to write a compiler backend myself.  
the linux D compiler, as it is, is still too early in the alpha stages to 
be of any use to me.  i'd rather use something more mature like GCC or the 
existing win32 D compiler than this.  even if only for experimentation.

-- 
Charles "grey wolf" Banas

"Sean L. Palmer" <seanpalmer directvinternet.com> writes:

"Charles Banas" <greywolf greyfade.net> wrote in message
news:oprmdipc0b8ctebf news.digitalmars.com...
 On Thu, 20 Mar 2003 15:27:50 -0800, Walter <walter digitalmars.com> wrote:
 and the only other two linkers that seem to support it are Intel's and
 Borland's.  i don't have access to Intel's.  i never ceases to amaze me
 when some standards are never used by everyone.

It's because Microsoft wants to own all the standards.

Sean

Charles Banas <greywolf greyfade.net> writes:

On Fri, 21 Mar 2003 10:44:05 -0800, Sean L. Palmer 
<seanpalmer directvinternet.com> wrote:

 "Charles Banas" <greywolf greyfade.net> wrote in message
 news:oprmdipc0b8ctebf news.digitalmars.com...
 On Thu, 20 Mar 2003 15:27:50 -0800, Walter <walter digitalmars.com> 
 wrote:
 and the only other two linkers that seem to support it are Intel's and
 Borland's.  i don't have access to Intel's.  i never ceases to amaze me
 when some standards are never used by everyone.

 It's because Microsoft wants to own all the standards.

which is something i'd like to prevent. :)

-- 
Charles "grey wolf" Banas

"Walter" <walter digitalmars.com> writes:

"Charles Banas" <greywolf greyfade.net> wrote in message
news:oprmdipc0b8ctebf news.digitalmars.com...
 The object file format used by D and the linker is OMF (Object Module
 Format). It's a standard put out by Intel.

 and the only other two linkers that seem to support it are Intel's and
 Borland's.  i don't have access to Intel's.  i never ceases to amaze me
 when some standards are never used by everyone.

OMF was the only one used on the PC by probably 50 compilers for about 12
years. It's a pretty well worn standard. The only compilers I know of that
supports Microsoft's version of COFF is Microsoft's compilers.

 since i work on multiple platforms, i need the ability to work with more
 common object file formats (like elf and coff) and to output a raw 32-bit
 binary file (with no header, relocation, or segment data) for my project.
 i guess i'll have to wait for the GCC frontend before i can use D.  suck.

The PE format is pretty simple, and it shouldn't be too hard to write an
EXE2BIN for it, assuming that you've programmed out any references to DLLs,
etc.

Charles Banas <greywolf greyfade.net> writes:

On Fri, 21 Mar 2003 15:27:24 -0800, Walter <walter digitalmars.com> wrote:

 "Charles Banas" <greywolf greyfade.net> wrote in message
 news:oprmdipc0b8ctebf news.digitalmars.com...
 The object file format used by D and the linker is OMF (Object Module
 Format). It's a standard put out by Intel.

 and the only other two linkers that seem to support it are Intel's and
 Borland's.  i don't have access to Intel's.  i never ceases to amaze me
 when some standards are never used by everyone.

 OMF was the only one used on the PC by probably 50 compilers for about 12
 years. It's a pretty well worn standard. The only compilers I know of 
 that
 supports Microsoft's version of COFF is Microsoft's compilers.

the thing is, not only is it well-worn, but it's worn-out.  those 50 
compilers are obsolete and OMF was pretty much replaced by ELF, COFF, et. 
al.  and since i don't even use Microsoft's version of COFF (since i rarely 
use the MS compiler), your last statement held little meaning to me.

 since i work on multiple platforms, i need the ability to work with more
 common object file formats (like elf and coff) and to output a raw 32- 
 bit
 binary file (with no header, relocation, or segment data) for my 
 project.
 i guess i'll have to wait for the GCC frontend before i can use D.  
 suck.

 The PE format is pretty simple, and it shouldn't be too hard to write an
 EXE2BIN for it, assuming that you've programmed out any references to 
 DLLs,
 etc.

true, but i was hoping to avoid that.

although i would like to write everything in D, my projects simply aren't 
ready for that, so i want to write certain modules in D.  therefore, i want 
to use GCC for the majority of the project - and the object formats it 
supports.

-- 
Charles "grey wolf" Banas

Mark Evans <Mark_member pathlink.com> writes:

...use GCC for the majority of the project - and the object formats it 
supports.
Charles "grey wolf" Banas

I second the motion.  Object formats are a problem with Digital Mars tools
generally and the whole topic also folds into a Linux port of DMC++ which would
be an absolute Godsend for Linux.

Mark