From Newsgroup: comp.os.vms

Arne Vajh|+j <arne@vajhoej.dk> wrote:

On 7/21/2025 3:11 PM, Chris Townley wrote:

On 21/07/2025 19:22, Arne Vajh|+j wrote:

$ cc test
$ anal/obj/disasm/out=hello.lis hello

VSI is trying to get it in release notes for all compilers.

I thought it was...

It is in all the recent ones I have seen.

But I don't know if it is all and how far back it goes.

On Alpha:

ANALYZ A01-07

*** This file is not an ELF object file

I guess that by VSI policy Alpha is not recent enough. OTOH cc/list/machine_code works on Alpha...
--
Waldek Hebisch
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.os.vms

On 9/1/2025 3:59 PM, Waldek Hebisch wrote:

Arne Vajh|+j <arne@vajhoej.dk> wrote:

On 7/21/2025 3:11 PM, Chris Townley wrote:

On 21/07/2025 19:22, Arne Vajh|+j wrote:

$ cc test
$ anal/obj/disasm/out=hello.lis hello

VSI is trying to get it in release notes for all compilers.

I thought it was...

It is in all the recent ones I have seen.

But I don't know if it is all and how far back it goes.

On Alpha:

ANALYZ A01-07

*** This file is not an ELF object file

I guess that by VSI policy Alpha is not recent enough. OTOH cc/list/machine_code works on Alpha...

Itanium and x86-64 use ELF format.

VAX and Alpha use the VMS custom "double var" format. But
slightly different some definitions are xxx$ on VAX and
Exxx$ on Alpha.

Arne

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.os.vms

On 9/1/2025 3:59 PM, Waldek Hebisch wrote:

Arne Vajh|+j <arne@vajhoej.dk> wrote:

On 7/21/2025 3:11 PM, Chris Townley wrote:

On 21/07/2025 19:22, Arne Vajh|+j wrote:

$ cc test
$ anal/obj/disasm/out=hello.lis hello

VSI is trying to get it in release notes for all compilers.

I thought it was...

It is in all the recent ones I have seen.

But I don't know if it is all and how far back it goes.

On Alpha:

ANALYZ A01-07

*** This file is not an ELF object file

I guess that by VSI policy Alpha is not recent enough. OTOH cc/list/machine_code works on Alpha...

The Alpha/Itanium compilers support /LIST/MACHINE. [The Itanium C++ uses ANAL/OBJ/DISASSEMBLE behind the scenes]. For x86, hacking LLVM to do
BOTH an object file and a machine code listing is awkward. For x86
compilers only, you can also say /SWITCH=ASSEMBLY and get an .S file and
not an OBJ file. Most of the time the S file should work with llvm-mc,
but all of the time [which is why we don't document it - long story]

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.os.vms

On 9/2/2025 6:07 AM, hb0815 wrote:

On 9/2/25 00:52, Arne Vajh|+j wrote:

On 9/1/2025 3:59 PM, Waldek Hebisch wrote:

On Alpha:

ANALYZ A01-07

***-a This file is not an ELF object file

I guess that by VSI policy Alpha is not recent enough.-a OTOH
cc/list/machine_code works on Alpha...

Itanium and x86-64 use ELF format.

VAX and Alpha use the VMS custom "double var" format. But
slightly different some definitions are xxx$ on VAX and
Exxx$ on Alpha.

This indicates, that for (VAX and) Alpha nobody saw the need for a disassembler (integrated into ANALYZE).

I don't know if there are (freeware) tools that can disassemble such
object files. (A quick search didn't find any. I'm sure ChatGPT can
write one if you ask it!) It shouldn't be so hard (for VSI) to write
such a tool. The debugger and SDA can disassemble: it's just a matter of knowing the API and calling the corresponding code (which as far as I
can see is not exposed in a shareable image). But I doubt that anyone at
VSI wants to spend the time to create such a tool for Alpha.

If you really need to see the machine code that is in an object file,
for which you don't have the source code ...

DEC did not provide a disassembler.

But at least for VAX other did:

https://www.digiater.nl/openvms/freeware/v80/dism32/

worked pretty well as I remember it.

But Alpha is a different story.

Arne

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.os.vms

hb0815 <mw40171@mucweb.de> wrote:

On 9/2/25 00:52, Arne Vajh|+j wrote:

On 9/1/2025 3:59 PM, Waldek Hebisch wrote:

On Alpha:

ANALYZ A01-07

***-a This file is not an ELF object file

I guess that by VSI policy Alpha is not recent enough.-a OTOH
cc/list/machine_code works on Alpha...

Itanium and x86-64 use ELF format.

VAX and Alpha use the VMS custom "double var" format. But
slightly different some definitions are xxx$ on VAX and
Exxx$ on Alpha.

This indicates, that for (VAX and) Alpha nobody saw the need for a disassembler (integrated into ANALYZE).

I don't know if there are (freeware) tools that can disassemble such
object files. (A quick search didn't find any. I'm sure ChatGPT can
write one if you ask it!) It shouldn't be so hard (for VSI) to write
such a tool. The debugger and SDA can disassemble: it's just a matter of knowing the API and calling the corresponding code (which as far as I
can see is not exposed in a shareable image). But I doubt that anyone at
VSI wants to spend the time to create such a tool for Alpha.

objdump can disassemble, just use 'objdump -d' (or 'objdump -D' is you
expect code in nominally non-code sections). Works for me
with Alpha VMS .obj files (and executables). My main use is to check potentially confusing assembler mnemonics (which encoding is used
by the assembler).
--
Waldek Hebisch
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.os.vms

On 9/2/25 00:52, Arne Vajh|+j wrote:

On 9/1/2025 3:59 PM, Waldek Hebisch wrote:

On Alpha:

ANALYZ A01-07

***-a This file is not an ELF object file

I guess that by VSI policy Alpha is not recent enough.-a OTOH
cc/list/machine_code works on Alpha...

Itanium and x86-64 use ELF format.

VAX and Alpha use the VMS custom "double var" format. But
slightly different some definitions are xxx$ on VAX and
Exxx$ on Alpha.

This indicates, that for (VAX and) Alpha nobody saw the need for a disassembler (integrated into ANALYZE).

I don't know if there are (freeware) tools that can disassemble such
object files. (A quick search didn't find any. I'm sure ChatGPT can
write one if you ask it!) It shouldn't be so hard (for VSI) to write
such a tool. The debugger and SDA can disassemble: it's just a matter of knowing the API and calling the corresponding code (which as far as I
can see is not exposed in a shareable image). But I doubt that anyone at
VSI wants to spend the time to create such a tool for Alpha.

If you really need to see the machine code that is in an object file,
for which you don't have the source code ...

In an own program, you can read the code psect(s) of the object
module(s) into memory and start the debugger (remember the ZAP tool?) to disassemble it

There is also SRM_CHECK. It can disassemble code sections in Alpha
images. But that was not the main purpose of this tool. So I don't know
how good the disassembly output really is. However it's available as
freeware and the ZIP archive contains the source code.

And yes, ANALYZE/DISASSEMBLE (/OBJECT is default) is only for ELF
(Executable and Linkable Format) objects. There was obviously a need for
this and adding a disassembler to a tool that already knew the format of
the object module was the preferred solution.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.os.vms

On 2025-09-02, Arne Vajhoj <arne@vajhoej.dk> wrote:

On 9/2/2025 6:07 AM, hb0815 wrote:

This indicates, that for (VAX and) Alpha nobody saw the need for a
disassembler (integrated into ANALYZE).

I don't know if there are (freeware) tools that can disassemble such
object files. (A quick search didn't find any. I'm sure ChatGPT can
write one if you ask it!) It shouldn't be so hard (for VSI) to write
such a tool. The debugger and SDA can disassemble: it's just a matter of
knowing the API and calling the corresponding code (which as far as I
can see is not exposed in a shareable image). But I doubt that anyone at
VSI wants to spend the time to create such a tool for Alpha.

If you really need to see the machine code that is in an object file,
for which you don't have the source code ...

DEC did not provide a disassembler.

[snip]

But Alpha is a different story.

What about srm_check ? It apparently has an undocumented disassembler
built into it.

Simon.
--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.os.vms

On 9/4/25 14:22, Simon Clubley wrote:

What about srm_check ? It apparently has an undocumented disassembler
built into it.

As mentioned previously, srm_check only works with image files - for
obvious reasons.

If you really want to use it, the best you can do is to link your object module(s) with /NOREPLACE, exclude any default library and shareable
image, ignore all the warnings about undefined symbols and hope that the linker generates an image file.

Also, if you search for "srm_check," you may find complaints about the disassembly output not always being correct. I don't know if this was fixed. --- Synchronet 3.21a-Linux NewsLink 1.2