[OT] lcc-win32 and GNU

[Kenny McCormack]

Presumably you are saying he should have grokked lcc-win32 to produce
assembler source code? If so, he still has the problem of writing in
ELF code generation into his assembler.

Probably the easiest fix for him is to require each user to install
libbfd separately; and in most Linuxes it comes preinstalled.[/QUOTE]

Which is what I've been saying all along.

 

[santosh]

someone is offering a compiler for Linux, complaining about the
license of libbfd which he used to write object files, but if he'd
made any effort to adapt to the local way of doing things (instead
of just dropping his big MS-style bomb into foreign territory) he
wouldn't have been writing object files at all and libbfd would not
have been involved!

Presumably you are saying he should have grokked lcc-win32 to produce
assembler source code? If so, he still has the problem of writing in
ELF code generation into his assembler.

There's no reason a serious competitor to the GNU compiler couldn't
feed its output to the GNU assembler.[/QUOTE]

I rather feel that this option would entail more work for jacob than
implementing object code generation directly.

(Not that I think a non-free x86-only compiler is a serious
competitor)

It could conceivably be a competitor on the x86. Right now though, the
main competitors to gcc on the x86 are MS's and Intel's compilers.

 

[Walter Roberson]

Since this apparently wooshed over everyone's head, the point was:
someone is offering a compiler for Linux, complaining about the license
of libbfd which he used to write object files, but if he'd made any
effort to adapt to the local way of doing things (instead of just
dropping his big MS-style bomb into foreign territory) he wouldn't have
been writing object files at all and libbfd would not have been
involved!

I use a true Unix system (Unix 95 certified) most of the time.
The vendor compilers do not produce assembly code unless specifically
requested. Instead, the "compile-only" pass first runs a front end
which produces a binary WHIRL file and then the language-independant
back-end part of the "compile-only" pass transforms that to object code.
It's been that way since at least 1996 (if not earlier.)
http://www-unix.mcs.anl.gov/OpenAD/open64A.pdf

Note that this is -not a "MS-style bomb into foreign territory";
this was used for highly optimizing Unix compilers.

So I don't know which "local way of doing things" you are talking
about.

 

[Alan Curry]

I use a true Unix system (Unix 95 certified) most of the time. [...]

So I don't know which "local way of doing things" you are talking
about.

The local way on GNU systems (see Subject line) which jacob was actually
talking about (which happens to follow old unix tradition). I didn't notice
any mention of "true UNIX(r)" until now.

 

[Walter Roberson]

I use a true Unix system (Unix 95 certified) most of the time. [...]

So I don't know which "local way of doing things" you are talking
about.

The local way on GNU systems (see Subject line) which jacob was actually
talking about (which happens to follow old unix tradition).

What are "GNU systems" ? I thought Jacob's target in this instance
was Linux; Linux is not GNU.

I didn't notice
any mention of "true UNIX(r)" until now.

You indicated that Jacob was introducing a MS 'bomb'; I demonstrated
that, on the contrary, real official Unix systems used the
direct-to-binary technique before MS.

You speak of "unix tradition", but not even -one- Linux version
has been Unix certified. How do the compilers on actual Unix systems
operate?
http://www.opengroup.org/openbrand/register/catalog.htm

 

[Richard]

I use a true Unix system (Unix 95 certified) most of the time. [...]

So I don't know which "local way of doing things" you are talking
about.

The local way on GNU systems (see Subject line) which jacob was actually
talking about (which happens to follow old unix tradition).

What are "GNU systems" ? I thought Jacob's target in this instance
was Linux; Linux is not GNU.

Huh? Linux "systems" are nearly always Gnu systems. Or should I say most
Gnu Operating Systems use the Linux kernel.

 

[Mark McIntyre]

So - what of the attackee?

You should have stopped here. The bit below felt uncomfortably like a
self-justification, and was unnecessary as we all know you are very
unlikely to stoop to sockpuppetry.
--
Mark McIntyre

"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it."
--Brian Kernighan

 

[CBFalconer]

santosh wrote:

.... snip all Navia baiting ...

Of course not. But as I'm sure you're aware, in these days where
distributing software on the internet is so cheap and easy, free
software developers make most of their money through support
contracts. Obviously there's a big problem with that business model

No, there are other methods. For example, the copyright holder can
release his code under GPL (or equivalent), making it free to all
who are willing to follow the license, and can still allow others
to use the same code under a different license. This can be useful
to enable code use without making the using source available to the
world, and MAY provide income to the copyright holder.

 

[Richard Heathfield]

Mark McIntyre said:

You should have stopped here. The bit below felt uncomfortably like a
self-justification, and was unnecessary as we all know you are very
unlikely to stoop to sockpuppetry.

Tell that to Richard Bos.

 

[CBFalconer]

Depends on how he defines most. Most compilers I have encountered
did *not* use the assembler. The first compiler that did use
assembler that I encountered was the C compiler on Unix vs. 6, but
that was about 10 years after playing with many other compilers.

In general, using an assembler can be a nuisance, because of
external dependencies, and greatly extended compiler processing
time. However use of a separate assembler is very handy when the
(basically) same compiler is to be used on various architectures.
It also leaves an immediately recognizable (and thus debuggable)
intermediate statement of the program.

 

[Tor Rustad]

Stop replying to yourself.

Nothing in R.H. c.l.c posting history I know about (from years back),
would suggest such a behavior.

Methinks, Richard Bos need to put forward an apology.

 

[CBFalconer]

Alan Curry wrote:
.... snip ...


I rather feel that this option would entail more work for jacob than
implementing object code generation directly.


It could conceivably be a competitor on the x86. Right now though,
the main competitors to gcc on the x86 are MS's and Intel's
compilers.

Note that Jacob was NOT writing a Linux compiler; he was writing an
X86 compiler usable only on X86's running Linux.

 

[jameskuyper]

Stop replying to yourself.

I'm curious. Would you care to identify the procedure you used to
identify rosewater as a sock puppet? Does it involve any evidence that
could survive even mildly skeptical inquiry? Or is it based entirely
upon percieved similarities between the points of view expressed?

 

[Ben Bacarisse]

Alan Curry wrote:

Presumably you are saying he should have grokked lcc-win32 to produce
assembler source code? If so, he still has the problem of writing in
ELF code generation into his assembler.

Probably the easiest fix for him is to require each user to install
libbfd separately; and in most Linuxes it comes preinstalled.

This is a misconception. lcc (the compiler driver) is linked again
libbfd, so it is a "derived work". (It is also linked again libc and
libm, but these are LGPL, I think.)

The GPL does not prevent anyone from distributing GPL source (or
binaries provide the source is made available). Heck, half the point
of the GPL is to _encourage_ distribution. What is prevents, is
building on the GPL work without making your code GPL too.

 

[Bart van Ingen Schenau]

[snip]

However use of a separate assembler is very handy when the
(basically) same compiler is to be used on various architectures.

I don't agree to that. Assembly language is tied too tightly to a
particular processor architecture to be useful as intermediate
language in a compiler that targets multiple architectures.
You need a slightly higher-level language, where things like processor
registers are not an important consideration. To my knowledge, this is
the kind of language that GCC uses between its (language dependent)
front-ends and its (platform dependent) back-ends.

It also leaves an immediately recognizable (and thus debuggable)
intermediate statement of the program.

True, but I would rather have that the code-generator can produce
either the binary op-codes for the compiler, or the equivalent
assembly mnemonics for assembly-output.

Bart v Ingen Schenau

 

[santosh]

This is a misconception. lcc (the compiler driver) is linked again
libbfd, so it is a "derived work". (It is also linked again libc and
libm, but these are LGPL, I think.)

Is this true even if the "linking" takes place at runtime with a shared
object of libbfd?

<snip>

 

[Alan Curry]

You speak of "unix tradition", but not even -one- Linux version
has been Unix certified. How do the compilers on actual Unix systems
operate?

Tradition is the stuff that goes back a long way, not the stuff that
currently has the UNIX(r) label on it. Those who are still using separate
processes for compiling (from C to assembly) and assembling (to create object
files) are in fact more traditional on this point. This has nothing to do
with your precious corporate certifications.

 

[Ben Bacarisse]

Is this true even if the "linking" takes place at runtime with a shared
object of libbfd?

I think so, but IANAL. It would seen strange to me if dynamic vs. static
linking altered the notion of what constitutes a derived work.

It may be time to ask in a group that knows.

 

[Chris Hills]

[On the few occasions in this reply where I have used the word "you", it
refers to Richard Bos. Where I mean a more generic "you", I have used the
appalling neologism "yougeneric".]

Richard Bos said:

(e-mail address removed) wrote:

Richard Heathfield wrote:

Stop replying to yourself.

I have not done so. I have no need to hide behind sock puppets. You have
made the mistake of believing one of the group trolls.

(rest of stupid denial - clipped)

Gee, a denial. Wow. Are we surprised?

From what I know of Richard Heathfield he can be an obtuse pedantic
pain in the arse (but not as bad as CBF but I believe him to be
honest and true to what he believes. He is consistent and stands by what
he says no mater how popular or unpopular.

I don't think it would even occur to him to use a fake id on here for
himself. If he wants to say something he does so under his own name.

 

[Chris Hills]

I'm curious. Would you care to identify the procedure you used to
identify rosewater as a sock puppet? Does it involve any evidence that
could survive even mildly skeptical inquiry? Or is it based entirely
upon percieved similarities between the points of view expressed?

Now it can be told..... Richard Heathfield is a sock puppet for Rose
Water

Rose not Richard is the real person.... Yes for years Rose has been
leading a double life..... the C guru and author we think we know and a
burger flipper at the all night Internet cafe in Detroit to make money
to feed 23 brothers and sisters.

The easy way to stop this silliness is for Rose to identify her/him
self. So Rose either publicly or two a couple of the regulars if you
would prefer not to tell everyone. Who will be able to vouch for the
fact that Rose and Richard are two people