Topicality

  • Thread starter Anders Wegge Keller
  • Start date
T

Tom St Denis

Le 24/02/11 14:58, Tom St Denis a crit :








Also, no offense, but your non-conforming compiler isn't all that
useful to most people.  I sell C code for a living.  Most customers
aren't even using the same processor let alone same compiler [or
version of the compiler] as I am.
No, the extensions are useless to you. Your situation isn't
particularly common.
I've never had a customer, ever, ask about using compiler extensions
in library code I was providing them.  Except to ensure I *wasn't*
using them.

Then, do not use compiler extensions.

The question is however:

Why do you need to denigrate my work at each post?

My compiler provides an ANSI compatible mode. Use that.

But of course that is not useful to you because some other obscure
reason that you will find in a second after you read this post.

Since when do I always denigrate your work? I rarely if ever even
mention your work at all.

Persecution complex much?

Tom
 
T

Tom St Denis

In message <[email protected]



Which one specifically?

Technically the latest is the "the spec" people should be aiming to
conform to. That said, C90 is basically what most people use.
Which GCC?  There are many.

Um any recent version from 3.x.y and up. I suppose if you're using a
version of GCC from 1987 before C90 was ratified it might not
conform...
Esoteric to you essential to others. So it is NOT a C99 compiler any
more than any other compiler is.  Therefore it does not conform to the
"ISO C spec"   It also has extensions of its own.

Except, that I can deal with a compiler not supporting VLAs (something
I'd never use). I can't make use of features [in a general sense]
that aren't part of the spec.

For instance, I do use GNU inline asm in some macros, but only if I
detect I'm building on those platforms. But things like VLAs [pre-
C99] I couldn't use since you can't really just #ifdef that away
cleanly.

Similarly, bignums or objects are not part of ANY C spec [let alone
the latest] and if you design software to use them you're pretty much
tied to that compiler.
CRAP   there are many GCC compilers.  They do not adhere to the ISO-c
standard any better than any other compiler out there and a lot worse
than most.  It is not even a particularly efficient compiler system

Um, I've never had a C90 support issue with any version of GCC from
3.4 on [I only mention that version since my experience with 2.8 and
3.2 is so old that I don't really recall specifics].

My work right now involves me mixing ARMcc and GCC and you know what,
GCC actually produces faster code in some circumstances [sometimes by
a long shot] when compiling portable C code [so it's not like I'm
cheating with inline asm]. ARMcc also produces faster code in some
circumstances. So it's hardly like GCC is out of contention in terms
of worthyness. In that it's free, portable, widely available, targets
many platforms, and actually produces fairly good output.

If anything, my big gripe about GCC [and GLIBC] is that its impossible
to build [specially as a cross-compiler]. Which is why companies like
Codesourcery have business...

Tom
 
C

Chris H

In message <[email protected]
s.com> said:
Technically the latest is the "the spec" people should be aiming to
conform to. That said, C90 is basically what most people use.

So C99 or C90 or something in between sort of.... In other words sort
of conforming to various standards over the last 20 years.....

Ie almost anything,
Um any recent version from 3.x.y and up. I suppose if you're using a
version of GCC from 1987 before C90 was ratified it might not
conform...

So a "recent" version or not and some don't conform...


so your version of the standard is a vague moving target and you are
only looking for partial conformance and some versions of GCC may or may
not conform to this partial sort of conformance....

Remind me what was your problem with LCC?
My work right now involves me mixing ARMcc and GCC and you know what,
GCC actually produces faster code in some circumstances

One swallow does not make a spring. BTW what is ARMcc
of worthyness. In that it's free,

That is a myth, you can just download it for free. Not the same thing
portable, widely available, targets
many platforms,

So are many other compilers.
and actually produces fairly good output.
Get serious!

If anything, my big gripe about GCC [and GLIBC] is that its impossible
to build [specially as a cross-compiler].

My time costs money.... so how "free" is this compiler? I know one
person who built GCC discovered that he had been timed doing it and when
they costed it out it had cost more than any of the relevant commercial
compilers in the running.

Unlike the GCC compiler they all came with 12 months support and a
library with a sensible license for commercial use.
 
K

Keith Thompson

jacob navia said:
My compiler provides an ANSI compatible mode. Use that.
[...]

I understand that lcc-win32 doesn't attempt to conform fully to C90
(for example, it always accepts // comments). How closely does the
"ANSI compatible mode" ("ISO compatible" would be more accurate)
conform to C99?
 
J

jacob navia

Le 24/02/11 16:18, Tom St Denis a écrit :
Since when do I always denigrate your work? I rarely if ever even
mention your work at all.

Persecution complex much?


Look:

I said:

That would be at last an interesting subject in this boring
group

YOU answered:

Also, no offense, but your non-conforming compiler isn't all that
useful to most people. I sell C code for a living. Most customers
aren't even using the same processor let alone same compiler [or
version of the compiler] as I am. Conforming to a standard is what
makes my job possible.


It was YOU that mentioned my compîler system in the first place.
Next time, just keep silent ok?
 
K

Kenny McCormack

jacob navia said:
Also, no offense, but your non-conforming compiler isn't all that
useful to most people. I sell C code for a living. Most customers
aren't even using the same processor let alone same compiler [or
version of the compiler] as I am. Conforming to a standard is what
makes my job possible.

It was YOU that mentioned my compîler system in the first place.
Next time, just keep silent ok?

Indeed.

It is a solid gimme and a dead certainty that whenever anyone says "no
offense", what follows is going to be both offensive and a put-down.

You are absolutely spot-on, as usual, about this group.

--
Windows 95 n. (Win-doze): A 32 bit extension to a 16 bit user interface for
an 8 bit operating system based on a 4 bit architecture from a 2 bit company
that can't stand 1 bit of competition.

Modern day upgrade --> Windows XP Professional x64: Windows is now a 64 bit
tweak of a 32 bit extension to a 16 bit user interface for an 8 bit
operating system based on a 4 bit architecture from a 2 bit company that
can't stand 1 bit of competition.
 
T

Tom St Denis

So C99 or C90 or something in between sort of....  In other words sort
of conforming to various standards over the last 20 years.....

Um, well I don't need to use every feature of C99 just because its
there. The slow adoption of new features in C99 [and C1X when it
comes out] are largely because of that. They're not always needed,
and in some industries not even applicable.

Like people writing C compilers for PICs might not have added "long
long" when they were C90 because a) it's not part of the spec and b)
not really applicable to writing code for a PIC. Whereas, many other
vendors added "long long" before C99 was ratified.

Roughly speaking if you code against C90 you're targeting basically
100% of all C environments [code/data memory limitations
notwithstanding].
Ie almost anything,

Um, no. It's fairly clear. ISO C90 is an exact specification. Using
"long long" [which is my only use outside C90] makes me a C99 user.
So if I was pushed I'd say "I develop C99 code." But practically
speaking I say "C90 with long long" as it better targets my market
[hint: many UNIX C90 compilers supported "long long"].
So a "recent" version or not and some don't conform...

Where did I say that? To use a version of GCC that was pre-C90 you'd
have to go back to v1.36. We're on v4.5.2 (after several major
compiler re-architectures!).

We actually tell customers that we use GCC 4.4 and up, and we make
SPECIAL circumstances for older compilers. Most customers have no
problems with this.
so your version of the standard is a vague moving target and you are
only looking for partial conformance and some versions of GCC may or may
not conform to this partial sort of conformance....

No, I would not use a version of GCC that doesn't at least support
C90. But since I use "long long" I need that as well [or C99
conformance].

There is no "vague moving target" about this.
Remind me what was your problem with LCC?

Which standard describes his add-on features? How am I not vendor
locked in if I use them? Does his compiler produce code for x86_64,
ppc, arm, mips, ?

[hint: GCC does]
One swallow does not make a spring.  BTW what is ARMcc

If you're purporting to have an opinion worth a damn and don't know
what ARMcc is (you can infer that it's a C compiler from ARM...) then
why again are you posting?
That is a myth, you can just download it for free. Not the same thing

GCC isn't free as in libre? I guess I missed something.
So are many other compilers.

Um, when LCC-win32 targets ARM and PPC [and PIC and AVR and ...] maybe
you'll have a point.

Or when Microsoft Visual C++ targets any of those ...
Get serious!

Well I have empirical data to back it up. You have nothing but random
thoughts weaving in and out of your mind.
If anything, my big gripe about GCC [and GLIBC] is that its impossible
to build [specially as a cross-compiler].

My time costs money.... so how "free" is this compiler?   I know one
person who built GCC discovered that he had been timed doing it and when
they costed it out it had cost more than any of the relevant commercial
compilers in the running.

You can buy commercial cross compilers based on GCC [hint: that's what
we do]. We also use the versions of GCC that come with our distros.
Unlike the GCC compiler they all came with 12 months support and a
library with a sensible license for commercial use.

Code Sourcery offers that with their builds of GCC. Your point?

The first week I had ARMcc [a really expensive but fairly good ARM
compiler] I broke it. It failed to build a valid C source. Then they
released a patch. I broke that one too. Don't get me wrong, ARM
support is fantastic. But lets not pretend like GCC is the only
compiler with faults.

I've also broken MSVC countless times ...

Tom
 
T

Tom St Denis

Le 24/02/11 16:18, Tom St Denis a crit :
Since when do I always denigrate your work?   I rarely if ever even
mention your work at all.
Persecution complex much?

Look:

I said:

That would be at last an interesting subject in this boring
group

YOU answered:

Also, no offense, but your non-conforming compiler isn't all that
useful to most people.  I sell C code for a living.  Most customers
aren't even using the same processor let alone same compiler [or
version of the compiler] as I am.  Conforming to a standard is what
makes my job possible.

It was YOU that mentioned my comp ler system in the first place.
Next time, just keep silent ok?

You bring up your non-conforming compiler. If you can't take the heat
get out of the kitchen. We're not all here to serve as your personal
pep rally.

You bring up LCC-win32 from time to time. In this thread you bitched
that we don't talk about new compiler ideas so I let you know what I
think about it.

Newsflash, not everyone is going to agree with what you're doing. I
never insulted your work though, I just said that non-conforming
compilers aren't that useful to most people. And it's true.

Tom
 
J

jacob navia

Le 24/02/11 17:25, Keith Thompson a écrit :
jacob navia said:
My compiler provides an ANSI compatible mode. Use that.
[...]

I understand that lcc-win32 doesn't attempt to conform fully to C90
(for example, it always accepts // comments).

And so what?
If you don't want // comments just do not use them.
If my compiler doesn't COMPLAIN about

// This is a single line comment

that doesn't mean that it is non conforming to C89.
It will accept and correctly compile all C89 code.

How closely does the
"ANSI compatible mode" ("ISO compatible" would be more accurate)
conform to C99?

I don't know. In any case there are no problems with C89.
 
J

jacob navia

Le 24/02/11 16:50, Chris H a écrit :
My time costs money.... so how "free" is this compiler? I know one
person who built GCC discovered that he had been timed doing it and when
they costed it out it had cost more than any of the relevant commercial
compilers in the running.

Unlike the GCC compiler they all came with 12 months support and a
library with a sensible license for commercial use.

My first (and last) contact with gcc support was with RedHat.
I wanted to know the format of the exception handling used by gcc
so that my JIT would be compatible with it under linux.

I called the developer at Red Hat. He told me that to answer
that question I had to subscribe to a one year license of full
support by RedHat for US$ 25 000.

It took me 4 weeks to figure it out. But I figured it out eventually.

What I was missing is that the exception handling is handled by a script
started by the linker, that compiles C source (during linking) that
implements the run time interpreter of the exception handling rewind
stack machine.

Imagine. The compiler generates byte codes that are interpreted by
an interpreter that is compiled in at link time...

Another "gem" was gnu's assembler that it sees

.byte 63
.byte 64

it will NOT output
63,64

but another number since it "OPTIMIZES" the assembler instructions
believing it is gcc that emitted them... That took me also a while
to find.

Happily under windows, doing the same was incredibly easy. Microsoft
had an API.
 
T

Tom St Denis

I don't know. In any case there are no problems with C89.

As a compiler vendor I'd assume your test harness includes conformance
testing for [at least] C99, let alone in addition to C90 and/or C89.

Tom
 
K

Keith Thompson

Chris H said:
My time costs money.... so how "free" is this compiler? I know one
person who built GCC discovered that he had been timed doing it and when
they costed it out it had cost more than any of the relevant commercial
compilers in the running.

I haven't built gcc from source in several years, since binary builds
are easily available for the systems I use.

If I did need to build it from source, I could fire off the build
and work on something else until it's done. (It tooks several
hours last time I did it, but computers are faster these days.)

If using gcc imposes more costs *for you* than the commercial
alternatives, then by all means use one of the commercial
alternatives.
 
T

Tom St Denis

If using gcc imposes more costs *for you* than the commercial
alternatives, then by all means use one of the commercial
alternatives.

As I pointed out a few times in this thread already, there are
commercial vendors of GCC for various architectures. So if the cost
were an issue you weigh it against buying [or just downloading since
it's GPL anyways] one of them.

Not that my position of advocacy is to only use GCC, but the notion
from Chris H that GCC is utterly worthless is not supported by the
evidence. It beats ARMcc in some applications, is more portable than
most other compilers, it's the default compiler of Freescale, etc...

Tom
 
K

Keith Thompson

jacob navia said:
Le 24/02/11 17:25, Keith Thompson a ecrit :
jacob navia said:
My compiler provides an ANSI compatible mode. Use that.
[...]

I understand that lcc-win32 doesn't attempt to conform fully to C90
(for example, it always accepts // comments).

And so what?
If you don't want // comments just do not use them.
If my compiler doesn't COMPLAIN about

// This is a single line comment

that doesn't mean that it is non conforming to C89.

Yes it does. The C89/C90 standard requires a diagnostic for any
violation of a syntax rule. This is clearly stated in C90 5.1.1.3
(it's probably 2.1.1.3 in the ANSI C89 standard).
It will accept and correctly compile all C89 code.

jacob, read the following sentence very carefully.

I WAS NOT COMPLAINING ABOUT LCC-WIN32'S LACK OF C90 CONFORMANCE.

You are under no obligation whatsoever to produce a conforming C90
compiler, any more than I am, and I personally don't care whether you do
or not. I merely made an observation.

I think there are some obscure corner cases where a C99 // comment
can appear to be some other valid construct in C90; my guess is
that lcc-win32 handles these in the C99 manner, and thus fails to
conform fully to C90. If you want to argue that these corner cases
aren't worth worrying about, I won't disagree, but it does mean that
your claim that "It will accept and correctly compile all C89 code"
isn't 100% correct.

And a conforming C90 compiler must issue diagnostics for any violations
of C90 constraints or syntax rules. I doubt that lcc-win32 does so.
(This is not to imply that it should.)

Finally, how does lcc-win32 handle the following strictly conforming C90
program:

int main(void)
{
int restrict = 1;
int inline = 1;
return 0;
}

One more time, it doesn't bother me in the slightest if lcc-win32 isn't
a fully conforming C90 compiler.

But I would find it very interesting if it were a fully conforming
C99 compiler (since there are so few of them), thus my question:
I don't know. In any case there are no problems with C89.

You don't know? Seriously?

You've been bragging about lcc-win32's C99 conformance here for years
(and I've congratulated you for it).

Ok, you're under no more obligation to produce a conforming
C99 compiler than you are to produce a conforming C90 compiler.
If you've given up on C99 conformance, I find that disappointing,
but since I am neither a paying customer nor a user of the free
version there's no real reason my disappointment should concern you.

Still, if you're making no effort to conform to C99, then I suggest
that comp.lang.c is not the best place to discuss your product.
(Yes, I know you weren't the first person to bring it up in this
thread, but you've certainly mentioned it enough in other threads.)
 
C

Chris H

In message <[email protected]
..com> said:
So C99 or C90 or something in between sort of....  In other words sort
of conforming to various standards over the last 20 years.....

Um, well I don't need to use every feature of C99 just because its
there. The slow adoption of new features in C99 [and C1X when it
comes out] are largely because of that. They're not always needed,
and in some industries not even applicable.

Like people writing C compilers for PICs might not have added "long
long" when they were C90 because a) it's not part of the spec and b)
not really applicable to writing code for a PIC. Whereas, many other
vendors added "long long" before C99 was ratified.

Roughly speaking if you code against C90 you're targeting basically
100% of all C environments [code/data memory limitations
notwithstanding].

So your idea of a standard is "roughly" and approximately" something
that suits you and is not standard at all. .
Ie almost anything,

Um, no. It's fairly clear. ISO C90 is an exact specification. Using
"long long" [which is my only use outside C90] makes me a C99 user.
So if I was pushed I'd say "I develop C99 code." But practically
speaking I say "C90 with long long" as it better targets my market
[hint: many UNIX C90 compilers supported "long long"].


EQUIVOCATION AGAIN. It is either standard of it is not.
Where did I say that? To use a version of GCC that was pre-C90 you'd
have to go back to v1.36. We're on v4.5.2 (after several major
compiler re-architectures!).
We actually tell customers that we use GCC 4.4 and up, and we make
SPECIAL circumstances for older compilers. Most customers have no
problems with this.

So GCC is variable too. (And non standard as well.... I HAVE seen test
results for GCC against the industry standard test suits.)
so your version of the standard is a vague moving target and you are
only looking for partial conformance and some versions of GCC may or may
not conform to this partial sort of conformance....

No, I would not use a version of GCC that doesn't at least support
C90. But since I use "long long" I need that as well [or C99
conformance].
There is no "vague moving target" about this.

Yes it is. You have an approximate idea of the standard and a compiler
that is not standards compliant. Having a flag that says C90 or C99 is
not the same as passing the test suits.
Which standard describes his add-on features?

None but then no ISO standard describes the many variations of the GCC
extensions.
How am I not vendor
locked in if I use them? Does his compiler produce code for x86_64,
ppc, arm, mips, ?
[hint: GCC does]

I don't care I don't use those targets.

How does GCC produce code for ARM. Cofdfire and 8051? IAR does (and
with a validated compiler) ....
If you're purporting to have an opinion worth a damn and don't know
what ARMcc is (you can infer that it's a C compiler from ARM...) then
why again are you posting?

OK it is a compiler for ARM. So what.
GCC isn't free as in libre? I guess I missed something.

You did.
So are many other compilers.

Um, when LCC-win32 targets ARM and PPC [and PIC and AVR and ...] maybe
you'll have a point.
Or when Microsoft Visual C++ targets any of those ...

Why should any of them?

So LCC''s fault is that it has the same targets as Microsoft. Why does
it need to target multiple targets?
Well I have empirical data to back it up. You have nothing but random
thoughts weaving in and out of your mind.

I doubt you do. I have seen plenty of benchmarks however GCC usually
fairs very badly without a lot of work.... and time cost money.
If anything, my big gripe about GCC [and GLIBC] is that its impossible
to build [specially as a cross-compiler].

My time costs money.... so how "free" is this compiler?   I know one
person who built GCC discovered that he had been timed doing it and when
they costed it out it had cost more than any of the relevant commercial
compilers in the running.

You can buy commercial cross compilers based on GCC [hint: that's what
we do]. We also use the versions of GCC that come with our distros.

Version of.... all of them non standard... different and have a very
restrictive library.
Unlike the GCC compiler they all came with 12 months support and a
library with a sensible license for commercial use.

Code Sourcery offers that with their builds of GCC. Your point?

The first week I had ARMcc [a really expensive but fairly good ARM
compiler] I broke it. It failed to build a valid C source.

Really... However your idea of valid source seems variable and flexible.
Then they
released a patch. I broke that one too. Don't get me wrong, ARM
support is fantastic. But lets not pretend like GCC is the only
compiler with faults.

It isn't but it has more than most .
 
C

Chris H

In message <[email protected]
s.com> said:
I don't know. In any case there are no problems with C89.

As a compiler vendor I'd assume your test harness includes conformance
testing for [at least] C99, let alone in addition to C90 and/or C89.


Why? Most compiler are C95 and NOT C99. You don't test a C95
compiler with a C99 test suit.
 
K

Keith Thompson

Chris H said:
In message <[email protected]
s.com> said:
How closely does the
"ANSI compatible mode" ("ISO compatible" would be more accurate)
conform to C99?

I don't know. In any case there are no problems with C89.

As a compiler vendor I'd assume your test harness includes conformance
testing for [at least] C99, let alone in addition to C90 and/or C89.

Why? Most compiler are C95 and NOT C99. You don't test a C95
compiler with a C99 test suit.

Because jacob has been bragging here for years about lcc-win32's C99
conformance.
 
T

Tom St Denis

Roughly speaking if you code against C90 you're targeting basically
100% of all C environments [code/data memory limitations
notwithstanding].

So your idea of a standard is "roughly" and approximately"  something
that suits you and is not standard at all. .

I don't get what sort of standard of response you are looking for. I
already said multiple times that I base most of my work on C90 with
one exception, but stated that if pushed I'd say I was a C99
developer. There is no "approximate" about that. If you want a
precise answer I'm a C99 developer. Practically speaking though [of
more use to real developers] I'm C90 with "long long."

Again, if you had real experience developing portable software this
would make sense to you.
Um, no.  It's fairly clear.  ISO C90 is an exact specification.  Using
"long long" [which is my only use outside C90] makes me a C99 user.
So if I was pushed I'd say "I develop C99 code."  But practically
speaking I say "C90 with long long" as it better targets my market
[hint: many UNIX C90 compilers supported "long long"].

EQUIVOCATION AGAIN.  It is either standard of it is not.

C99 is a standard. Now you're just being obtuse.
So GCC is variable too. (And non standard as well....  I HAVE seen test
results for GCC against the industry standard test suits.)

Last I checked GCC conforms to C90 and most of C99. For my purposes
GCC is a C99 compiler since I don't use the features it lacks in that
domain. More so, I claim I write C99 code because I make use of
features only found in C99 [and nowhere else]. I don't force the use
of features that are not part of C99.

LCC-win32 is not C99 compliant. Nor is ARMcc, nor is MSVC, nor is ...
Yes it is. You have an approximate idea of the standard and a compiler
that is not standards compliant. Having a flag that says C90 or C99 is
not the same as passing the test suits.

Um, if my C90 [or C99] compliant code compiles correctly it's good
enough for me. And I have never seen GCC fail to accept C90 code. I
have ICE'ed the compiler [optimization bugs] but syntactically it
ALWAYS has accepted my C90 code.

Also the GCC team has a test suite for conformance to C90 and C99.
Check it out.
None but then no ISO standard describes the many variations of the GCC
extensions.

And I don't force my users to use them. The difference between a bit
of inline-asm [using GNU's syntax] and using a whole design
methodology is vast. Take a look at my math library "TomsFastMath."
From the same C code I support portable C99, ARM, MIPS, PPC, X86, and
AVR32. In fact, new ports usually only take minutes to write (I wrote
the PPC port while ssh'ed into a host in the UK on a lark).

Now suppose I used LCC-win32s "objects". Could I write a program such
that I both use objects and not use objects to work with both his and
C99 compliant compilers? No. A whole program re-write would be
required. Therefore, if I spend time developing apps with his non-
standard additions I get vendor locked in.
 How am I not vendor
locked in if I use them?  Does his compiler produce code for x86_64,
ppc, arm, mips, ?
[hint: GCC does]

I don't care I don't use those targets.

And because you [and he] don't care about that it makes the compiler
less than ideal for a lot of people.
How does GCC produce code for  ARM. Cofdfire and 8051?  IAR does (and
with a validated compiler) ....

Um? GCC targets ARM. I wouldn't use a C compiler for an 8051, but
then again I've developed programs for an 8051 so I know WTF I'm
talking about ...
OK it is a compiler for ARM. So what.

It was my example of a really expensive commercial compiler.
Why should any of them?

Because most processors on this planet don't run x86 natively?
So LCC''s fault is that it has the same targets as Microsoft. Why does
it need to target multiple targets?

Because if you want to have proprietary language extensions you ought
to target more than one platform. Otherwise it's just left there.
Basically people who don't write win32 programs can't use his compiler
even if they really wanted to.
I doubt you do. I have seen plenty of benchmarks however GCC usually
fairs very badly without a lot of work.... and time cost money.

I don't know what to say here. I have benchmark after benchmark that
shows it's not uniformly better or worse. Believe it if you will. I
don't give a rats ass.
You can buy commercial cross compilers based on GCC [hint: that's what
we do].  We also use the versions of GCC that come with our distros.

Version of.... all of them non standard... different and have a very
restrictive library.

That's so retarded I don't even know ... words ... lost.

GLIBC is a LGPL such that your applications are not "restricted."

And they're not all different versions. We have both GCC 4.5.1 for
ARM and PPC. It's the same compiler just targets different
processors. All you are showing now is the complete and utter lack of
experience you have with the tools.
Code Sourcery offers that with their builds of GCC.  Your point?
The first week I had ARMcc [a really expensive but fairly good ARM
compiler] I broke it.  It failed to build a valid C source.

Really... However your idea of valid source seems variable and flexible.

Where did I say that?
It isn't but it has more than most .

I haven't ICE'ed GCC in a while actually. My point was that no
software is perfect. Few commercial compilers are actually C99
compliant just like GCC isn't. The all have errata, bugs, patches,
etc.

I have no idea why you have such a raging hard on to dis GCC, and
frankly I don't care. People are entitled to their opinions, just not
their own facts.

Tom
 
T

Tom St Denis

In message <[email protected]
As a compiler vendor I'd assume your test harness includes conformance
testing for [at least] C99, let alone in addition to C90 and/or C89.

Why?  Most compiler are C95 and NOT C99.    You don't test a C95
compiler with a C99 test suit.

Except if you claim your compiler is C99 compliant...

Tom
 
L

lawrence.jones

jacob navia said:
If you don't want // comments just do not use them.
If my compiler doesn't COMPLAIN about

// This is a single line comment

that doesn't mean that it is non conforming to C89.
It will accept and correctly compile all C89 code.

Including pathological code like:

float x = 3.0//* this is NOT a // comment */5.0;

Conforming means more than making an effort to do the right most of the
time.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Staff online

Members online

Forum statistics

Threads
474,085
Messages
2,570,597
Members
47,218
Latest member
GracieDebo

Latest Threads

Top