Post #171,121
8/26/04 6:07:59 PM
|
What is with gcc, anyway?!?
OK folks...Here's today's quiz: Given the following C code: \n#define MAX(a, b) ((a) > (b) ? (a) : (b))\n\nint main (int argc, char ** argv)\n{\n assert(MAX(0x7FFFFFFF, 0x80000000) == 0x7FFFFFFF);\n return 0;\n}\nWhen compiled for a 32-bit architecture (x86), why would this assert? When compiled with gcc (V3.3.1), it does. According to the ANSI standard:
The type of an integer constant is the first of the corresponding list in which its value can
be represented.
suffix |
Decimal
Constant |
Octal or Hex
Constant |
|---|
None |
int
long int
long long int |
int
unsigned int
long int
unsigned long int
long long int
unsigned long long int |
Well, we have no suffix, and both values can be represented as ints, so they should both be ints, and so therefore 0x7FFFFFFF is greater than 0x80000000, so the assertion should be true. So what's the deal? Is it that the gcc people can't be bothered with ANSI compliance? Or is there some kind of error here that I'm just not seeing? (To be fair, this was gotten via cygwin; perhpas the cygwin people dicked it up?) (And don't even get me started about the fact that they can't seem to be bothered implementing __STDC_VERSION__....) /me is not pleased! jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,123
8/26/04 6:22:56 PM
8/26/04 6:25:16 PM
|
What is typeof(0x80000000)?
[edit note: I guess I did read it backwards after all - NeverMind]. I'm assuming that you think the number should be negative? By your table, - int
- unsigned int
- long int
- unsigned long int
- long long int
- unsigned long long int
I don't see how you expect to ever have a constant of type of unsigned, since the number of bits involved is always the same as the signed counterpart? That is, how could you ever get an unsigned integer between 0 & 255, when the signed integer would go from -128 to 127 over those same 8 bits? Perhaps I'm missing something obvious though.
Edited by ChrisR
Aug. 26, 2004, 06:25:16 PM EDT
|
Post #171,125
8/26/04 6:30:08 PM
|
right, add an L
-drl
|
Post #171,126
8/26/04 6:34:49 PM
|
Ain't that a UL
My See is awful rusty.
|
Post #171,140
8/26/04 9:11:16 PM
|
This is why C is sometimes called "high-level assembler."
Although I enjoyed using C to poke around inside MS-DOS and Windows, nowadays I prefer more modern languages. With things like Big Integers. :-)
Wade.
Is it enough to love
Is it enough to breathe
Somebody rip my heart out
And leave me here to bleed
| |
Is it enough to die
Somebody save my life
I'd rather be Anything but Ordinary
Please
| -- "Anything but Ordinary" by Avril Lavigne. |
|
Post #171,206
8/27/04 9:35:47 AM
|
Re: This is why C is sometimes called "high-level assembler.
C is basically the lingua franca for embedded these days (although C++ is starting to make inroads there, much to Todd's chagrin ;-) )
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,209
8/27/04 9:43:22 AM
|
Actually, that would be a U (or u)
The 'L' or 'l' means long and implicitly signed. Note, however, that \nint main (int argc, char ** argv)\n{\n assert(MAX(0x7FFFFFFFL, 0x80000000L) == 0x7FFFFFFFL);\n return 0;\n}\n\n also asserts. And while LordBeatnik's respone below makes sense and probably explains my original bitch, it certainly doesn't explain the above. <mumble> jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,922
8/31/04 11:34:46 AM
|
What about casting the values?
something like: \n assert(MAX((long)0x7FFFFFFF, (long)0x80000000) == (long)0x7FFFFFFF);\n
|
Post #171,942
8/31/04 11:55:24 AM
|
Yes, that was the solution I used
This seems to be the best solution; it gives me the explicit bit pattern I'm looking for, and the explicit instruction to treat the patters as signed. Hopefully, when the first draft of the C++0x spec comes out, they will have cleaned up the spec with regards to this particular issue.
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,158
8/26/04 11:05:40 PM
|
Re: What is with gcc, anyway?!?
It looks reasonable to me that 0x8000000 got promoted to unsigned int. If it wasn't, what would be the point of having the unsigned types in the list? That's why the unsigned types are not listed for decimal numbers but are for hex. 0x80000000 wouldn't be positive as an int, so it got promoted to unsigned int. Thus, the MAX returns 0x80000000.
Dave "LordBeatnik"
|
Post #171,207
8/27/04 9:39:19 AM
|
I guess...
What you say makes sense. The problem here is the phrase in the standard The type of an integer constant is the first of the corresponding list in which its value can be represented. [emphasis added] 0x8000000 certainly "can be represented" as an int, so silly /me thought that, if it "can be represented" as an int, it would be. Thanx for the input! jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,402
8/27/04 6:51:20 PM
|
If there was no ambiguity...
If there were no ambiguities in standards, what reason would they have to host another party^Wstandards meeting? 8)
Dave "LordBeatnik"
|
Post #171,751
8/30/04 5:29:20 PM
|
For the next set of standards?
/me slaps head in disbelief that I even thought that...
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,225
8/27/04 11:32:17 AM
|
Hex constants are assumed to be positive
0x80000000 would make an int to be negative, so it cannot be represented as int.
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,262
8/27/04 1:15:45 PM
|
Re: Hex constants are assumed to be positive
Sez who?
(OK, let me rephrase that with the smart-ass control turned down...Where in the ANSI standard does it say that hex constants are assumed to be positive?)
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,268
8/27/04 1:27:28 PM
|
Setting the high bit is processor dependent.
It's not carved in stone anywhere that that's the only way to do it. Most do, but that's not specified generally anywhere.
|
Post #171,280
8/27/04 1:53:36 PM
|
OK, then the standard should be saying
"presentable in the implementation's architecture". What do you do with those nits, eat them?
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,304
8/27/04 2:27:25 PM
|
No I actually write code based on them
and then spend entirely too much time recovering from other's inability to use the English language effectively. What do you do with them?
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,329
8/27/04 3:24:32 PM
|
Ignore them and use my common sence
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,752
8/30/04 5:29:56 PM
|
I'll bet debugging your code is a real joy...
;-)
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,885
8/31/04 10:04:06 AM
|
Nope
The joy comes when you read your standard, forget your common sence and assume that same things work the same way everywhere.
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,899
8/31/04 11:04:10 AM
|
And so tell me, O Oracle of Common Sence [sic]
What exactly is it that is so nonsensical about defining a hex value and assuming it is treated as a signed value? We supplicate ourselves breathlessly at your feet awaiting such pearls (or is that perls) of wisdom as you may deign to proffer.
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,902
8/31/04 11:10:45 AM
|
offensive foul, ball to Ark
You violated the "be kind to non-native English speakers" (who do extremely well in a totally different language and make us look stupid by comparison) rule.
Just point out that it's "sense".
BTW Ark - "licence" and "license" are both right. One is British. Likewise "offence" and "offense", "defence" and "defense". But "electric fense" is always wrong.
-drl
|
Post #171,909
8/31/04 11:17:52 AM
|
Rest assured,
I'd be making the same mistakes in Russian as well.
Por syntax checus in Cium - Deo Gratie.
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,913
8/31/04 11:25:55 AM
|
:) ok
-drl
|
Post #171,914
8/31/04 11:26:24 AM
8/31/04 11:26:59 AM
|
OK, OK...
I'll take my yellow card and give up an indirect free kick....
;-)
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
Edited by jb4
Aug. 31, 2004, 11:26:59 AM EDT
|
Post #171,904
8/31/04 11:14:05 AM
|
Well, while the tone is disagreable, the question is
(almost) legit \n#include <stdio.h>\n\n\nint main(int argc, char**argv) {\n int c = -0xFF;\n printf("signed: %d, hex: %X\\n", c, c);\n return 0;\n}\n\n$ gcc -o test test.c; ./test\nsigned: -255, hex: FFFFFF01\n And the answer is <drrrrum rrrroll>: The hex constants are treated as positive because they lack the "minus" in front. Tada! --
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,912
8/31/04 11:25:35 AM
|
Point missed.
The point is not that I can generate a negative number using hex values, its that I can generate a specific bit pattern directly that is interpreted correctly using two's compliment arithmetic. In my example, I want the bit pattern 0x80000000. Interpreted as an int. Which is what I should expect, given that incomplete pile of verbage that passes as the C(99) standard in section 6.4.4.1, paragraph 5. Your example doesn't match, as you explicitly change the value of your bit pattern by applying an operator to it (the unary minus).
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,915
8/31/04 11:26:40 AM
|
If C were sane, TRUE = -1 and problem vanishes
-drl
|
Post #171,917
8/31/04 11:28:42 AM
|
Ermm...say What?
You got me on that one, Ross. How does the value of a boolean have anything to do with how a constant token is interpreted?
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,924
8/31/04 11:37:26 AM
|
Re: Ermm...say What?
It forces strict attention to signed vs. unsigned, because TRUE is signed.
IOW we don't need no steenkin Booleans.
-drl
|
Post #171,929
8/31/04 11:43:34 AM
|
Much like Jewish law
Forces strict attention to matters divine because it's so easy to transgress. Any takers?
You recommend a practice because it's easy to violate? Ouch.
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,961
8/31/04 1:01:31 PM
|
So you prefer how VB does it??
To deny the indirect purchaser, who in this case is the ultimate purchaser, the right to seek relief from unlawful conduct, would essentially remove the word consumer from the Consumer Protection Act
- [link|http://www.techworld.com/opsys/news/index.cfm?NewsID=1246&Page=1&pagePos=20|Nebraska Supreme Court]
|
Post #171,962
8/31/04 1:04:18 PM
|
Point: Ben...
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,972
8/31/04 1:47:19 PM
|
How it used to do it.
VB.Net had to play a little nicer with the other languages, so they switched to the way it's represented in C#.
|
Post #171,991
8/31/04 3:18:32 PM
|
Hehee!
VB was "embraced and extended" by C! Oh, the irony!
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,921
8/31/04 11:31:50 AM
8/31/04 11:32:44 AM
|
Why do you expect hex to be treated differently from decimal
?
When you see a decimal constant 4294967294, do you expect it to mean -2?
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
Edited by Arkadiy
Aug. 31, 2004, 11:32:44 AM EDT
|
Post #171,926
8/31/04 11:42:16 AM
|
Re: Why do you expect hex to be treated differently
In FORTH this issue never arises. An integer is what it is in relation to base 2 and the number of available bits in it. Signed vs. unsigned is a stupid, needless complication.
So: TRUE is 111111...1 and FALSE is 000000...0. There is only one TRUE and one FALSE and NOT TRUE is FALSE. Something like 1101001..1 has no logical value.
The programmatic representation is immaterial.
-drl
|
Post #171,953
8/31/04 12:37:46 PM
|
Man, that's a good question!
I suppose because I can specify a bit pattern directly, and I expected that bit pattern to be interpreted by the underlying hardware directly. But I must say, I now see your point, and it is a good one.
Thanx!
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #172,013
8/31/04 4:42:17 PM
|
Now that's clarity!
Alex
"If you can control the meaning of words, you can control the people who must use the words." -- Philip K. Dick, US science fiction writer
|
Post #172,017
8/31/04 5:00:03 PM
|
Nope, it's bad math
A 32 bit integer has 4...mumble possible values. If you really want integers and not whole numbers, then implicitly these are signed, which requires a bit. So there are 2...mumble positive values, 2...mumble-1 negative values, and zero.
If you want 4...mumble integers you need more bits.
-drl
|
Post #172,022
8/31/04 5:19:13 PM
|
But, counters of real things don't need negative values.
Alex
"If you can control the meaning of words, you can control the people who must use the words." -- Philip K. Dick, US science fiction writer
|
Post #172,023
8/31/04 5:20:08 PM
|
I'd like to add a negative number of votes for Bush. ;-)
To deny the indirect purchaser, who in this case is the ultimate purchaser, the right to seek relief from unlawful conduct, would essentially remove the word consumer from the Consumer Protection Act
- [link|http://www.techworld.com/opsys/news/index.cfm?NewsID=1246&Page=1&pagePos=20|Nebraska Supreme Court]
|
Post #172,025
8/31/04 5:23:34 PM
|
That only works on Diebold machines. :)
Alex
"If you can control the meaning of words, you can control the people who must use the words." -- Philip K. Dick, US science fiction writer
|
Post #172,024
8/31/04 5:21:51 PM
|
Certainly
But then, Ark's comment loses its pithiness (\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd\ufffd)
-drl
|
Post #172,046
8/31/04 6:59:29 PM
|
"Expressive", not "pithy" :)
"pithy" I guess would be better translated as poisonous, ядовитый - it's an idiomatic usage.
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #172,047
8/31/04 7:03:32 PM
|
Re: "Expressive", not "pithy" :)
Why can I see your Cyrillic but not mine? Some encoding mojo?
-drl
|
Post #172,172
9/1/04 9:33:56 AM
|
Dunno.
I see yours fine in Mozilla on Win 2000 after I select the proper code page - Cyrillic Windows-1251
--
"...was poorly, lugubrious and intoxicated."
-- Patrick O'Brian, "Master and Commander"
|
Post #171,339
8/27/04 4:06:40 PM
|
Wasn't trying to piss you off...
From my point of view, hex and octal are just more convienient ways to write binary.
Hex/octal/binary numbers are not negative without some convention to make them so. This varies.
The C language, and those that borrowed from it, make the tacit assumption that the high bit indicates a negative number unless it is declared unsigned, in which case it is just a binary number in the range of 0 - UINT_MAX (from limits.h) I don't know of any languages where you can write -0x123 for a negative hex number.
I should mention that I am not particularly a language pedant. I currently use C/C++ for some application work and drivers, I write firmware for a propriatary processor in a much modified, self maintained assembler, and do scripting in perl as necessary. Plus whatever needs be done. I'm not a purist. I really wasn't trying to pick nits.
|
Post #171,760
8/30/04 5:51:55 PM
|
Interesting assumption:
The C language, and those that borrowed from it, make the tacit assumption that the high bit indicates a negative number unless it is declared unsigned... An interesting assumption. I don't believe C makes that assumption at all. The underlying hardware makes that assumption, because that's simply the way it works. C (the language) doesn't care. An implementation of C cares, because it has to map tokens that represent negative values to the actual binary representation of the value. Now it has been several Decades since I saw a machine that doesn't uset two's complement arithmetic. The first one was an NCR 315 that our local bank used when I was in high school in 1968...It was a DECIMAL machine that used XS-3 notation internally. the last one I saw was a CDC Star, which is a 60-bit (!) machine that used ones complement artihmetic...which has a unique value for negative zero. The NCR machine antedated C by something like a decade, but were a C compiler to exist for that machine, it could not assume the high-order bit of a value made it negative. jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #171,379
8/27/04 5:35:03 PM
|
Hmm...while the datatype will be first reached,
IIRC, C standard does not state the size of the number, merely that shorts will be a minimum of a certain size, longs will be a minimum of a certain size, etc.
So you're short may handle sizes past 32,768.
|
Post #171,761
8/30/04 5:55:52 PM
|
True, but...
...that's not the issue here. The table indicates the progression starts at int, through long, then through long long. So we know that the value is "at least" an int. Since the target machine is a 32-bit machine, we know that the value is treated in some form of 32-bit value.
jb4
shrub\ufffdbish (Am., from shrub + rubbish, after the derisive name for America's 43 president; 2003) n. 1. a form of nonsensical political doubletalk wherein the speaker attempts to defend the indefensible by lying, obfuscation, or otherwise misstating the facts; GIBBERISH. 2. any of a collection of utterances from America's putative 43rd president. cf. BULLSHIT
|
Post #172,093
8/31/04 11:03:21 PM
8/31/04 11:10:58 PM
|
Re: What is with gcc, anyway?!?
Depends on whether the compiler implicitly promotes them to signed or unsigned integers.
The answer could change depending on whether your compiler treats 0x7fffffff and 0x80000000 as signed or unsigned integers.
I think most compilers would treat them as signed, but gcc might not. Also, in a 64 bit compiler, both would still be positive integers.
There was a guy I worked with at BEA that had a set of about 100 macros that would promote the values and test, and then dump out the compiler assumptions in a report. Very useful when you have to write portable code.
His name is David Tribble. I wonder where he is working now.
Glen
Edited by gdaustin
Aug. 31, 2004, 11:10:58 PM EDT
|
