Implicit assumptions (was: Re: Some fun with -O2)

[Mark Millard]

On 2021-Jan-15, at 11:22, Walter von Entferndt <walter.von.entferndt at posteo.net> wrote:

> At Freitag, 15. Januar 2021, 19:35:40 CET Mark Millard wrote:
>> Have you read a (fairly modern) C standard or its officially
>> published rationle? You might want to.
>> 
> Honestly, no.  The price to download the official standard PDF from https://
> www.iso.org/standard/74528.html is ~200 CHF.  If you can send me a link to 
> download a copy, I'd be thankful.

The rationale vintage that I have a copy of is available at:

http://www.open-std.org/jtc1/sc22/wg14/www/C99RationaleV5.10.pdf

The rationale is normally a better starting point than the standard
for guiding understanding of the standard --but is not normative.

What makes the rationale a matter starting point for interpreting
the standard is that, while the standard should have the more
overall properties documented in the rationale, one has to infer
various properties from more detailed rules that were designed to
have the properties (and may be better at covering limiting
conditions than the more overall rationale material).

But the rationale can not be used to answer all questions.

As for drafts of vintages of the standard:

http://www.open-std.org/jtc1/sc22/wg14/www/projects#9899

has links (the first 4 N???? links) to final drafts (or
very-early next-standard drafts) of the standards for C2x,
C17, C11, and C99.

My understanding is that C17 mostly dealt with clarification
requests, other than the ATOMIC_VAR_INIT one leading to
depreciating ATOMIC_VAR_INIT instead.

As I understand, FreeBSD basically targets C99 (possibly without
things made optional or depreciated in C11 or later: future
compatibility biased).


> Any other good reference manual either in 
> PDF or HTML (tarball or thelike) would also be fine.  I showed my source 
> (https://code-reference.com/c/) which I quickly looked up on the net.
> 
>> From the officially published C99 rationale (page labeled 11,
>> Terms and definitions):
>> 
>> QUOTE
>> ) All objects in C must be representable as a contiguous sequence of bytes,
>>  each of which is at least 8 bits wide.
>> 
>> ) A char whether signed or unsigned, occupies exactly one byte.
>> 
> That means it does not make any difference to use either NBBY or CHAR_BIT?  
> Maybe CHAR_BIT is preferable, because it is C standard (guaranteed to exist on 
> all platforms), whereas NBBY is not since it's in include/sys?

I'd say that using C's CHAR_BIT means less variation in the code from system
to system and less knowledge of system specifics required. So in code not
intended to be system specific, CHAR_BIT would be preferred. That does not
cover code intended to be system specific.

> Beside that, can you affirm the fix I suggested is correct & portable?

Before I could make any grand overall claim about the code, I'll
need to take time to think about all of it, instead of just reporting
things that I happen to quickly notice while scanning the exchange.

I may get to this, but I'm not sure when.


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)