Re: [Bug-wget] [PATCH] Invalid Content-Length header in WARC files, on some platforms

[Tim Ruehsen]

Am Monday 26 November 2012 schrieb Ángel González:
> On 26/11/12 00:04, Giuseppe Scrivano wrote:
> > that is a good idea, we can add something like:
> > 
> > #define MAX_INT_TO_STRING_LEN(x) (sizeof(x) * 24082 / 10000) + 2
> > 
> > where `x' in our case will be off_t.
> > 
> > I am not sure if such a function already exists somewhere.
> > 
> > Does it improve readability?
> 
> Well, that just brings the question «Where does 24082 come from?» but if
> that used for eg.

simply said (well, not simply - it had to refresh my knowledge ;-)

sizeof(x) = log256(max(x))
=> sizeof(x) = log10(max(x)) / log10(256)
=> sizeof(x) * log10(256) = log10(max(x))  [log10(max(x)) is what we need !]
=> sizeof(x) * log10(256) = log10(max(x))
=> sizeof(x) * 2,4082... = log10(max(x))
~= sizeof(x) * 24082 / 10000 = log10(max(x))

and for completeness: plus one for trailing \0 and plus one for the sign

But why so complicated ? 32 is a good value even for 64bit and aligns 
succeeding memory to a 16 byte boundary.
If you want to save memory, you could calculate the needed size replacing 
max(x) by your value and make a inline function out of MAX_INT_TO_STRING_LEN - 
ok, but then you need the C99 feature 'variable length arrays' ;-)

MAX_INT_TO_STRING_LEN should be renamed to make clear that we are talking 
about converting int to a DECIMAL string, e.g. MAX_INT_TO_DEC_LEN.

Regards, Tim