From: H. Peter Anvin <hpa@zytor.com>
Date: Tue, 10 May 2005 14:52:54 GMT
Subject: Re: [PATCH] minor Makefile and local-pull.c edits for Darwin
Message-ID: <4280CAC6.4000805@zytor.com>
In-Reply-To: <4280C883.6080209@zytor.com>

H. Peter Anvin wrote:
> Junio C Hamano wrote:
> 
>>
>> HPA> This is just plain WRONG.  st.st_size is longer than long on many
>> HPA> architectures, including Linux/i386.
>>
>> HPA> The easiest way to deal with it is to #include <inttypes.h>, use %jd
>> HPA> and cast it to (intmax_t).  That is, however, a C99-ism.
>>
>> Actually the easiest way is to stop reporting the size.  Nobody
>> else in core GIT reports st.st_size in their error messages.
>>
>> Although I agree with you that what you say about the size of
>> st.st_size is correct, in GIT world view, apparently "unsigned
>> long" is big enough to hold st.st_size all over the code.  Would
>> you recommend tackling that assumption as well?
> 
> Probably.  It's an off_t.
> 

That being said, there are also a whole bunch of assumptions that any 
object can be memory-mapped *plus* fit uncompressed in memory... that's 
obviously not going to be the case for large files.

On the other hand, one has to start cleaning up somewhere...

	-hpa