Re: [PATCH 09/13] drop objects larger than --blob-limit if specified

From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Apr 6, 2007, 15:49 UTC
Message-ID: <Pine.LNX.4.64.0704060845120.6730@woody.linux-foundation.org>
In-Reply-To: <alpine.LFD.0.98.0704052257200.28181@xanadu.home>

On Thu, 5 Apr 2007, Nicolas Pitre wrote:

Show 6 quoted lines

> 
> > (2) Pack the object any way and let the packfile size exceed
> >     my specification.  Ignoring a clear preference from the user
> >     doesn't seem good.
> 
> It is not indeed.

Well, I think there is an easy solution.

Just go back and say that when the user limits the size, it limits the offset at which objects can *start*.

Not only is that the only thing that the index file itself cares about, it also means that

 - you don't ever need to undo anything at all (because you know the 
   starting offset) when you begin packing a new object.

   This should simplify your patch a lot.

 - the object size issue just goes away. Sure, the pack-file limit looks a 
   bit strange (it's no longer a hard limit on the *size* of the 
   pack-file, just on the object offsets), but let's face it, we really 
   really don't care.

And in practice, by setting the pack-file limit to 2GB (or even 1GB), you never ever have to worry about the 32-bit filesystem limits any more, unless your repository is fundamentally so screwed that you simply *cannot* reporesent it well on something like FATFS (ie any object that is 2GB in size will probably have a blob that is even bigger, and FATFS already has problems with it).

So in practice, just limiting the index offsets is what you really want anyway.

		Linus