Blair Zajac wrote:
> Hyrum K. Wright wrote:
>> Hi all.
>> Anyway, I'm soliciting feedback on the implementation and usage of
>> this feature.
>> Comments welcome.
> How does it handle the case when individual revisions are each less than
> the maximum file size for the build (say 2 Gbytes), but the packed size
> is larger than that? Is it possible to have a packed set of revisions,
> followed by an unpacked set, followed by a packed set?
Not currently. I thought about that scenario, but decided against it initially.
Actually doing the packing wouldn't be too difficult (just detect the proper
error and move on to the next shard). Rather, the problem is keeping track of
which shards are packed and which aren't. We use this information to enable
opening of pack files without having to attempt the unpacked version first. We
currently just use a single file (max-packed-rev) in the file system to track
the border between packed revisions and unpacked ones.
Adding support for a heterogeneous packing environment shouldn't be hard, we'd
just need a way to index which shards have been packed, and which haven't.
After that, it's just a SMOP. ;)
Received on 2008-11-29 20:04:54 CET