From: Jeff Hostetler Date: Tue, 26 Sep 2017 14:25:16 GMT Subject: Re: RFC: Design and code of partial clones (now, missing commits and trees OK) (part 3) Message-ID: In-Reply-To: <20170922155802.ab79717818578a23cc31f6fe@google.com> On 9/22/2017 6:58 PM, Jonathan Tan wrote: > On Fri, 22 Sep 2017 17:32:00 -0400 > Jeff Hostetler wrote: > >> I guess I'm afraid that the first call to is_promised() is going >> cause a very long pause as it loads up a very large hash of objects. > > Yes, the first call will cause a long pause. (I think fsck and gc can > tolerate this, but a better solution is appreciated.) > >> Perhaps you could augment the OID lookup to remember where the object >> was found (essentially a .promisor bit set). Then you wouldn't need >> to touch them all. > > Sorry - I don't understand this. Are you saying that missing promisor > objects should go into the global object hashtable, so that we can set a > flag on them? I just meant could we add a bit to "struct object_info" to indicate that the object was found in a .promisor packfile ? This could be set in sha1_object_info_extended(). Then the is_promised() calls in fsck and gc would just test that bit. Given that that bit will be set on promisOR objects (and we won't have object_info for missing objects), you may need to adjust the iterator in the fsck/gc code slightly. This is a bit of a handwave, but could something like that eliminate the need to build this oidset? > >>> The oidset will deduplicate OIDs. >> >> Right, but you still have an entry for each object. For a repo the >> size of Windows, you may have 25M+ objects your copy of the ODB. > > We have entries only for the "frontier" objects (the objects directly > referenced by any promisor object). For the Windows repo, for example, I > foresee that many of the blobs, trees, and commits will be "hiding" > behind objects that the repository user did not download into their > repo. >