3.2 would be rushing and this is a major change which is not fixing a new regression. I noticed that the cache now gets the wiki name from the document whereas it used to get it from the context. Does anyone see any potential problems with this? +1 for 3.3 Caleb On 10/05/2011 04:40 AM, Denis Gervalle wrote:
On Tue, Oct 4, 2011 at 18:10, Denis Gervalle <[email protected]> wrote:
On Tue, Oct 4, 2011 at 17:25, Denis Gervalle <[email protected]> wrote:
Hi devs,
I need to fix, and I have a proposed commit at https://github.com/dgervalle/xwiki-platform/commit/068c24793dd77daa826765e30...
Following remarks from Thomas, a better commit is at https://github.com/dgervalle/xwiki-platform/commit/298c87ea8b5957d539d9a6323...
Basically, it does the following:
a) Add XWikiDocument#getKey() which return a unique text key like 8:fullname2:fr
b) Add XWikiDocument#getKey(context) which return a unique text key like 5:xwiki8:fullname2:fr
Following Thomas comments, I have reviewed this commit again, the new one is at https://github.com/dgervalle/xwiki-platform/commit/a80fe5de8dd21830aa8ae7fbf...
and point a) and b) are now:
a) Add XWikiDocument#getLocalKey() which return a local unique text key like 5:space4:name2:fr
b) Add XWikiDocument#getKey() which return a unique text key like 5:xwiki5:space4:name2:fr
c) Change XWikiDocument#getId() to use the lowest 64bits of an MD5 hash of XWikiDocument#getKey()
d) Deprecate all XWikiCacheStore#getKey(...) and use XWikiDocument#getKey(context) for the XWikiCacheStore map
e) Introduce a data migration to convert existing database, also fixing links, rcs, attachements and deleted attachements use of docid.
I am +1 only after XWIKI-7006 is fixed (see my previous vote), since any improper access to an outdated database would corrupt it. Currently the patch is ready for 3.3, but if XWIKI-7006 goes in 3.2, I am also +1 to have this in 3.2.