[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

svn 1.1.0-rc2 status

From: Ben Collins-Sussman <sussman_at_collab.net>
Date: 2004-08-02 23:24:58 CEST

Hello folks,

I'd really like to get a 1.1.0-rc2 tarball out the door soon... like,
in the next couple of days.

Here's a list of things that are 'new' in rc2 so far (I'll be putting
them in the CHANGES file, don't worry):

* fixed: segfault when history-following hits 'empty' revision (r10368)
* fixed: bug in symlink timestamp setting (r10351)
* fixed: prevent canonicalized URIs from ending in '/' (r10317)
* fixed: make svnserve's authn work on usernames with spaces (r10385)
* fixed: have svnserve use repos UUID as default authn realm (r10394)
* fixed: don't assume DEFAULT_OS_TYPE is bdb (r10405)
* fixed: win32 prompting bug (r10431)
* if no BDB configured, default to fsfs back-end. (r10299, 10338)
* tweak description of --without-berkeley-db (r10313)
* tweak description within repos/README.txt (r10296)
* tweaks to HACKING (r10264)
* many .po translation updates

Last week, we (okay, really maxb) spent quite a lot of energy hunting
down the mysterious BDB 'invalid-argument' bug (issue #1968). We've
been treating it as a 1.1rc2 showstopper up till now, but I'd like to
propose that we remove that label. If you look at the issue, you'll
notice that about half of the issues are probably due to a Redhat
BDB/NPTL problem. The other half of the reports are undetermined, but
many could be explained by accidental BDB upgrade/downgrade.
Regardless,

    1. ...our use of BDB hasn't changed in ages
    2. ...we have no evidence that this problem is specific to 1.1. Some
       of the reports are about 1.0.x, for example.
    3. ...we have no way of gathering detailed information, because
       of the way we handle BDB errors. ("invalid arg" masks the real
       error message)
    4. ... if there really *is* something wrong with BDB or our use of
       it, then we need to get fsfs released as soon as we can!

The point is, we're no longer gaining anything by stalling rc2 on this
problem. In fact, the *best* thing we can do is apply a forthcoming
patch from maxb which adds verbose BDB error-reporting. This will
help us capture details if any rc2 users can reproduce.

So, back to the rc2 release: here's the TO-DO list, AFAICT:

* BDB 'invalid-arg' (#1968): get maxb verbose-error patch applied.
* STATUS file backports:
   - path-encoding bugfix
   - win32 build fix
   - bunches of perl fixes
   - mailer.py: stop calling *fs_berkeley*()
   - IRI support
* Release Management fixes:
   - ability to generate a win32 zipfile (#1982)
   - libtool 1.4.3 on OSX build bug (#1981)

The path-encoding bugfix, IRI support, and BDB verbose-error changes
probably justify restarting a 4-week soak. That seems to be the
popular consensus.

(Yes, breser, we'll be careful not to call it "IRI" support, but
rather, "commandline client support for spaces and non-ascii chars in
URLs".)

Maxb: can you get that verbose-error BDB patch committed?

Breser: need any help with the release management issues?

Erik: need any help with .po reviews/backports?

Everyone: review the stuff in STATUS, if you can.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Aug 2 23:27:13 2004

This is an archived mail posted to the Subversion Dev mailing list.