Re: [DBD-SQLite] SQLite 3.7.0 now in SVN - new bugs to fix

2010-07-22 Thread Adam Kennedy
As soon as someone fixes the tests, ping the list and I'll cut the dev release.

Adam K

On 22 July 2010 12:54, Darren Duncan dar...@darrenduncan.net wrote:
 Hello,

 Today the next major version of SQLite was released, and below the dashed
 line is a copy of D. Richard Hipp's release announcement.

 I have committed that version into the DBD::SQLite Subversion repository, so
 when you svn-up you should have that version.

 From my own testing, the simple substitution of 3.7.0 for the previous
 3.6.23.1 broke a couple of our tests:

 t/08_busy.t                            (Wstat: 512 Tests: 23 Failed: 2)
  Failed tests:  11, 22
  Non-zero exit status: 2
 t/28_schemachange.t                    (Wstat: 512 Tests: 9 Failed: 2)
  Failed tests:  7-8
  Non-zero exit status: 2

 ... and these cite 'disk I/O error' as the problem.

 I did not debug the problem aside from that I tested both with the
 Subversion head plus the latest CPAN release 1.30_03; for both, the above 2
 errors occur with 3.7.0 and not with the prior SQLite version.

 Separately, Subversion head had a few other failures in make test on my
 system while the pristine 1.30_03 had no errors; those would just be related
 to DBD::SQLite changes.

 Given the major changes in 3.7.0, I recommend cutting a new CPAN developer
 release 1.30_04 as soon as all the tests pass.

 For my part, I can at least try testing on another system or 2 in case for
 some reason the new failures are specific to my first machine, though if
 anyone else can reproduce the above test failures then that won't be
 necessary.

 -- Darren Duncan

 -

 SQLite version 3.7.0 is now available on the website

     http://www.sqlite.org/

 The most important change in version 3.7.0 is that SQLite now supports
 write-ahead logs as an optional method for transaction control, for improved
 performance and concurrency.  Additional information can be found here:

    http://www.sqlite.org/wal.html

 We are actually already using the write-ahead logging feature on the SQLite
 website itself, in the Fossil DVCS that tracks all changes to the SQLite
 source tree. (Yes, the SQLite write-ahead log code is stored in an SQLite
 write-ahead log database - how's that for recursion!)

    http://www.sqlite.org/src

 The added concurrency of the write-ahead log journaling mode allows multiple
 users to be doing extended read operations, such as checking out historical
 versions of the SQLite code or looking at extended timelines simultaneously
 with developers making new checkins, adding or editing tickets, or actually
 rebuilding the entire 10-year source code database.  The write-ahead log
 code has performed very well so far for us.

 Version 3.7.0 also marks the official cut-over to our new SQLite logo and a
 new color scheme for the website.  We hope you like the new look.

 The 114-day time span since the previous release (version 3.6.23.1) is the
 longest span between consecutive releases in the 10-year history of SQLite.
 Much of that time was spent testing and stressing the new write-ahead log
 feature.  This is probably the most thoroughly tested release of SQLite that
 we have every produced.  And so even though the write-ahead logging feature
 is entirely new, we are very hopeful that the 3.7.0 release will prove to be
 stable and robust and ready for production use.

 Of course, if you do happen to run into problems, please let me know at
 once. Thanks!

 D. Richard Hipp
 d...@sqlite.org


 ___
 DBD-SQLite mailing list
 DBD-SQLite@lists.scsys.co.uk
 http://lists.scsys.co.uk/cgi-bin/mailman/listinfo/dbd-sqlite


___
DBD-SQLite mailing list
DBD-SQLite@lists.scsys.co.uk
http://lists.scsys.co.uk/cgi-bin/mailman/listinfo/dbd-sqlite


[DBD-SQLite] [Fwd: Re: [sqlite] SQLite version 3.7.0]

2010-07-22 Thread Darren Duncan



 Original Message 
Subject: Re: [sqlite] SQLite version 3.7.0
Date: Thu, 22 Jul 2010 18:31:42 +0700
From: Dan Kennedy danielk1...@gmail.com
To: General Discussion of SQLite Database sqlite-us...@sqlite.org


On Jul 22, 2010, at 1:08 PM, Darren Duncan wrote:
snip

On that note, I got this report from someone on Windows:

Latest SVN trunk tested on win32 Strawberry perl v1.12.1 : all  
tests pass,

no problem.

... and I was using a Unixen.


Is there any way your tests could be deleting a database
file while there is still an open sqlite connection to it?

With 3.7.0, if the underlying database file is unlinked
while you are connected to it, then you try to write to
the database, you get SQLITE_IOERR_FSTAT. Earlier versions
would continue writing without causing an error.

You cannot delete a file while it is open on windows, so
this doesn't come up on win32.

This happened with a couple of Tcl tests too.


___
DBD-SQLite mailing list
DBD-SQLite@lists.scsys.co.uk
http://lists.scsys.co.uk/cgi-bin/mailman/listinfo/dbd-sqlite