Dear Sympa users,
Tha latest stabilisation version fo Sympa 6.0 is now out.
Download
Sympa 6.0.2 now!
Check
the release notes!
This version mainly fixes bugs that were reported since Sympa 6.0.1 was
released.
Here is a summary of the bugs fixed:
- [Submitted by S. Ikeda] When bulk_max_count was increased, some
message packets were processed more than once. Fixed to check that a
packet is really not locked before trying to send it and using a
prepared statement to quicken lock process.
- When a packet was found which corresponded to no message in the
spool table, then an empty message was sent anyway. Should such a case
arise again, the packet would be deleted.
- Reverting a part of revision #6227. This lead to datasources
loosing all their subscribers when using a single datasource and the
synchronizing failed.
- When calling send_notify_to_owner, the '$param' parameter was
used as a hash, wihout checking first that it was actually a hash. in
some cases though, such as in set_status_family_closed(), it was a ref
to an array. This caused Sympa to crash with a "Can't coerce array into
a hash" error. Fixed by moving the bad line in the hash-specific part
of the sub.
- [Submitted by M. Brethous, Orange FT Group] When a file was used
as datasource for subscribers inclusion, no test was done on the data
found in the file. This led to 100% CPU usage when, for example, a very
large excel file was used instead of a normal text file - a likely
error when sources are handled by list owners. Fixed by cancelling
inclusion when no valid email is found in the 50 first non empty lines
of the source file.
- A bug reported by K Clair (riseup) and Thomas Berry (nasa) : no
purge were done in bulk_spool_table.
- [Submitted by R. ANDERSSON] With older version of DBD::Mysql,
bulk.pl happened to die with an "Out of memory!" cry. Fixed by
increasing the version to the oldest fixing this problem. See
http://bugs.mysql.com/bug.php?id=19453 for more details.
- [Submitted by T. Berry, NASA] Some SQL statements with Sybase
failed when the actual backend was Microsoft SQL Server 2005 sp3. This
was due to a double quote in the definition of a constant breaking the
query. Fixed by replacing double quotes by slash-escaped simple quotes.
- Using Oracle 10, the logs for the list were never displayed in
the web interface. This was due to Oracle returning 0 to function
$sth->rows, before any row are fetched. fixed by first trying to
fetch rows before counting the number of rows.
- [Submitted by K. Maneva-Jakimoska, univ. Montclair] when a user
had set her name, she could not delete it. the only was to set it as '
'. fixed by setting it to undef if no value is provided by the form.
- [Reported by M. Deranek, slonko.net] Upgrade.pm had three
problems :
- a test used the wrong DBD name for SQLite, leading to a part of the
upgrade bein ignored,
- we tried to update the primary key, which is impossible in SQLite,
- we used the xrong syntax to update the indexes.
Fixed by removing the primary key update attempts and fixing the syntax.
- [Reported by J. Bazik, Brown University] In the subroutine
Lock::_lock_nfs, if File::NFSLock failed, the else clause attempted to
close a filehandle that was never set. Fixed by removing this closing
attempt.
- Makefile.am: During the make isntall, the Makefile used a line
that tested the existence of the data_structure.version file in the
sympa subdirectory of etc. If the file did not exist, then it was
created. Problems:
1- this directory was never created.
2- it was never used : in other places of the code, the actual location
used was etc, not etc/sympa.
- [Submitted by M. Deranek, slonko.net] The upgrade process failed
with SQLite, because DBD::SQLite >= 1.26_02 returns the table names
prefixed with "main".
- [Submitted by F. Lachapelle, Inverse] If the db_type is set to
Oracle and db_port is specified in the main configuration file, than
the port would appear two times in the connection string. Fixed by
removing useless isolated port insertion in the connect string, as it
is done for any connection in a central place.
- [reported by F.Fareneau, univ Valenciennes] shared
documents/folders names were badly encoded while using Encode >=
2.19. The issue was related to a low level subroutine
(Encode::MIME::Header::_encode_q()) that slightly changed its API
(expecting unicode instead of bytes parameter). Now using
MIME::EncWords::encode_mimewords() instead.
- [#6552] [Soji IKEDA] Wwsympa would stall while loading session
data including metacharacters. Existing code (tools::string_2_hash())
would loop on a regular _expression_ match.
--
David Verdin
Comité réseau des universités
|
