Current state of VIEW bugs in MySQL 5.0

While the majority of the really serious and crashing bugs are fixed right now in MySQL 5.0, there are still a lot of bugs that only surface in some very specific queries. One area that seems to be more heavily affected in particular are VIEWs.

To keep track of the ongoing progress I just compiled a list of my VIEW bugs that have just been fixed with the current releases or are probably going to be fixed in one of the next. Interestingly, I hit five of them (#19077, #19089, #19111, #19114, #19714) with just one single set of nested VIEWs. I bet you believe me, that it took me some time to figure out why my final query just wouldn't return the result I expected.

Bug #15950 (no schedule for fix available yet)
Functions returning the current date or time (such as NOW()) get optimized away to a constant when used inside BETWEEN ... AND ... at the time of the creation of a VIEW. This will of course render the VIEW useless as time progresses.
This is the longest standing of my VIEW bugs, posted back at the end of last year, and still in verified state. But you can easily work around it by just using <= and >= operators instead of BETWEEN ... AND ....
Bug #19111 (no schedule for fix available yet)
A TRIGGER that selects from a VIEW on the same table that fires that TRIGGER fails with ERROR 1443 (HY000): The definition of table 'v' prevents operation INSERT on table 't'. The strange thing is, that this worked (at least for some cases) in versions before 5.0.20. If you convert the VIEW to an inlined SELECT statement directly inside the TRIGGER it also works.
Bug #19714 (no schedule for fix available yet)
COALESCE() returns a wrong data type in some VIEWs, e.g. BIGINT even if only INTs are listed as arguments.
Bug #19089 (to be fixed in 5.0.23 or 5.0.24)
There are problems with VIEWs regarding DEFAULT values of underlying tables. The bug affects some CREATE TABLE ... SELECT statements selecting from a VIEW. MySQL could fail with an ERROR 1067 (42000): Invalid default value on such a command or could succeed but wouldn't propagate the DEFAULT settings to the newly created table.
This is currently in patch approved state and will probably be fixed in 5.0.23 or 5.0.24.
Bug #19077 (fixed in 5.0.23)
Nested derived tables (built from unnamed views or real VIEWs alike) returned wrong results for certain data types if they were using certain aggregate functions (such as MAX() or MIN()).
This will be fixed in 5.0.23.
Bug #19114 (fixed in 5.0.21)
DESCRIBE and information_schema was broken for VIEWS with a stored FUNCTION in the SELECT list to which the user had no explicit EXECUTE rights but only implicit through the use of the VIEW defined as SQL SECURITY DEFINER.
This had been detected in 5.0.20 and just disappeared in 5.0.21.
Bug #15153 (fixed in 5.0.21)
A CONVERT_TZ() failed if used in the SELECT list of a VIEW (even after a corresponding earlier bug had been fixed).
This had been fixed in 5.0.21.

Please note that there were and are other issues with VIEWs. This is just the list of bugs that I discovered.

I also think there might still be some more of the especially minor bugs (such as data type issues) hidden inside VIEWs. So please pay attention when you discover something that just seems to look strange and try to post repeatable bug reports. The developers do a great job in fixing them.