fixes before 3.4gamma

Here's my latest summary of (mostly) important, final fixes for 3.2beta, 
before going to gamma.  Scott is perhaps compiling these already, but I 
figured I would try to summarize them also since I will be transitioning 
into update-stream-maintainer soon.

Most of these are appropriate beta fixes, in the sense that they fix 
recently introduced bugs, not long-standing bugs.  Hopefully, this will 
be it.

A rough schedule: After these go to the internal stream, they should 
probably go public soon (within a day).  Then, if no further immediate 
problems crop up within the next 3-4 days, we could declare 3.4 as 
gamma. (and Scott would release a candidate image)  Then, I believe we 
were going to wait at least a week or so as gamma, and if there are 
still no problems, it can be declared final.

Does this sound about right?  I guess we are slightly behind our goal of 
releasing 3.4 at the beginning of the year, but we're not doing too 
badly just yet. :-)


1. [FIX] Another Acorn FileDir bug fix (21 December 2002) * Tim Rowledge 
(21 December 2002)
- fix for AcornFileDirectory only which was apparently meant to appear 
in update #5141.

2. [FIX] MoreWeakMsgSendFixes-nk (21 December 2002) * [email protected] 
(21 December 2002)
- a fix to #5155 which was recommended by Ned and Rob Withers. (or, 
could use Rob's further tweaked MoreWeakMsgSendFixes-nk-rww.2.cs posted 
to SqF list on 12/30, if that's okay with Ned)

3. [FIX] ProjectArchiveFixes-nk (20 December 2002) * Ned Konz (20 
December 2002)
- fixes a serious recently-introduced bug with project loading & saving

4. [FIX] minor VM fixes to 3.4   (MorePluginTermination-JMM.1.cs.gz & 
vmparameterTo25-JMM.1.cs.gz, posted to SqF list on 12/20)
- VM fixes agreed to by John McIntosh and Tim (which will apparently 
help Alan's demos).  Will only affect VM builders, so it probably won't 
hurt to go in 3.4.

5. "Trouble with 5150UUID-fix"
- include Tim's fix to #makeUnixSeed (UUIDGEnBugFix.1.cs), posted to 
squeak-dev on 12/20.
- include my fix to remove the TimeStamp dependency by replacing 
"TimeStamp current asSeconds" with "Time totalSeconds" in 
UUIDGenerator>>makeSeed.
- John mentioned an "Improper store..." error which he had seen, but we 
may have to let that go for now if there's no fix, and Tim sez his 
UUIDGEnBugFix.1.cs probably fixes it.  (Although will catching a 
FileStreamException work for this? hmm.)  I guess we should just leave 
this alone for now.  (An alternative is to dump #makeUnixSeed entirely.)

In addition to these, Ned just mentioned a couple others which perhaps 
stretch the "critical" bugfix criteria a bit, but they seem localized 
enough that they would be safe to add:

6. VMMakerNicerFor3.4-nk (16 December 2002)
- VMMaker-only fix, won't hurt anyone else.  (Oops, wait a minute, this 
changeset seems to open VMMaker as a side-effect of executing "VMMaker 
initialize" in the postscript?  We don't want that!  Help, Ned. :) )

7. TestRunnerInWorldMenu-nk (16 December 2002)
- I also agree this should be in the open... menu.  It does the safe 
menu-registering thing.

And one more, which maybe should wait:

8. FileListCleanup-nk (13 December 2002)
- I wonder if this should wait until 3.5alpha.  It's basically just a 
refactoring which removes a bunch of methods, and it's conceivable that 
external code could reference these?


Also, I posted last month on squeak-dev about various SUnit failures:
-------------------------

- TestUUIDPrimitives>>testCreationRandom - Only occurs under Windows.  
This failure should be expected under Windows in some cases, according 
to Andreas. (see his Dec 22 post on squeak-dev)  So, we should probably 
ignore this problem for now, and perhaps alter the test in 3.5alpha.  
(Although I sent some generated UUID's from my machine to John to see if 
he wanted to look at the problem.)

- FileDirectoryTests>>testDeleteDirectory
- FileDirectoryTests>>testDirectoryExists
- FileDirectoryTests>>testExists

These need to be fixed by the VM maintainers.  Already fixed under 
Windows, John is looked at this for the Mac OS X carbon VM, and I guess 
Ian should look at this one too for the Unix/Linux VMs.  (See Ned's 
12/21 post on squeak-dev)

So basically, I guess we don't need to do anything via updates for any 
of these SUnit failures.

- Doug