STS woes

I'm experiencing problems with STS 3.1.2 (the latest version) and I'm
wondering if anyone else has experienced the same or is using it with no
problems.  I'm not able to force the problems to appear (i.e., reproduce
them) but after awhile, one or more of the following symptoms appear: 1)
Versioning a package doesn't work, it seems to start but just hangs there
until I cancel; 2) Source code for some methods shows in red (uh oh, we know
what that means); 3) I get a message in the transcript that a zombie
PackageBrowserShell has been detected; and 4) Dolphin crashes.

I thought it could be a D5.1 issue since I had switched to D5.1 at around
the same time, but now am pretty sure it's STS-related because I tried all 4
combinations of D5.03, D5.1, STS 3.1.2 and the previous STS.  I started all
"tests" with a clean image and closed all windows after installing STS.  I'm
not messing around with the source, image or change files, and I'm (still)
using Win98.

Life is tough without STS so I'm back to D5.03 with the "older" STS which
has been working fine on the stuff I'm working on.  A few unrelated
observations: 1) STS now has an XML export, but I don't see any import that
works on these - I tried importing, but hit a "self halt"; 2) D5.1 packages
may be incompatible with D5.0 since the former now include "named:" for
PoolConstants (not a biggie, but you do have to edit the package file to
remove "named: blahblah" if you want to go backwards, if you know what I
mean); 3) In my very limited time using D5.1 it seemed a bit slower - but I
don't have any objective data to back that up.

-- Louis

Louis,
"Louis Sumberg" <[hidden email]> wrote in message
news:bauecq$3r247$[hidden email]...
> I'm experiencing problems with STS 3.1.2 (the latest version) and I'm
> wondering if anyone else has experienced the same or is using it with no
> problems.  I'm not able to force the problems to appear (i.e., reproduce
> them) but after awhile, one or more of the following symptoms appear: 1)
> Versioning a package doesn't work, it seems to start but just hangs there
> until I cancel; 2) Source code for some methods shows in red (uh oh, we
know
> what that means); 3) I get a message in the transcript that a zombie
> PackageBrowserShell has been detected; and 4) Dolphin crashes.
[...]
FWIW, I've using D5.1 and STS 3.1.2 (since it came out) with no problems.
I'm using Windows XP, though.

Don

"Louis Sumberg" <[hidden email]> wrote in message
news:bauecq$3r247$[hidden email]...
>...2) D5.1 packages
> may be incompatible with D5.0 since the former now include "named:" for
> PoolConstants (not a biggie, but you do have to edit the package file to
> remove "named: blahblah" if you want to go backwards, if you know what I
> mean);

The easiest way around that is to add a #named: method to the class side of
PoolConstantsDictionary in D5 implemented as '^self new' (i.e. just ignore
the argument). This can be defined (if not present) from the package
pre-install script.

>... 3) In my very limited time using D5.1 it seemed a bit slower - but I
> don't have any objective data to back that up.

This might be an illusion caused by the "delayed opening of message boxes"
bug (1249), patch available from
http://www.object-arts.com/Lib/Update/5.1/1249.st.

In general there should be some mild performance improvements in D5.1; the
VM is essentially the same, but some compiler improvements were back ported
from D6. If you know of any specific cases where this is not so, we'd be
interested to hear about them.

Regards

Blair

Don,

> FWIW, I've using D5.1 and STS 3.1.2 (since it came out) with no problems.
> I'm using Windows XP, though.

Thanks for the feedback.  Since my post the other day I've done a lot of
testing using D5.1 with and without STS and am coming to the conclusion that
it's not STS that's the problem, but perhaps Win98.  Sorry, David, for
pointing my finger at STS, which we all know is super duper fantabulous!

-- Louis

Blair,

Thanks for the pointers on adding #named: to D5 and patch 1249.  I've also
filed in patch 1272 and will do some more testing when i get back from
vacation next week.  As I indicated to Don in another post, I'm thinking
it's a Win98 problem with D5.1, based on testing without STS and what I'm
seeing from other posters.  I did have a problem with the patches installed
but I wasn't starting with a squeaky clean image, so more testing is needed.
What does seem fairly consistent (though not absolutely reliable where I can
repeat it at will) is that I'll run unit tests (about 240 now, on Gdiplus),
step into some methods, and then find that if I close and reopen the
PackageBrowser I get a warning there's a zombie PackageBrowserShell.  If I
hit panic at that point, everything will usually disappear, with a "Dolphin"
process still running which I need to kill.  If instead of panic, I save the
image and restart it, everything seems fine, including panic.  Another
symptom that things have gone awry is that the SUnit browser will have
commands and toolbar buttons grayed out that should be available.  Bottom
line is that I'll do more testing, there may still be a red flag on this
issue, and I need to get me a new laptop *s*.  Thanks again.

-- Louis

Louis

You wrote in message news:bb3arp$54gj5$[hidden email]...
>
> Thanks for the pointers on adding #named: to D5 and patch 1249.  I've also
> filed in patch 1272 and will do some more testing when i get back from
> vacation next week.  As I indicated to Don in another post, I'm thinking
> it's a Win98 problem with D5.1, based on testing without STS and what I'm
> seeing from other posters.  I did have a problem with the patches
installed
> but I wasn't starting with a squeaky clean image, so more testing is
needed.
> What does seem fairly consistent (though not absolutely reliable where I
can
> repeat it at will) is that I'll run unit tests (about 240 now, on
Gdiplus),
> step into some methods, and then find that if I close and reopen the
> PackageBrowser I get a warning there's a zombie PackageBrowserShell.  If I
> hit panic at that point, everything will usually disappear, with a
"Dolphin"
> process still running which I need to kill.  If instead of panic, I save
the
> image and restart it, everything seems fine, including panic.  Another
> symptom that things have gone awry is that the SUnit browser will have
> commands and toolbar buttons grayed out that should be available.  Bottom
> line is that I'll do more testing, there may still be a red flag on this
> issue, and I need to get me a new laptop *s*.  Thanks again.

The failure to update the status of the toolbar buttons does sound
suspiciously like the Win98 issue reported by Steve Waring, which we believe
to be corrected by #1249 and #1272.

However, I have a suspicion that you may also be exhausting system resources
when running your tests, either because of a resource leak, or due to
delayed finalization. If resources become exhausted then all sorts of
strange things can start to happen, not just strange visual effects (such as
the fonts suddenly changing), but also Dolphin and/or Win98 can crash
completely. Personally I don't think it worth running Win98 as main
development environment, it is just too unreliable, but when doing so I
always run a resource meter to see when things are getting critical. I also
go against my usual style by sticking to 3 or 4 windows rather than my usual
20 odd.

Regards

Blair