[AG-TECH] AGVCR - Windows bug fixed, released 1.5.10
Derek Piper
dcpiper at indiana.edu
Tue May 2 11:49:51 CDT 2006
I've found and fixed the problematic Windows 'hang' bug, it was caused
by an error in getting the current time. The value would 'wrap' far
sooner than the 49.7 days that Windows has for its timeGetTime() value
(about 1000 times faster.. roughly every hour after starting, but sooner
if the machine had been up longer) and it would cause the time that came
back from the function to be less than what was originally recorded.
This threw off the code that called the relevant GTK functions to update
the display and thus the display appeared 'frozen' since it wasn't
responding to windowing system events any more. It also messed up the
timestamps of a recorded file.
In the course of tracking down the bug I ended up converting AGVCR to
work in a 'threaded' model, although the current release 1.5.10 does not
have that enabled and works (unthreaded) just as previous releases. If I
don't go mad thinking about threaded code, maybe I'll move to that for
real releases. It works for the most part, just a few 'niggles' (mainly
with popups interoperating between threads).
Anyway, I thought I'd release this now to get the bugfix out there for
Windows users.
Derek
Derek Piper wrote:
>
> To further the investigation, I have found that AGVCR doesn't
> actually crash or hang at all, at least not in of itself. By enabling
> the telnet mode, I was able to telnet to it even though the window was
> 'not responding'. It was still receiving and recording to disk just
> fine. Issuing a command 'help' actually got things 'unstuck' on the GUI
> side, so that may be a work around until I can track down what's
> happening. It seems to be again something to do with the GTK stuff
> though and not actually anything with the underlying AGVCR operations
> themselves.
> This may have been why upgrading GTK made some slight difference to
> the frequency of the apparent 'hangs'.
> An update to AGVCR Launcher (0.4) allows you to enable the 'remote
> control' mode from the Venue Client.
>
> http://iri.informatics.indiana.edu/~dcpiper/agvcr/
>
> If anyone has had anything similar happen with GTK apps, let me know.
>
> Derek
>
> Phillips, Joseph wrote:
>
>> Hi Derek,
>> We'd certainly appreciate anything you can uncover- the AGVCR seems
>> to crash at random, even when it's not playing or recording anything.
>> Sometimes it will record fine for 45 minutes and never freeze, and
>> other times it will lock up after 10 minutes, recording or not. When
>> it does freeze while recording, it always corrupts the file being
>> recorded. I agree that it's probably some windows-specific glitch-
>> unfortunately, that's what we're running on our machines, which are
>> 'former' InSors cubes. Any help you can offer would be great- thanks
>> again for getting back to me so fast with the GTK update. Even with
>> the crashes, AGVCR is still an impressive piece of work!
>>
>>
>> Joseph Phillips
>> Simulation Technician
>> Center for Simulation and Safety in Healthcare
>> University of Chicago
>>
>> ------------------------------------------------------------------------
>> *From:* Derek Piper [mailto:dcpiper at indiana.edu]
>> *Sent:* Mon 4/24/2006 7:27 AM
>> *To:* Phillips, Joseph
>> *Cc:* ag-tech at mcs.anl.gov
>> *Subject:* Re: [AG-TECH] RE: AGVCR
>>
>>
>> Hi Phillip,
>>
>> Still crashing though?.. that's not good. I'll try and work on
>> that. It
>> runs for hours on Linux without problems (that's the machine I have with
>> the most disk space so it's where I record our meetings).
>> Something about Windows or windows-specific code, or GTK under
>> Windows ...
>>
>> Derek
>>
>> Phillips, Joseph wrote:
>> > Thanks for the quick help Derek- it worked! It seems to be working
>> fine
>> > now. The program has still crashed twice during playback since I made
>> > the switch to GTK 2.8.9, but it doesn't seem to be a regular thing
>> like
>> > before.
>> > Thanks again- looking forward to more of your additions in the future!
>> > > Joe
>> > >
>> > Joseph Phillips
>> > Center for Simulation and Safety in Healthcare
>> > University of Chicago
>> > jphillips at dacc.uchicago.edu <mailto:jphillips at dacc.uchicago.edu>
>> > > >
>> ------------------------------------------------------------------------
>> > *From:* Derek Piper [mailto:dcpiper at indiana.edu]
>> > *Sent:* Thu 4/20/2006 2:40 PM
>> > *To:* Phillips, Joseph
>> > *Cc:* ag-tech at mcs.anl.gov
>> > *Subject:* Re: AGVCR
>> >
>> >
>> > Hi Phillip,
>> >
>> > I was able to recreate the problem on a windows machine (AGVCR
>> > appeared
>> > to lock up after 16 minutes), and it was running GTK 2.6.10 as a
>> > runtime. I then removed GTK 2.6 and installed GTK 2.8.9 (the latest
>> > available Windows installer) and re-tried the test in the same venue
>> > (Venue Client didn't even get shut down between tests) and AGVCR has
>> > been running fine for over 45 minutes and still going as I write this
>> > email, so I would suggest trying GTK 2.8.9 to see if it helps you too.
>> > I think I will need to update the Windows GTK runtime to
>> 2.8 and
>> > possibly put a note on the website. Given the apparent GTK bug with
>> > AGVCR running on GTK 2.6 I may even move to compiling the Windows
>> build
>> > against 2.8 to make it 'compulsory' to use 2.8.
>> >
>> > Derek
>> >
>> > Phillips, Joseph wrote:
>> > > We've been attempting to use AGVCR 1.5.9 for video feedback in our
>> > > medical simulation center, but the application window seems to
>> go blank
>> > > and lock up after letting it run for more than 10 minutes or so.
>> > > Invariably, the file that was being recorded when the application
>> > > crashed can't be played back later, even after restarting. We're
>> using a
>> > > WindowsXP (formerly InSors) machine with a Videum capture card
>> and four
>> > > cameras.
>> > > If we can get it working, we'd like to use it for recording
>> our sim
>> > > debriefings as well, including the playback or the earlier
>> recording.
>> > > Has anyone tried using AGVCR to record playback streams from
>> AGVCR on
>> > > another machine? It seems like it should work, right? Any help
>> or advice
>> > > is greatly appreciated- thanks!
>> > >
>> > > Joe Phillips
>> > > Center for Simulation and Safety in Healthcare
>> > > University of Chicago
>> > > jphillips at dacc.uchicago.edu <mailto:jphillips at dacc.uchicago.edu>
>> >
>> > --
>> > Derek Piper - dcpiper at indiana.edu - (812) 856 0111
>> > IRI 323, School of Informatics
>> > Indiana University, Bloomington, Indiana
>> >
>>
>> --
>> Derek Piper - dcpiper at indiana.edu - (812) 856 0111
>> IRI 323, School of Informatics
>> Indiana University, Bloomington, Indiana
>>
>
--
Derek Piper - dcpiper at indiana.edu - (812) 856 0111
IRI 323, School of Informatics
Indiana University, Bloomington, Indiana
More information about the ag-tech
mailing list