flash (macromedia): I'm experiencing a browser crash using FileReference to upload files.
I'm using code based on the FileUpload.fla example code from Macromedia.
The example code works ok in isolation, but when incorporated into the large
app I get crashes on IE WIndows XP, FireFox Windows XP, Safari Mac OSX.
After repeatedly upload files, I eventually get a crash.

I've summited a bug report, but I'd be curious if others have experienced
problems using Fileupload.
Also I'd like any pointers on what's the best way to get the failing code in
the hands of macromedia engineering.

Here's a the first few lines from the crash in Safari:

Date/Time: 2005-10-05 15:13:05.856 -0400
OS Version: 10.4.2 (Build 8C46)
Report Version: 3

Command: Safari
Path: /Applications/Safari.app/Contents/MacOS/Safari
Parent: WindowServer [70]

Version: 2.0.1 (412.5)
Build Version: 7
Project Name: WebBrowser
Source Version: 4120500

PID: 382
Thread: 0

Exception: EXC_BAD_ACCESS (0x0001)
Codes: KERN_PROTECTION_FAILURE (0x0002) at 0x0000001c

Thread 0 Crashed:
0 ...romedia.Flash Player.plugin 0x054a2cc0
native_ShockwaveFlash_TCallFrame + 429396
1 ...romedia.Flash Player.plugin 0x054a2904
native_ShockwaveFlash_TCallFrame + 428440
2 ...romedia.Flash Player.plugin 0x054cfa2c
native_ShockwaveFlash_TCallFrame + 613056
3 ...romedia.Flash Player.plugin 0x054a4c5c
native_ShockwaveFlash_TCallFrame + 437488
4 ...romedia.Flash Player.plugin 0x053b749c Flash_EnforceLocalSecurity +
872036
5 ...romedia.Flash Player.plugin 0x052ec9e0 Flash_EnforceLocalSecurity +
41896
6 ...romedia.Flash Player.plugin 0x052e43c0 Flash_EnforceLocalSecurity +
7560
7 com.apple.WebKit 0x959069e8 -[WebBaseNetscapePluginView
sendEvent:] + 280
8 com.apple.WebKit 0x959087b8 -[WebBaseNetscapePluginView
sendNullEvent] + 144
9 com.apple.Foundation 0x92885cfc __NSFireTimer + 116
10 com.apple.CoreFoundation 0x9075ed10 __CFRunLoopDoTimer + 184
...

i have quite the same problem:
isolated my upload class works properly, but when implemented in my
application it causes crash on IE & Firefox, WinXp.
IE closes immediately. Firefox runs about 30 seconds, then it becomes very
slow (actually the swf's framerate drops from 28 to 1) so i have to close it
via taskmanager.

I have encountered quite the same problem. After repeatedly upload files, the
browser crashes - this happens in both Mozilla Firefox as well Opera. Uploading
200 files at a time is not a problem, but 10 times uploading 20 files in a
session will result in a crash.

The problem does not seem to be memory related (the problem occurs both in
very large as well very small flash movies), nor is it related to software
running on my PC, since it seems to be happening at other PC's as well. It
happens both at the filereference as well the filereferencelist.

Another nice thing I came around is the following try:
- create a new flash movie,
- add a button to it with the following action:
on(release) {
getURL(http://labs.oinam.com/flash8/fileuploaddownload,"_blank");
}
- open the flash movie and click the link in a browser that uses tabs, such as
FireFox 1.5 or Opera
- in the window that opens, upload 2 or more images, and close that tab
- you'll notice that the original movie (the one containing the button you've
created) is now suddenly unloaded, showing only the message "movie not loaded"
upon right-click.

This happens with serval flash 8 file upload examples that can be found on the
net (the link provided is just an example). The behavior does not occur when
uploading only a single file.



Since I was building a web-application quite depending on file upload, I have
used the following workaround for the crashes:
- I have placed all code related to file uploads in a separate fileupload.swf
file
- I load this SWF when necessary (using loadMovieNum)
- On each new upload, I load this upload-movie again, in a new level (I run a
counter to increase the level to load the fileupload.swf in for each step)
- I do not unload the previous loaded fileupload.swf files, they simply remain
unused but loaded in levels in the flash movie. This is bad for memory, but
seems to be necessary because the crashes otherwise just continue to happen.



I am still investigating, so above assumptions are not necessarily 100% true,
but this were my observations so far.

greetz,
Barend