This issue is still consistenly reproducible on current git version of Gnash
when download simulator is in use.

I have made test SWF files (using libming's MakeSWF), one loader that tries
to download another Flash file from a custom download simulation server
which introduces brief pause on specific positions in the stream:

My loaded SWF in this case is 37288 bytes long. Server output lines begin
with `htcheckpoint:`, loader SWF's output lines begin with `testloader:`.

Following is a trace log from Flash Player (interleaved with server output):

testloader: starting download
htcheckpoint: request received, pausing 2000 ms
htcheckpoint: sent 0 byte(s), pausing 2000 ms
testloader: download started
htcheckpoint: sent 1 byte(s), pausing 2000 ms
Warning: Not a known player download type, http://localhost:8085/testpattern.swf
testloader: downloading: 0/0
htcheckpoint: sent 1621 byte(s), pausing 2000 ms
testloader: downloading: 1621/37288
htcheckpoint: sent 9322 byte(s), pausing 2000 ms
testloader: downloading: 5717/37288
testloader: downloading: 9322/37288
htcheckpoint: sent 18644 byte(s), pausing 2000 ms
testloader: downloading: 17514/37288
testloader: downloading: 18644/37288
htcheckpoint: sent 27966 byte(s), pausing 2000 ms
testloader: downloading: 26836/37288
testloader: downloading: 27966/37288
htcheckpoint: sent 37287 byte(s), pausing 2000 ms
testloader: downloading: 32062/37288
testloader: downloading: 36158/37288
testloader: downloading: 37287/37288
htcheckpoint: sent 37288 byte(s), pausing 2000 ms
testloader: downloading: 37288/37288
testloader: download completed: status 200
testloader: started action

Following is a trace log from Gnash (interleaved with server output):

126 TRACE: testloader: starting download
htcheckpoint: request received, pausing 2000 ms
htcheckpoint: sent 0 byte(s), pausing 2000 ms
htcheckpoint: sent 1 byte(s), pausing 2000 ms
htcheckpoint: sent 1621 byte(s), pausing 2000 ms
6146 TRACE: testloader: download started
6147 TRACE: testloader: downloading: 20/37288
6147 TRACE: testloader: download completed: status 0
6147 TRACE: testloader: started action
htcheckpoint: sent 9322 byte(s), pausing 2000 ms
htcheckpoint: sent 18644 byte(s), pausing 2000 ms
htcheckpoint: sent 27966 byte(s), pausing 2000 ms
htcheckpoint: sent 37287 byte(s), pausing 2000 ms
htcheckpoint: sent 37288 byte(s), pausing 2000 ms

You'd see that on Gnash...

• `OnLoadStart()` was called when first frame is completed rather than when HTTP response was received.
• `OnLoadProgress()` was only called only one time when first frame is completed, rather than every time a block of data is received. Reported number of bytes downloaded was also inaccurate (20 instead of 1621).
• `OnLoadComplete()` was called even when the download was not yet completed. HTTP response code reported to the callback was also incorrect (0 instead of 200).

Test SWFs, source code, download simulator script, and full debug logs
are attached as `movieloader-progresscheck.zip`.

Gnash: 0.8.11dev (git 15d61a8 26-Dec-2015)
Flash Player: 11.2 r202 (11.2.202.491) Standalone
System: Debian GNU/Linux 7.0 Wheezy i386

P.S. The download simulation server runs on TCP port 8085 by default;
this it can be changed by adding `-p PORTNUMBER` option to server's
command line. URL in loader SWF must be adjusted accordingly if the port
is changed.


(file #35912)

Sandro,

I load a 6.7MB SWF hosted at http://archive.org with an average download speed of 300KB/s. It takes more than 30 seconds to be loaded. And the whole script works with the Adobe player.

You are the expert here and I don’t know how onLoadComplete is handled internally by Gnash. But it works as expected.

The script attached (still a manual testcase) uses onLoadStart, onLoadError, onLoadProgress and onLoadComplete.

I’m not sure whether onLoadStart works fine. And onLoadProgress doesn’t performs it contents.

onLoadError works fine (at least with the script attached).

And onLoadComplete works fine, since lines 106 to 110 in the attached script are only executed once the external SWF has been completely loaded.

I don’t know what I’m missing here. But if onLoadComplete is executed after the SWF header is read, wouldn’t it be possible that the SWF header can be read only after complete loading?

Many thanks for your help.

(file #29551)

Sandro,

according to the Adobe documentation (http://help.adobe.com/en_US/AS2LCR/Flash_10.0/help.html?content=00001383.html) onLoadComplete is ”[i]nvoked when a file that was loaded with MovieClipLoader.loadClip() is completely downloaded.” This part seems to work fine in Gnash.

onLoadInit is “[i]nvoked when the actions on the first frame of the loaded clip have been executed” (http://help.adobe.com/en_US/AS2LCR/Flash_10.0/help.html?content=00001385.html),

onLoadStart is “[i]nvoked when a call to MovieClipLoader.loadClip() has begun to download a file” (http://help.adobe.com/en_US/AS2LCR/Flash_10.0/help.html?content=00001387.html).

onLoadError works also fine (at least with the script I provided).

And onLoadProgress is “[i]nvoked every time the loading content is written to the hard disk during the loading process (that is, between MovieClipLoader.onLoadStart and MovieClipLoader.onLoadComplete)” (http://help.adobe.com/en_US/AS2LCR/Flash_10.0/help.html?content=00001386.html).

I hope it helps,


Pablo

Besides, if you read the FIXME lines in the C++ code you see that they also request to research about semantics of onLoadComplete argument.

In particular it's not clear to me if onLoadComplete should be called when the whole resource was loaded or only after the first bytes arrive.

Sandro,

sorry for having expressed myself wrong: I haven’t reported the sound issue yet. (I have to clear other issues before.)

I cannot code and C++ is Greek to me. I can read basic ActionScript (understanding is something different ;-)) and I can write very simple ActionScript. So, I see which FIXME you mean in libcore/MovieLoader.cpp, but not more than that.

I have tried to add checks to the script I provided yesterday, In order to understand how to do it, I wanted to compile testsuite/misc-swfc.all/sound.sc, but I found two issues:

First, on line 8 there is a C++ include directive. As far as I know this cannot be complied with swfc.

Second, as far as I know (I can check this back with the mailing list), line 11 should read
    .sound audio "sound1.wav"
instead of
    .sound audio MEDIA(sound1.wav)"

I guess I have to be missing something here.

With all these limits, how can I help best to the bug resolution?

Many thanks for your help,


Pablo

Note that onLoadProgress should be implemented, somewhat tested in testsuite/actionscript.all/MovieClipLoader.as

Sandro,

attached you have a test case.

listen_loading.onLoadProgress seems to be the part not working.

I hope it helps,

Pablo
          



(file #29533)

Sandro, I am the author of both SWF files.

How show I make the test case? Do I have to meet any standards specified somewhere?

Many thanks for your help.

FYI: the logged ERROR was downgraded to a DEBUG with http://git.savannah.gnu.org/cgit/gnash.git/commit/?id=1606ab48c70dcdda3ca95ff5212b836418b01c3d