[TIMOB-1042] Regression in HTTPClient/1.3

would it be possible to get the URL to your gzip for testing? If not, that's fine - just give me more details about the size so i can try and re-create. thanks

Here's the code to reproduce it:

   var debug = Ti.API.debug;
   
   var file = Titanium.Filesystem.getFile(
       Titanium.Filesystem.applicationSupportDirectory + "/tmp.file");
   
   // work around the bug where onload events are sometimes not delivered
   var utils = {};
   utils.clearXhrTimer = function () {
       if (utils.currentXhrTimer) {
           clearInterval(utils.currentXhrTimer);
           utils.currentXhrTimer = null;
       }
   };
   
   // a 6MB file to demonstrate the problem
   var url = "http://ankisrs.net/tmp/";
   
   var xhr;
   function downloadFile() {
       xhr = Titanium.Network.createHTTPClient();
       xhr.setTimeout(60000);
       xhr.onerror = function(e) {
           utils.clearXhrTimer();
           alert("Error: " + e.error);
       };
       xhr.onload = function() {
           debug(xhr.status);
           try {
               utils.clearXhrTimer();
               if (xhr.status == 200) {
                   if (file.exists()) {
                       file.deleteFile();
                   }
                   file.write(this.responseData);
                   downloadFile();
               }
           } catch (err) {
               alert (err);
           }
       };
       xhr.ondatastream = function (e) {
           debug(e.progress);
       };
       utils.clearXhrTimer();
       utils.currentXhrTimer = setInterval(
           function () {}, 1000);
       xhr.open('POST', url);
       xhr.send();
   }
   
   downloadFile();
   

(it's actually 40MB of random data zipped this time, but same problem)

OK, i'll take a look at this ASAP.

tried 5-6 times to reproduce ... seems to work all the time. no crashes...

can you possibly try building and testing from head?

Compiled from head - at first I thought I'd done something wrong as you've made the build process much faster! I'm assuming the version # is supposed to be 1.3.1?

Unfortunately the problem is still there. I discovered that I'd set the delay to 2 seconds during testing. I was still able to reproduce it with that short delay, but I've set it back to 20 seconds now in case it's a timing issue. Could you give it another go?

I also tried creating a new project and copying the code above into app.js (deleting the default stuff), and still got the problem.

Did some more testing. I can reproduce this every time when running from the simulator, but was unable to reproduce this even one when running on a device. If I comment out the ondatastream function, the error doesn't occur either. I guess it's a race condition?

I've been trying to squeeze a bit more performance out of HTTPClient - download speeds are about 3x slower on a device than in the simulator, both running of the wireless network. Perhaps I'm just being unrealistic about what a 2nd gen touch can do. I have a 1st gen touch which is much slower again, but it's been acting funny recently, so I'm not sure if its poor performance is due to a hardware fault or not.

Anyway, I wrote a patch to reduce the frequency the progress handlers were called, in the hope that would speed things up. I ended up reverting it because it had a negligible impact on download speed, but one nice side effect it had was to stop this problem from popping up. I mention it in case that makes it any easier to debug. OTOH, since this doesn't seem to appear outside the simulator, it's not a big deal.

http://github.com/dae/titanium_mobile/commit/f6a50354b4f6987d9035762837b9648730bedb28"> http://github.com/dae/titanium_mobile/commit/f6a50354b4f6987d903576...

can you try with the latest from master?
i've updated to the latest ASIHTTPRequest which seems to have some performance improvements.

The latest ASI has fixed the download issue I was seeing in the emulator (and I saw it happen once on a real device, but only once), so you can mark this issue as closed now. Thanks for the fix!