View Issue Details
|ID||Project||Category||View Status||Date Submitted||Last Update|
|0002519||JEDI VCL||00 JVCL Components||public||2005-01-17 06:40||2005-01-24 04:36|
|Product Version||3.00 BETA 2|
|Target Version||Fixed in Version||3.00|
|Summary||0002519: JvHttpUrlGrabber freezes application|
|Description||I run into trouble when I use JvHttpUrlGrabber in a chain.|
Like following. I think itÂ´s a problem with threads - But how should I do
procedure TForm1.JvHttpUrlGrabber1DoneFile(Sender: TObject;
FileName: String; FileSize: Integer; Url: String);
If fileexists(filename) then
.... open the file and read some infos and then download second file
The application freezes.
I think it is because, the "first" thread of the grabber is not finished
until it is still in procedure JvHttpUrlGrabber1DoneFile.
|Tags||No tags attached.|
Your assumption is right. Could you try this:
In JvUrlListGrabber.pas, at line 878, replace this
Then rebuild the JVCL packages and your application.
Note that this introduces a memory leak (the thread is never freed) and as such must not be put into production code. But once I know that this works, I'll be able to further work on this to get rid of the memory leak.
no I think this is not the right way. It should not be possible to start same grabber in that time. For this way we should use an additional grabber.
Therefore the grabber shouldnÂ´t start again, before "old" grabber thread is ended. I also run in troubles, when I watch "status" of the grabber until is not set to gsstopped, like this:
grabber1.start; //download file1
while grabber1.status=gsstopped do
grabber1.start; //download file2
Run into exceptions....
At this point I would prefer to ask if thread is terminated to if status is gsstopped
But when you are in the OnDoneFile event, the thread is NOT terminated yet. It is still running, but halted to be synchronized with the main thread. Now, if you call Start, it will try to kill that thread, and hence be blocked because it's inside itself.
My change ensures that there is no blocking if you call start from an event. But that change is not clean.
What it should actually do is to call Stop before starting a new one. I'll investigate that.
As to your code, it is very dangerous to have a loop that calls ProcessMessage like that. You should use the chaining of events.
In the end, you, as the user, don't need to know there is a thread. This is just an implementation detail that should have no impact on your code. Basically, I could decide that the thread is stopped much later after the end of the download.
I'll investigate that and will do further testing.
||Isn't it possible to use the urlgrabber without a thread, and then make the thread handling outside manually?|
For my usage it is a very easy and smoothy way to have a httpgrabber that do not slowdown my application until it grabs something from internet.
I really like this component. It is one of my favorits.
(that anonymous post was from me)
Jens: No it is not even an option. It is being used by the TJvUrlListGrabber component which in essence is asynchronous.
I won't change that.
What I will fix though is the way the thread is used so that calling Start on an already started thread from one its own events doesn't freeze the application.
Not too hard, I just need some time to implement it.
||@Olievier:I didn't want to remove the threads. But what about an property to work without thread!|
One more bug:
Stopping a JvHttpUrlGrabber through method stop when grabber isnÂ´t started causes an AV!
But I donÂ´t know why.
> @Olivier:I didn't want to remove the threads. But what about an
> property to work without thread!
But then, how do you make it asynchronous when the API calls are synchronous?
> Stopping a JvHttpUrlGrabber through method stop when grabber
> isnÂ´t started causes an AV!
> But I donÂ´t know why.
That one was easy, it's as stupid bug because the thread is not created yet. This is now fixed in CVS.
Further, the chaining problem is fixed as well, and the fix does not introduce any memory or resource leak AFAIK.
Please use the latest version from CVS and let us know how it goes
> > @Olivier:I didn't want to remove the threads. But what about an
> > property to work without thread!
> But then, how do you make it asynchronous when the API calls are synchronous?
This exactly what i want. I want to use it synchronous. For example in the TJvProgramVersionCheck i already have implemented a thread for handling the download's and check's. In this case i need it synchronous.
> This exactly what i want. I want to use it synchronous. For example in
> the TJvProgramVersionCheck i already have implemented a thread for handling
> the download's and check's. In this case i need it synchronous.
I understand, but those components are not meant to be synchronous. If they were, they would be locking up the application for the duration of the download, which can be quite long.
Use the events, and you should be able to do what you want, but have the advantage to not freeze the application.
great work. It works now as far as I Â´ve tested good.
@Jens: What and where is the TJvProgramVersionCheck?
The class from Jens is under dev/Jens/JvProgramVersionCheck in CVS. Alternatively, you can see it online here:
|2005-01-17 06:40||Dierk||New Issue|
|2005-01-17 06:41||obones||Status||new => assigned|
|2005-01-17 06:41||obones||Assigned To||=> obones|
|2005-01-17 12:46||obones||Note Added: 0006188|
|2005-01-17 12:46||obones||Status||assigned => feedback|
|2005-01-18 00:15||Dierk||Note Added: 0006196|
|2005-01-18 00:50||anonymous||Note Added: 0006197|
|2005-01-18 01:19||jfudickar||Note Added: 0006198|
|2005-01-18 01:25||Dierk||Note Added: 0006199|
|2005-01-18 01:39||obones||Note Added: 0006200|
|2005-01-18 02:23||jfudickar||Note Added: 0006202|
|2005-01-19 08:30||anonymous||Note Added: 0006228|
|2005-01-22 05:16||obones||Note Added: 0006247|
|2005-01-24 00:24||jfudickar||Note Added: 0006262|
|2005-01-24 01:22||obones||Note Added: 0006263|
|2005-01-24 01:25||Dierk||Note Added: 0006265|
|2005-01-24 04:36||obones||Status||feedback => resolved|
|2005-01-24 04:36||obones||Fixed in Version||=> 3.00|
|2005-01-24 04:36||obones||Resolution||open => fixed|
|2005-01-24 04:36||obones||Note Added: 0006267|