Bug 198376 – [GStreamer] Cannot play Bert's Bytes radio stream from http://radio.dos.nl/

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Alberto Garcia]

1. Go to http://radio.dos.nl/
2. Click on the menu button on the lower right corner
3. Select "Jazz"
4. Click on the "Bert's Bytes" icon.

This should play this radio stream, but it does nothing with WebKitGTK 2.24.2.

It reportedly works fine with 2.24.0 and fails with 2.24.1. It is likely related to bug 197410.

This bug was originally reported in Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=929749

[Alberto Garcia]

(In reply to Alberto Garcia from comment #0)
> 2. Click on the menu button on the lower right corner

Sorry, I meant the lower *LEFT* corner.

[Philippe Normand]

That's not a live stream though, http://www.bertvandenbrink.com/bertsbytes/concert.mp3

[Richard Lucassen]

No, it's not a live stream, but 2.24.0 plays it and 2.24.1 does not.

Richard

[Philippe Normand]

Actually it might if you wait long enough :D
It's an issue with the on-disk buffering, I started a patch.

[Philippe Normand]

Created attachment 373217 [details]
Patch

[Xabier Rodríguez Calvar]

Comment on attachment 373217 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=373217&action=review

> Source/WebCore/ChangeLog:16
> +        No new tests, existing media tests cover this patch.

Why is no test turning green associated with the bug then?

> Source/WebCore/platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:1474
> +    auto updateMaxTimeLoaded = [&]() {

Sometimes it is hard to choose when to use a lambda like this or just another regular object method. My preference in this case would be the latter.

> Source/WebCore/platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:1508
> +    case GST_BUFFERING_TIMESHIFT:
> +    case GST_BUFFERING_LIVE:

maybe just "default:" ?

[Xabier Rodríguez Calvar]

Comment on attachment 373217 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=373217&action=review

>> Source/WebCore/ChangeLog:16
>> +        No new tests, existing media tests cover this patch.
> 
> Why is no test turning green associated with the bug then?

There is no test and there should be. I know we are in a hurry for the release, so let's land it and open a bug for the test.

[Aakash Jain]

I am testing gtk queue on new EWS and it seems to fail for this patch.

Can someone look if EWS caught a real issue or it's a false positive?

https://ews-build.webkit-uat.org/#/builders/7/builds/281
Retry: https://ews-build.webkit-uat.org/#/builders/7/builds/282

[Philippe Normand]

(In reply to Aakash Jain from comment #8)
> I am testing gtk queue on new EWS and it seems to fail for this patch.
> 
> Can someone look if EWS caught a real issue or it's a false positive?
> 
> https://ews-build.webkit-uat.org/#/builders/7/builds/281
> Retry: https://ews-build.webkit-uat.org/#/builders/7/builds/282

I'll check it, thanks for the heads-up!

[Philippe Normand]

It's a WPE issue, not GTK. I forgot an include:

GLibUtilities.h:38:15: error: ‘GUniquePtr’ does not name a type; did you mean ‘GUniquePtr_h’?

[Aakash Jain]

(In reply to Philippe Normand from comment #10)
> It's a WPE issue, not GTK. 
Yeah, I confused that. Thanks for the correction.

[Philippe Normand]

Created attachment 373236 [details]
Patch

[Philippe Normand]

Comment on attachment 373236 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=373236&action=review

> Source/WTF/wtf/glib/GUniquePtr.h:27
> +#include <memory.h>

That was the EWS issue I think. Let's see if EWS likes it now.

[Philippe Normand]

Still red, I'll check it :)

[Philippe Normand]

Created attachment 373239 [details]
Patch

[Philippe Normand]

One of those weird issues with Forwarding headers and unified builds :(

[Philippe Normand]

Created attachment 373242 [details]
Patch

[Philippe Normand]

Committed r247010: <https://trac.webkit.org/changeset/247010>