diff mbox series

[meta-arago,master,3/4] meta-arago-test: Add curl-cffi and yt-dlp to ti-test-extras

Message ID 20260302205436.1117890-3-t-lannynd@ti.com
State Changes Requested
Delegated to: Ryan Eatmon
Headers show
Series [meta-arago,master,1/4] meta-arago-test: Add curl-cffi recipe | expand

Commit Message

Lannynd, Thorsten March 2, 2026, 8:54 p.m. UTC
Add curl-cffi and yt-dlp to the TI_TEST_EXTRAS packagegroup for additional
testing capabilities.

- curl-cffi: Python library for curl-impersonate functionality
- yt-dlp: Enhanced youtube-dl fork with additional features

Signed-off-by: Thorsten Lannynd <t-lannynd@ti.com>
---
 meta-arago-test/recipes-core/packagegroups/ti-test.bb | 2 ++
 1 file changed, 2 insertions(+)

Comments

Andrew Davis March 2, 2026, 10:55 p.m. UTC | #1
On 3/2/26 2:54 PM, Thorsten Lannynd via lists.yoctoproject.org wrote:
> Add curl-cffi and yt-dlp to the TI_TEST_EXTRAS packagegroup for additional
> testing capabilities.
> 
> - curl-cffi: Python library for curl-impersonate functionality
> - yt-dlp: Enhanced youtube-dl fork with additional features
> 

Didn't youtube-dl run into some legal issues.. Could you expand on what these
"additionaltesting capabilities" will be?

Andrew

> Signed-off-by: Thorsten Lannynd <t-lannynd@ti.com>
> ---
>   meta-arago-test/recipes-core/packagegroups/ti-test.bb | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/meta-arago-test/recipes-core/packagegroups/ti-test.bb b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
> index f2ae1aef..d1d049c2 100644
> --- a/meta-arago-test/recipes-core/packagegroups/ti-test.bb
> +++ b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
> @@ -77,10 +77,12 @@ TI_TEST_EXTRAS_OPENCL = " \
>   "
>   
>   TI_TEST_EXTRAS = " \
> +    curl-cffi \
>       python3-pillow \
>       pytesseract \
>       python3-numpy \
>       wayland-utils \
> +    yt-dlp \
>       ${@bb.utils.contains('DISTRO_FEATURES', 'opengl', '${TI_TEST_EXTRAS_OPENGL}', '', d)} \
>       ${@bb.utils.contains('DISTRO_FEATURES', 'opencl', '${TI_TEST_EXTRAS_OPENCL}', '', d)} \
>   "
Lannynd, Thorsten March 2, 2026, 11:49 p.m. UTC | #2
On 3/2/26 4:56 PM, Andrew Davis wrote:
> On 3/2/26 2:54 PM, Thorsten Lannynd via lists.yoctoproject.org wrote:
> > Add curl-cffi and yt-dlp to the TI_TEST_EXTRAS packagegroup for
> > additional testing capabilities.
> >
> > - curl-cffi: Python library for curl-impersonate functionality
> > - yt-dlp: Enhanced youtube-dl fork with additional features
> >
> 
> Didn't youtube-dl run into some legal issues.. Could you expand on what
> these "additionaltesting capabilities" will be?
> 
These packages enable automated Chromium testing to verify video streams display correctly. Our usage is minimal:

yt-dlp -g https://player.vimeo.com/video/253905163

This extracts only the direct CDN URL to Big Buck Bunny (CC-licensed) hosted on Vimeo—no downloading occurs via yt-dlp itself. We then use this URL to:

  - Extract specific frames at known timestamps as golden references
  - Validate hardware video decoding and Chromium rendering pipeline
  - curl-cffi provides the modern TLS/HTTP capabilities required by video hosting platforms

This approach eliminates the need to download, store, and/or host golden reference videos and allows testing with Vimeo directly. We can programmatically extract the exact frames needed for validation directly from the source, ensuring our hardware video decoding and Chromium rendering pipeline work correctly.

Thorsten
> Andrew
> 
> > Signed-off-by: Thorsten Lannynd <t-lannynd@ti.com>
> > ---
> >   meta-arago-test/recipes-core/packagegroups/ti-test.bb | 2 ++
> >   1 file changed, 2 insertions(+)
> >
> > diff --git a/meta-arago-test/recipes-core/packagegroups/ti-test.bb
> > b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
> > index f2ae1aef..d1d049c2 100644
> > --- a/meta-arago-test/recipes-core/packagegroups/ti-test.bb
> > +++ b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
> > @@ -77,10 +77,12 @@ TI_TEST_EXTRAS_OPENCL = " \
> >   "
> >
> >   TI_TEST_EXTRAS = " \
> > +    curl-cffi \
> >       python3-pillow \
> >       pytesseract \
> >       python3-numpy \
> >       wayland-utils \
> > +    yt-dlp \
> >       ${@bb.utils.contains('DISTRO_FEATURES', 'opengl',
> '${TI_TEST_EXTRAS_OPENGL}', '', d)} \
> >       ${@bb.utils.contains('DISTRO_FEATURES', 'opencl',
> '${TI_TEST_EXTRAS_OPENCL}', '', d)} \
> >   "
Andrew Davis March 3, 2026, 12:58 a.m. UTC | #3
On 3/2/26 5:49 PM, Lannynd, Thorsten wrote:
> On 3/2/26 4:56 PM, Andrew Davis wrote:
>> On 3/2/26 2:54 PM, Thorsten Lannynd via lists.yoctoproject.org wrote:
>>> Add curl-cffi and yt-dlp to the TI_TEST_EXTRAS packagegroup for
>>> additional testing capabilities.
>>>
>>> - curl-cffi: Python library for curl-impersonate functionality
>>> - yt-dlp: Enhanced youtube-dl fork with additional features
>>>
>>
>> Didn't youtube-dl run into some legal issues.. Could you expand on what
>> these "additionaltesting capabilities" will be?
>>
> These packages enable automated Chromium testing to verify video streams display correctly. Our usage is minimal:
> 
> yt-dlp -g https://player.vimeo.com/video/253905163
> 
> This extracts only the direct CDN URL to Big Buck Bunny (CC-licensed) hosted on Vimeo—no downloading occurs via yt-dlp itself. We then use this URL to:
> 
>    - Extract specific frames at known timestamps as golden references
>    - Validate hardware video decoding and Chromium rendering pipeline

A direct URL to a video file does that? This whole list reads like an
LLM generated response..

>    - curl-cffi provides the modern TLS/HTTP capabilities required by video hosting platforms

Yeah, I read the commit message too, this gives me no new information

> 
> This approach eliminates the need to download, store, and/or host golden reference videos and allows testing with Vimeo directly. We can programmatically extract the exact frames needed for validation directly from the source, ensuring our hardware video decoding and Chromium rendering pipeline work correctly.
> 

"hardware video decoding and Chromium rendering pipeline" LLM generated
summary of the above LLM generated list, don't waste my time.

I just went and asked Claude to write a script to validate my "hardware
video decoding and Chromium rendering pipeline". It spit out a script
with these lines near the top:

> import yt_dlp
> from curl_cffi import requests
> from selenium ..

So that would be the real answer I'm looking for, you're adding some
generated testing script that uses these libraries. You could just
say that or point me to the script itself.

Anyway, what is wrong with hosting some set of reference videos locally?
Are you planning on randomly picking videos to test against? Otherwise
having every DUT fetch the same videos from the external web every test
run is going to be a needless bandwidth hog.

Andrew

> Thorsten
>> Andrew
>>
>>> Signed-off-by: Thorsten Lannynd <t-lannynd@ti.com>
>>> ---
>>>    meta-arago-test/recipes-core/packagegroups/ti-test.bb | 2 ++
>>>    1 file changed, 2 insertions(+)
>>>
>>> diff --git a/meta-arago-test/recipes-core/packagegroups/ti-test.bb
>>> b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
>>> index f2ae1aef..d1d049c2 100644
>>> --- a/meta-arago-test/recipes-core/packagegroups/ti-test.bb
>>> +++ b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
>>> @@ -77,10 +77,12 @@ TI_TEST_EXTRAS_OPENCL = " \
>>>    "
>>>
>>>    TI_TEST_EXTRAS = " \
>>> +    curl-cffi \
>>>        python3-pillow \
>>>        pytesseract \
>>>        python3-numpy \
>>>        wayland-utils \
>>> +    yt-dlp \
>>>        ${@bb.utils.contains('DISTRO_FEATURES', 'opengl',
>> '${TI_TEST_EXTRAS_OPENGL}', '', d)} \
>>>        ${@bb.utils.contains('DISTRO_FEATURES', 'opencl',
>> '${TI_TEST_EXTRAS_OPENCL}', '', d)} \
>>>    "
>
Yogesh Hegde March 3, 2026, 9:13 a.m. UTC | #4
On 18:58-20260302, Andrew Davis via lists.yoctoproject.org wrote:
> On 3/2/26 5:49 PM, Lannynd, Thorsten wrote:
> > On 3/2/26 4:56 PM, Andrew Davis wrote:
> > > On 3/2/26 2:54 PM, Thorsten Lannynd via lists.yoctoproject.org wrote:
> > > > Add curl-cffi and yt-dlp to the TI_TEST_EXTRAS packagegroup for
> > > > additional testing capabilities.
> > > > 
> > > > - curl-cffi: Python library for curl-impersonate functionality
> > > > - yt-dlp: Enhanced youtube-dl fork with additional features
> > > > 
> > > 
> > > Didn't youtube-dl run into some legal issues.. Could you expand on what
> > > these "additionaltesting capabilities" will be?
> > > 
> > These packages enable automated Chromium testing to verify video streams display correctly. Our usage is minimal:
> > 
> > yt-dlp -g https://player.vimeo.com/video/253905163
> > 
> > This extracts only the direct CDN URL to Big Buck Bunny (CC-licensed) hosted on Vimeo—no downloading occurs via yt-dlp itself. We then use this URL to:
> > 
> >    - Extract specific frames at known timestamps as golden references
> >    - Validate hardware video decoding and Chromium rendering pipeline
> 
> A direct URL to a video file does that? This whole list reads like an
> LLM generated response..

Blender provides direct URL to BBB video files. See Links [0] & [1]. 
wget can fetch the video. 

I don't see a need for yt-dlp.

> 
> >    - curl-cffi provides the modern TLS/HTTP capabilities required by video hosting platforms
> 
> Yeah, I read the commit message too, this gives me no new information
> 
> > 
> > This approach eliminates the need to download, store, and/or host golden reference videos and allows testing with Vimeo directly. We can programmatically extract the exact frames needed for validation directly from the source, ensuring our hardware video decoding and Chromium rendering pipeline work correctly.
> > 
> 
> "hardware video decoding and Chromium rendering pipeline" LLM generated
> summary of the above LLM generated list, don't waste my time.
> 
> I just went and asked Claude to write a script to validate my "hardware
> video decoding and Chromium rendering pipeline". It spit out a script
> with these lines near the top:
> 
> > import yt_dlp
> > from curl_cffi import requests
> > from selenium ..
> 
> So that would be the real answer I'm looking for, you're adding some
> generated testing script that uses these libraries. You could just
> say that or point me to the script itself.
> 
> Anyway, what is wrong with hosting some set of reference videos locally?
> Are you planning on randomly picking videos to test against? Otherwise
> having every DUT fetch the same videos from the external web every test
> run is going to be a needless bandwidth hog.
> 

Agree with Andrew here. Additionally, I have these concerns with
downloading videos on every DUT/Test 
1. Security risk of these videos containing some malware, since the list
is public and threat actors know the exact URL from where we are
downloading and how we are downloading. 
2. Licensing risk of using yt-dlp.

If you plan to download the videos then, How are you planning to mitigate these?

NAK

[0]: https://download.blender.org/peach/bigbuckbunny_movies/
[1]: https://download.blender.org/demo/movies/BBB/

Regards 
Yogesh
diff mbox series

Patch

diff --git a/meta-arago-test/recipes-core/packagegroups/ti-test.bb b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
index f2ae1aef..d1d049c2 100644
--- a/meta-arago-test/recipes-core/packagegroups/ti-test.bb
+++ b/meta-arago-test/recipes-core/packagegroups/ti-test.bb
@@ -77,10 +77,12 @@  TI_TEST_EXTRAS_OPENCL = " \
 "
 
 TI_TEST_EXTRAS = " \
+    curl-cffi \
     python3-pillow \
     pytesseract \
     python3-numpy \
     wayland-utils \
+    yt-dlp \
     ${@bb.utils.contains('DISTRO_FEATURES', 'opengl', '${TI_TEST_EXTRAS_OPENGL}', '', d)} \
     ${@bb.utils.contains('DISTRO_FEATURES', 'opencl', '${TI_TEST_EXTRAS_OPENCL}', '', d)} \
 "