Page MenuHomePhabricator

Miniature images from og:image not loading in social media links
Closed, ResolvedPublicBUG REPORT

Assigned To
Authored By
Theklan
Mar 6 2024, 3:57 PM
Referenced Files
F42579606: irudia.png
Mar 12 2024, 6:57 PM
F42578917: irudia.png
Mar 12 2024, 6:16 PM
F42578910: irudia.png
Mar 12 2024, 6:16 PM
F42418542: irudia.png
Mar 6 2024, 3:57 PM

Description

Steps to replicate the issue (include links if applicable):

  • Open buffer.com social media management system
  • Add any link from Wikimedia

What happens?: When trying to schedule social media messages using Buffer, Wikimedia articles are not loading miniature images in the last weeks, but they used to do it before.

irudia.png (588×898 px, 66 KB)

What should have happened instead?: They should load the images from the metadata of the page. They have been doing this for months, but in the last couple of weeks it has failed every time. It doesn't fail with other pages that are not part of Wikimedia.

Software version (skip for WMF-hosted wikis like Wikipedia):

Other information (browser name/version, screenshots, etc.):

Event Timeline

I don't know where this code comes from but I see in the source code of an article this:

<meta property="og:image" content="https://upload.wikimedia.org/wikipedia/commons/9/9a/3.7_inch_AA_gun_and_crew_Fremantle_Nov_1943_AWM_060199.jpg">
<meta property="og:image:width" content="1200">
<meta property="og:image:height" content="909">
<meta property="og:image" content="https://upload.wikimedia.org/wikipedia/commons/9/9a/3.7_inch_AA_gun_and_crew_Fremantle_Nov_1943_AWM_060199.jpg">
<meta property="og:image:width" content="800">
<meta property="og:image:height" content="606">
<meta property="og:image:width" content="640">
<meta property="og:image:height" content="485">

I assume that an <meta property="og:image" content="https://upload.wikimedia.org/wikipedia/commons/9/9a/3.7_inch_AA_gun_and_crew_Fremantle_Nov_1943_AWM_060199.jpg"> is missing for the small one, but I don't know if this is related.

I don't think it is related, because I have checked the one in the example provided and it is there:

<meta property="og:image" content="https://upload.wikimedia.org/wikipedia/commons/thumb/5/59/Insausti_jauregia_Azkoitia_2006-12-29.JPG/1200px-Insausti_jauregia_Azkoitia_2006-12-29.JPG">
<meta property="og:image:width" content="1200">
<meta property="og:image:height" content="900">
<meta property="og:image" content="https://upload.wikimedia.org/wikipedia/commons/thumb/5/59/Insausti_jauregia_Azkoitia_2006-12-29.JPG/800px-Insausti_jauregia_Azkoitia_2006-12-29.JPG">
<meta property="og:image:width" content="800">
<meta property="og:image:height" content="600">
<meta property="og:image" content="https://upload.wikimedia.org/wikipedia/commons/thumb/5/59/Insausti_jauregia_Azkoitia_2006-12-29.JPG/640px-Insausti_jauregia_Azkoitia_2006-12-29.JPG">
<meta property="og:image:width" content="640">
<meta property="og:image:height" content="480">
Jdlrobson subscribed.

I'm not sure why this is tagged as web team. It sounds like an issue with buffer.com not being able to handle multiple og:image tags?

There is nothing wrong our side.

It worked till three weeks ago, showing all the images available (small, medium, large). I don't know if there has been any change in the code, that's why I opened the issue.

Anyway, having more than one og:image is not a standard protocol: https://ogp.me/

I have checked and the same error is happening when trying to automatically load an image for Facebook post.

irudia.png (422×608 px, 16 KB)

Also for Twitter post:

irudia.png (627×792 px, 47 KB)

Theklan renamed this task from Miniature images not loading in buffer to Miniature images from og:image not loading in social media links.Mar 12 2024, 6:17 PM

Anyway, having more than one og:image is not a standard protocol: https://ogp.me/

Yes it is. See the section named Arrays

Yes it is. See the section named Arrays

Sorry, you are right! Anyway, images are not loading, which should be solved.

Is this only the eu.wp links ? Or all language versions ? What have you tried ?

I tried eu, en, and de. The result is the same, so I think it is global

irudia.png (279×742 px, 24 KB)

Can confirm there is definitely something wrong.

If i remember correctly our metadata for this is added by PageImages, but I don't immediately see a cause there. It seems like it is always adding the full size image... and indeed the last item is missing...

The metadata seems correct, or not different from any other website I have visited. As an working hypothesis... could it be something related on Commons serving raw images?

The FB sharing debugger notes:

Provided og:image URL, https://upload.wikimedia.org/wikipedia/commons/7/75/Anne_Frank_passport_photo%2C_May_1942.jpg could not be processed as an image because it has an invalid content type.

I wonder if we are blocking AWS/Azure/GCP or something, and this is the fallout.

In the last couple of weeks it has failed every time. It doesn't fail with other pages that are not part of Wikimedia.

Hmm, for me facebook.com, FB ogp debugger, whatsapp and Twitter are not giving previews. Discord, LinkedIn and Telegram work.
Both Twitter and Facebook use AWS. The others GCP and Azure, if i remember correctly. Traffic did we recently block some more AWS traffic ?

I assume there's no news on this, but the problem is still there, as I have experienced today, so isn't something that would be gone automagically.

I am checking with traffic whether any crawling APIs from these services are being rate limited or blocked.

Anyway, having more than one og:image is not a standard protocol: https://ogp.me/

This is standard. It is mentioned in the Arrays section on that page.

I'm seing that in the last hour this has changed. I don't know if any change has been done somewhere, but now it is possible to choose images from the link. I will wait some days to claim it solved.

Hello,

This has been fixed with this patch, which I forgot to associate with this bug.

This traffic was originally blocked in T358455, in which Meta was using this same facebookexternalhit user-agent in an attempt to scrape all of the files in Commons, causing one of our network links to saturate and thus impacting reliability and performance for many other users.

CDanis claimed this task.

Thanks! I assume they were trying to do evil!