Hi all, heads up that one of the recent changes in the new PDF export experience has broken dynamic content macros across two of our Confluence apps in 4 of our 8 test environments and has been reported on several customer sites.
It appears that sites with the new engine skip embedding content macros (servers never receive the export request). The MS Word export works fine.
This is breaking for our signature app customers as they use this to display signatures within exported pages.
Atlassian ECO support has raised a bug: https://jira.atlassian.com/browse/CONFCLOUD-80251 but so far have not published an incident. ECOHELP-52585
If you have dynamic content macros in your apps recommend testing your PDF exports.
Chris Cairns
Digital Rose
PS. And numerous customers have raised concerns and problems with this “new” PDF export: See https://jira.atlassian.com/browse/CONFCLOUD-76035. It looks like this change was released before being fully baked.
3 Likes
I don’t know how it works now, but really the PDF export feature should just load the page in a headless browser and take a screenshot.
Or if text search is necessary then do that individually for each iframe on the page and replace in PDF with a screenshot.
Thanks @nathanwaters . We use the Macro render mode to support exports (https://developer.atlassian.com/cloud/confluence/modules/macro-render-modes/). Instead of the headless browser Confluence will call our servers and ask for an XHTML rendering of the macro in storage format.
At least that is how it used to work. 
1 Like
Update: The Atlassian dev team found and rolled back the change the caused this problem. PDF Exports are working again.
Thank you Atlassian
3 Likes
Thanks for keeping us informed @Chris_at_DigitalRose. 
Update: The Atlassian dev team appears to have added the same code 3 months later and re-introduced the same problem. We are again getting reports and have confirmed on our test environments that Confluence PDF exports in some sites are ignoring the manifest render mode AGAIN.
Raised new ECOHELP-61607.
2 Likes
I’ve been trying to remind them that this is still an issue too: https://jira.atlassian.com/browse/CONFCLOUD-78447
Might be limited to the UI Kit 2 component and src processing. But still a bug.
As you mentioned above, if the PDF renderer would just use a webkit headless browser (that allows javascript to run) a lot of these issues would go away.