hugolib: Fix "borrow content from another language" issue

If a content file contains shortcode(s), we have logic in place to re-render it per output format.

We also have logic in place that avoids making a copy of the content used for this process if we don't need it.

This was before this commit limited to server mode and if the page should be output to multiple formats.

But there is a third case: If a site (language) borrows and renders `.Content` from another language. This would, before this commit, behave oddly for content with shortcodes.

Fixes #4986

diff --git a/hugolib/page.go b/hugolib/page.go
index f45257c9d32b3e674c445e6cfa0a38e5b595baed..c00070c9aeab7f79bfda5a522736b48b94ee0234 100644 (file)
--- a/hugolib/page.go
+++ b/hugolib/page.go
@@ -1249,10 +1249,10 @@ func (p *Page) prepareForRender() error {
        // or a template or similar has changed so wee need to do a rerendering
        // of the shortcodes etc.
 
-       // If in watch mode or if we have multiple output formats,
+       // If in watch mode or if we have multiple sites or output formats,
        // we need to keep the original so we can
        // potentially repeat this process on rebuild.
-       needsACopy := s.running() || len(p.outputFormats) > 1
+       needsACopy := s.running() || len(s.owner.Sites) > 1 || len(p.outputFormats) > 1
        var workContentCopy []byte
        if needsACopy {
                workContentCopy = make([]byte, len(p.workContent))