Problems with Page.source_filename, Page.iter_source_filenames with alternatives are enabled

[dairiki]

In playing around to look at another issue, I noticed that when building the example project, the blog pages are always rebuilt, even if they have not been changed.

I.e.

$ cd example
$ lektor build
Started build
U de/index.html
[... lots of other pages ...]
U static/css/nix.css
Finished build in 1.07 sec
Started prune
Finished prune in 0.00 sec

Then, do it again. Since no source has been changed, nothing should be rebuilt. But:

$ lektor build
Started build
U de/blog/blog-post-5/index.html
U de/blog/blog-post-4/index.html
U de/blog/blog-post-3/index.html
U de/blog/blog-post-2/index.html
U de/blog/first-post/index.html
U blog/blog-post-5/index.html
U blog/blog-post-4/index.html
U blog/blog-post-3/index.html
U blog/blog-post-2/index.html
U blog/first-post/index.html
U fr/blog/blog-post-5/index.html
U fr/blog/blog-post-4/index.html
U fr/blog/blog-post-3/index.html
U fr/blog/blog-post-2/index.html
U fr/blog/first-post/index.html
Finished build in 0.32 sec
Started prune
Finished prune in 0.01 sec

The blog posts are always rebuilt.

Issue(s) Resolved

I believe the root cause of this is that Page.source_filename is, in some cases, returning bogus values — specifically
the name of non-existing files. E.g. if only a contents.lr exists, .source_filename still returns contents+de.lr for the alt="de" version of the page. Since this is used for dependency tracking, things break if this points to a non-existing file.

Looking further, I think that SourceObject.source_filename is, or should be deprecated in favor of SourceObject.iter_source_filenames. Again, since this is used for dependency tracking, we usually (always?) care about all the source files, not just the "primary" one.

Related Issues / Links

Not sure.

Description of Changes

• Wrote at least one-line docstrings (for any new functions)
• Added unit test(s) covering the changes (if testable)
• Link to corresponding documentation pull request for getlektor.com

[dairiki]

Some of the assumptions I made in this PR are incorrect.

Primarily: I now believe that SourceObject.iter_source_filenames should list all files which are checked for when reading the source — whether or not each of those files actually exists. When I wrote this PR, I was thinking that iter_source_filenames should list only those files that actually exist.

Closing in favor of #1007