Invalid scheme format: '; dark' in /usr/src/app/_layouts/default.html #305
Comments
This was referenced Jan 31, 2025
Merged
vincerubinetti
added a commit
that referenced
this issue
Feb 3, 2025
Closes #305 This explanation is a bit hand-wavy, because I still can't follow the logic exactly. I find it hard to debug the flow of things in Jekyll because, while there's a `--trace` flag, the trace seems to be just the under-the-hood Ruby code, and not e.g. "you used this `_include` in this markdown file on this line, which then used this other `_include`, which then ran this Ruby filter, etc.". But in general: Certain nested content (I think within the `post-excerpt` hidden `data-search` attribute) ended up making its way to `relative_url` in `content.html` (which should just be a url), but only in certain rare cases where the peculiarity of the regex matching in the `section` component didn't prevent it. - change regex matching of `section` params in `content.html` to non-greedy, as they should've always been - change string pattern of passing `section` params to `content.html` from `param: value;` to `<param>value</param>`. this should reduce the chances of incorrect parsing, because one of the params is a URL, which can have `;` characters in them, which would confuse the old format. - fix unrelated bug where home page tab title looks like `| Lab Website Template` when it should just be `Lab Website Template` because the page title is `""` - strip all html from `post-excerpt` `data-search` attribute, as it always should've been. is part of the main bug fix of this PR, but also will reduce false-positive search matches, because without it, `data-search` contains e.g. single `p` characters leftover from `<p>` tags. now `data-search` should truly only be the content of the post. While not fully understanding the full flow of the root issue here, I think I've added enough safety mechanisms in enough spots to prevent similar issues from occurring in the future. New template version checklist: - [x] I have updated CITATION and CHANGELOG as appropriate. - [x] I have updated lab-website-template-docs as appropriate. - [x] I have checked the testbed as appropriate.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Origination: greenelab/greenelab.com#139
Full error:
Starting from the current template version, to replicate this bug, add a
sectionbreak component to a blog post, and remove any section breaks in the mainblog/index.mdpage.Then, from that starting point, do any one of the following to make the bug go away again:
sectionfrom blog postsectionto main blog pagesectionfrom blog post and main blog pagecontent=d.contentfromlist.htmldata-search="{{ search }}"frompost-excerpt.htmlpost-excerpt.htmlon main blog page (withlookupparam) instead oflistcomponentrelative_urlfromstyle="--image: url('{{ image | relative_url | uri_escape }}')"incontent.htmlConsolidation/summary of the first three bullets: this bug only happens when
sectionis not in main blog page, and is in blog post. The other 3 permutations of these two booleans are fine.What appears to be happening is related to the
section.htmlcomponent and the non-conventional (by necessity) way thatcontent.htmlcaptures the section component parameters.The fix for this is probably replacing the e.g.
"dark: (.*);"regexes incontent.htmlwith"dark: (.*?);", because it seems like the capture group is greedily matching too far, i.e.background: some-imageˇ; dark: true; size: pageˇ;(from ˇ to ˇ).What I still don't understand, or just can't wrap my head around in my current mental state, is why I've never run into this before, and why the particular solutions bulleted above make the problem go away. This makes me uneasy that a "fix" for this won't be a true fix: some edge cases will remain error-prone, or something else new will break as a result.
The text was updated successfully, but these errors were encountered: