Closed kukimik closed 5 months ago
Hmm.. I don't know why the tests failed on Windows. I guess it may be LF vs CRLF inside the <pre>...</pre>
block? Is there any way to see what file was generated during the tests? I don't have a machine with Windows configuration, so it's difficult for me to try it out myself.
It was a \n
vs \r\n
(dos vs unix newlines) issue caused by git. Solved by adding a line to .gitattributes
. The problem is discussed in tasty-golden
docs.
I noticed that text files that are not imported by dhall files are not rendered as preformatted text, which often looks awful. See the
README.md
fordhall-semver
:The raw text looks much better: https://raw.githubusercontent.com/Gabriella439/dhall-semver/main/README.md. Moreover, displaying as preformatted text is much closer to the real contents of the file.
I can't be 100% sure, but this looks like an unintentional omission in #2407. The test didn't catch it because the
Plain.txt
file is imported byAsText.dhall
and thus treated differently (I didn't look for the part of code responsible for this behaviour, I'm guessing based on what I see).Side note: it's also rather easy to add support for rendering Markdown files as HTML (I have a working proof of concept, but it needs some polish). This could be nice for files like
README.md
, which are common in Dhall packages. On the other hand, Markdown is quite readable on its own. Anyway, that's a story for another PR.