mirror of
https://github.com/discourse/discourse.git
synced 2026-08-04 10:39:43 +08:00
4cf76d2ffc
markbridge 0.4.0 gives `AST::Code` a `block` flag and sets it for `[code]` and `[pre]`, which are code blocks by definition — `[tt]` stays inline teletype and still leaves the choice to the renderer. So a single-line `[code=ruby]x = 1[/code]` inside a link label is now lifted out as a fence, exactly like the multi-line case, and the link is left with an empty label. A spec asserted the old reading, that a language attribute alone never made code a block and only multi-line code was hoisted. That premise is gone, so the three shapes — single-line with a language, single-line without, and multi-line — are now one table asserting the hoist. The unit test further down still covers code that is inline to begin with, which is the case that can stay in a label. The per-gem Gemfile.lock is gitignored, so CI resolves the newest markbridge while a local checkout keeps whatever it locked earlier. That is why this only broke in CI. The gemspec floor moves to 0.4.0, since the specs now describe that behaviour.
19 lines
519 B
Ruby
Vendored
19 lines
519 B
Ruby
Vendored
# frozen_string_literal: true
|
|
|
|
Gem::Specification.new do |s|
|
|
s.name = "migrations-converters"
|
|
s.version = "0.0.1"
|
|
s.summary = "Discourse migrations: public converter implementations"
|
|
s.authors = ["Discourse Team"]
|
|
s.required_ruby_version = ">= 3.4"
|
|
|
|
s.files = Dir["lib/**/*"]
|
|
|
|
s.add_dependency "migrations-core"
|
|
s.add_dependency "activesupport"
|
|
s.add_dependency "colored2"
|
|
s.add_dependency "i18n"
|
|
s.add_dependency "markbridge", ">= 0.4.0"
|
|
s.add_dependency "pg"
|
|
s.add_dependency "zeitwerk"
|
|
end
|
