fix(engine): add workaround to markdown parsing to avoid crashing espanso with malformed markdown. Fix #759

This commit is contained in:
Federico Terzi 2021-10-21 21:52:22 +02:00
parent 85bd14236a
commit cd89915525

View File

@ -17,6 +17,8 @@
* along with espanso. If not, see <https://www.gnu.org/licenses/>.
*/
use log::error;
use super::super::Middleware;
use crate::event::{effect::HtmlInjectRequest, Event, EventType};
@ -37,23 +39,33 @@ impl Middleware for MarkdownMiddleware {
fn next(&self, event: Event, _: &mut dyn FnMut(Event)) -> Event {
if let EventType::MarkdownInject(m_event) = &event.etype {
// Render the markdown into HTML
let html = markdown::to_html(&m_event.markdown);
let mut html = html.trim();
// NOTE: we wrap the `to_html` call between catch_unwind because if the markdown is malformed,
// the library panics. Ideally, the library would return a Result::Err in that case, but
// for now it doesn't, so we employ that workaround.
// See also: https://github.com/federico-terzi/espanso/issues/759
let html = std::panic::catch_unwind(|| markdown::to_html(&m_event.markdown));
if let Ok(html) = html {
let mut html = html.trim();
// Remove the surrounding paragraph
if html.starts_with("<p>") {
html = html.trim_start_matches("<p>");
}
if html.ends_with("</p>") {
html = html.trim_end_matches("</p>");
}
// Remove the surrounding paragraph
if html.starts_with("<p>") {
html = html.trim_start_matches("<p>");
}
if html.ends_with("</p>") {
html = html.trim_end_matches("</p>");
}
return Event::caused_by(
event.source_id,
EventType::HtmlInject(HtmlInjectRequest {
html: html.to_owned(),
}),
);
return Event::caused_by(
event.source_id,
EventType::HtmlInject(HtmlInjectRequest {
html: html.to_owned(),
}),
);
} else {
error!("unable to convert markdown to HTML, is it malformed?");
return Event::caused_by(event.source_id, EventType::NOOP);
}
}
event