Regression in 2.3.8: routes that contain an escaped question mark silently fail

[jawj]

Is there an existing issue for this?

• I have searched the existing issues

Mithril.js Version

2.3.8

Browser and OS

macOS 15.7.3, Firefox 147.0, Chrome 143.0.7499.193 (this is not a browser/OS bug)

Project

https://manipulexity.com
https://github.com/jawj/wordtools

Code

example = (pattern: string, orderDir = 'freq/desc') =>
  m(m.route.Link, { href: `/pattern/${encodeURIComponent(pattern)}/single/${orderDir}/1`, selector: 'a.example' }, pattern)

// if pattern contains a ?, it is escaped, but the resulting link doesn't do anything

Steps to Reproduce

Visit manipulexity.com
Click the last link, captioned 'palindromes'

Expected Behavior

It shows palindromes

Observed Behavior

Because I have pinned 2.3.7, it works, and shows palindromes.
If I upgrade to 2.3.8, nothing happens.

Context

No response

[kfule]

@jawj Thank you for your report.
In v2.3.8, the URL decoder now decodes encoded ASCII (including question mark) in routing, which may be causing the issue.
It may be better not to decode encoded ASCII, as in v2.3.7 and earlier. Or, it seems necessary to properly handle question marks that appear as a result of decoding within the path string.

Edit: flems
flems (v2.3.7)
flems (v2.3.8)

[jawj]

Hi @kfule. Thanks for the quick response.

It may be better not to decode encoded ASCII, as in v2.3.7 and earlier.

I took this as meaning you might not fix the problem. Of course, this is open-source and you don't owe me anything. But I must say that this really made me worry for the future of Mithril, because:

• Accepting arbitrary ASCII characters, properly escaped, in route paths seems like pretty basic functionality (my use case here is for regular expressions).
• It worked fine for years, and broke unexpectedly in a patch release.
• I can't see any compelling reason for the change, and certainly not one compelling enough to break existing apps.

But I see you've now reverted some changes to fix the problem, so hopefully 2.3.9 will be along soon and work again?

[kfule]

@jawj Sorry for the confusion - my earlier comment was about decoding strategy/placement, not "we won’t fix this". I agree that a patch release regression here is not acceptable, and escaped ASCII in route paths should be supported.

I’ve verified against your application code that the revert-style change fixes the issue. I’m not a maintainer so I can’t speak to release timing, but it should be resolved once merged and released.

[dead-claudia]

@kfule Could you send a PR to revert #3050? I mistakenly merged it as a set of commits instead of as just one, so it's not a simple revert. Edit: already saw you did in #3065. Sorry for the noise.