Zulip Chat Archive
Stream: general
Topic: Stale web editor behavior
Eric Wieser (Nov 24 2024 at 22:16):
I'm seeing some confusing behavior in the web editor on mathlib stable, where
#eval Lean.versionString == "4.13.0"
#guard Lean.versionString == "4.13.0"
shows the eval
as being true, but fails the #guard
Eric Wieser (Nov 24 2024 at 22:16):
The UI also seems confused:
Eric Wieser (Nov 24 2024 at 22:16):
Refreshing and adjusting the dropdown seems to reliably reproduce
Mario Carneiro (Nov 24 2024 at 22:18):
I just tried opening that on the playground and consistently got a false
from eval and a failure of the guard
Eric Wieser (Nov 24 2024 at 22:18):
Even after switching the playground to "mathlib stable"?
Mario Carneiro (Nov 24 2024 at 22:19):
oh, I missed that dropdown
Mario Carneiro (Nov 24 2024 at 22:19):
with it, both seem to say true
Eric Wieser (Nov 24 2024 at 22:19):
I think the failing sequence is:
- Open the example above with the default version
- Switch the dropdown to "stable"
- Refresh the page
Mario Carneiro (Nov 24 2024 at 22:20):
indeed, it seems there are now two LSPs running on both versions
Eric Wieser (Nov 24 2024 at 22:20):
(where step 3 often happens because I forgot about the tab for a while, and the server shutdown in the meantime)
Kyle Miller (Nov 24 2024 at 22:21):
I've seen this error a few times now. Somehow there's a second LSP that gives additional underlines, but you don't get messages from it. There's no error message on the #guard
, right? just the red squiggle?
Eric Wieser (Nov 24 2024 at 22:21):
Mario Carneiro said:
indeed, it seems there are now two LSPs running on both versions
Ah, you're right; editing the code leads to up-to-date printouts from both versions, not just stale printouts from the old version
Mario Carneiro (Nov 24 2024 at 22:22):
If I do
#guard Lean.versionString == "4.13.0"
#guard Lean.versionString != "4.13.0"
I get red squiggles from both
Eric Wieser (Nov 24 2024 at 22:22):
@Kyle Miller: it looks like the editor is connected to both LSPs, but the infoview is connected to only the right one
Kyle Miller (Nov 24 2024 at 22:22):
@Joachim Breitner was guessing that it has to do with https://github.com/leanprover-community/lean4web/issues/41
Joachim Breitner (Nov 25 2024 at 08:48):
Indeed, and @Jon Eugster mentioned that he’ll have a bit more time to work on lean-related stuff the coming weeks.
Last updated: May 02 2025 at 03:31 UTC