Fix stale hover reuse for unchanged regions#1533
Conversation
|
What are the new conditions to avoid reusing a previoulsy computed hover request? Is it now always recomputed? Could you specify if as part of the commit / PR message? |
LSPTextHover reused a completed hover request when the viewer and hover region were unchanged. As a result, a repeated hover over the same region could return stale content and avoid issuing a fresh textDocument/hover request. A previous request is still reused while it is in progress for the same viewer/region, but any completed requests are now treated as stale, so a subsequent hover on the same region starts a fresh textDocument/hover request instead of reusing the old completed result indefinitely. Fixes eclipse-lsp4e#1514
Hi, good question. No it is not always recomputed. I've updated the commit message appropriately. Cheers John |
|
Hi @rubenporras , I see the following failure for my PR: continuous-integration/jenkins/pr-merge — This commit cannot be built I'm not sure what I need to do, if anything. Is it simply waiting for a committer to approve the run? |
Hi @rubenporras , |
|
I haven't reviewed the code (yet), but I approved the workflow runs |
I think this will trigger many unneeded recomputations. Could we do it so that we invalidate the last computed hover request only if the editor looses focus? That should cover the case of external resources and be less agressive on the computation side. |
LSPTextHover reused a completed hover request when the viewer and hover region were unchanged. As a result, a repeated hover over the same region could return stale content and avoid issuing a fresh textDocument/hover request.
Fixes #1514