From: Kevin Kappelmann <cl-isabelle-users@lists.cam.ac.uk>
Hi all,
in Isabelle/Scala, Headless.Resources uses the default append_path,
which does not resolve symlinks. The same file can thus become two
nodes, e.g. with a symlink "/tmp/a -> /tmp/b" and a headless session
session.use_theories(List("/tmp/a/Scratch"))
session.use_theories(List("/tmp/b/Scratch"))
loads and checks Scratch twice as two separate nodes.
Isabelle/jEdit, in contrast, resolves symlinks in its append_path since
e212e2001b2a. Could Headless.Resources do the same, e.g. with .canonical
instead of .expand? Or could the constructor of Headless.Resources be
made public, so that applications, like PIDE MCP, can override
append_path themselves?
Best wishes,
Kevin
From: Kevin Kappelmann <cl-isabelle-users@lists.cam.ac.uk>
Brief update: this is not relevant for my application anymore as I
decided to write my own Resources layer for other reasons anyway.
Best wishes,
Kevin
On 06.11.26 09:59, Kevin Kappelmann wrote:
Hi all,
in Isabelle/Scala, Headless.Resources uses the default append_path,
which does not resolve symlinks. The same file can thus become two
nodes, e.g. with a symlink "/tmp/a -> /tmp/b" and a headless sessionsession.use_theories(List("/tmp/a/Scratch"))
session.use_theories(List("/tmp/b/Scratch"))loads and checks Scratch twice as two separate nodes.
Isabelle/jEdit, in contrast, resolves symlinks in its append_path since
e212e2001b2a. Could Headless.Resources do the same, e.g. with .canonical
instead of .expand? Or could the constructor of Headless.Resources be
made public, so that applications, like PIDE MCP, can override
append_path themselves?Best wishes,
Kevin
Last updated: Nov 19 2026 at 01:46 UTC