* src/itree.c (struct check_subtree_result): Remove `complete`.
(check_subtree): Remove `max_depth` arg (and adjust callers).
Use 0 as black-depth of empty tree.
Remove redundant `node->parent` check (already performed by the caller).
(check_tree): Replace with `check_tree_common` (update all callers).
Check the root's `parent` field.
(check_tree_no_rb): Delete function, inlined in its sole caller.
(interval_tree_remove): Add call to `check_tree` (without RB checks)
before `interval_tree_remove_fix`. Move update of `size`
field accordingly.
With this change all fields in the itree_null sentinel are read only.
This makes accessing itree_null thread safe in an obvious way.
Because it took two commits from two peole to get this right, I think
we can call this design fragile and difficult to reason about.
Another benefit of this commit is as preparation for removing sentinel
node completely, and just using NULL.
* src/itree.c (itree_null): Statically initialize itree_null.parent to
NULL. It is never accessed.
(null_is_sane): Assert parent == NULL.
(interval_tree_remove_fix): Remove unecessary assignments to parent
from node->parent. These were the last places itree_null.parent were
read.
(interval_tree_remove): Avoid an assignment to itree_null.parent
through min->right->parent.
(interval_tree_transplant): Avoid an assignment to itree_null.parent
through source->parent.
Stefan recently disabled this but I happened to want it back soon
after.
* src/itree.c (check_subtree): new arg: allow_red_red
(check_tree_common): renamed from check_tree, pass allow_red_red
through.
(check_tree): new function, pass allow_red_red=false
(interval_tree_insert): check_tree -> check_tree_common with
allow_red_red=true.
* src/itree.c (struct check_subtree_result): new struct returned by
check_subtree.
(check_subtree): new function, renamed from recurse_check_tree. Add
new black height assertions.
(check_tree): assert that the tree has non-negative size, assert that
limiting to interval_tree_max_height(tree) levels is enough to
traverses the complete tree.
Move args between `build_overlay` and `add_buffer_overlay`,
to try and keep buffer positions together with their buffer.
Be more strict in the `otick` values passed to `interval_tree_insert`.
Move a few things around to try and reduce dependencies through `.h` files.
Fix a thinko bug in `check_tree`.
* src/alloc.c (build_overlay): Remove `begin` and `end` args.
* src/buffer.c (add_buffer_overlay): Move from `buffer.h`.
Add `begin` and `end` args.
(copy_overlays): Adjust accordingly.
(Fmake_overlay): Use BUF_BEG and BUF_Z; adjust call to `build_overlay`
and `add_buffer_overlay`.
(Fmove_overlay): Use BUF_BEG and BUF_Z; Use the new `begin` and `end`
args of `add_buffer_overlay` so we don't need to use
`interval_node_set_region` when moving to a new buffer.
(remove_buffer_overlay, set_overlay_region): Move from `buffer.h`.
* src/buffer.h (set_overlay_region, add_buffer_overlay)
(remove_buffer_overlay): Move to `buffer.c`.
(build_overlay): Move from `lisp.h`.
(maybe_alloc_buffer_overlays): Delete function (inline into its only
caller).
* src/itree.c (interval_tree_insert): Move declaration `from buffer.h`.
(check_tree): Fix initial offset in call to `recurse_check_tree`.
Remove redundant check of the `limit` value.
(interval_node_init): Remove `begin` and `end` args.
(interval_tree_insert): Mark it as static.
Assert that the new node's `otick` should already be uptodate and its
new parent as well.
(itree_insert_node): New function.
(interval_tree_insert_gap): Assert the otick of the removed+added nodes
were uptodate and mark them as uptodate again after adjusting
their positions.
(interval_tree_inherit_offset): Check that the parent is at least as
uptodate as the child.
* src/lisp.h (build_overlay): Move to `buffer.h`.
* src/itree.h (interval_node_init): Adjust accordingly.
(interval_tree_insert): Remove declaration.
(itree_insert_node): New declaration.
The insertion code tried to manipulate the offset in its own way,
and apparently there was a bug in it. Replace that with a call to
`interval_tree_inherit_offset`, making the whole logic a bit simpler,
and fixing a bug along the way (not sure where the bug was, to be honest).
* src/itree.c (interval_tree_insert): Use `interval_tree_inherit_offset`.
Check the tree before insert_fix.
(recurse_check_tree): Don't check RB invariants.
(itree_limits_are_stable): Delete function (subsumed by `check_tree`).
* src/itree.c (interval_tree_remove_fix): Add a `parent` argument.
Change the loop so it always keeps both `node` and `parent` in sync,
thus avoiding the use of `node->parent` on the initial node (since
that one can be null).
(interval_tree_remove): Manually keep track of the `broken` node's
parent to pass it to `interval_tree_remove_fix`.
In some cases, `interval_tree_remove` could cause some nodes to
inherit fewer (or additional) offsets than the should because nodes
were transplanted between two parts of the tree where offsets had not
been propagated "equally". So we remove/apply all offsets along the
path between the two points of a transplant before doing the transplant.
* src/itree.c (interval_tree_subtree_min): Move before first use; delete
the declaration; add an `otick` argument, and use it to update offsets
along the way.
(interval_tree_remove): Update all offsets on the way from `node` to `min`.
Reorder some of the operations so that when we transplant `min` to `node`
those nodes are in the proper state where `interval_tree_transplant`
can do its sanity checks.
(itree_total_offset): New function.
(interval_tree_transplant): Use it to sanity check that improper
offsets aren't accidentally inherited/lost because of the transplant.
(itree_newlimit): New function.
(itree_limit_is_stable, interval_tree_update_limit)
(interval_tree_propagate_limit): Use it.
(null_is_sane): Remove `inline` annotation; it's not needed.
(interval_tree_inherit_offset): Sanity check that `offset` is 0 when
`otick` is uptodate. Skip the unneeded increments when the offset is 0.
(interval_tree_insert_fix): Add sanity check that we indeed have 2 reds.
`interval_tree_remove` called `interval_tree_propagate_limit (subst)`
and `interval_tree_propagate_limit (min_right)` but both of those nodes
are moved without touching their subtrees, so their `limit`s are
"stable" causing `interval_tree_propagate_limit` to do nothing.
Indeed we don't need to update those nodes's `limit`s but we *do*
need to update their parents since those nodes have been moved.
Incidentally, this removes some uses of `null->parent` :-)
There are more uses of `null->parent`, tho, so I added more comments
explaining them (with the help of the matching section of the book
from which the algorithm was taken).
* src/itree.c (interval_tree_update_limit): Remove unused arg `tree`.
(interval_tree_rotate_left, interval_tree_rotate_right): Adjust callers.
(interval_tree_contains): Mark as static.
(itree_limit_is_stable, itree_limits_are_stable): New functions.
(interval_tree_remove): Fix incomplete update of `limit`s in corner
cases.
(interval_generator_next): Add sanity check to make sure the `limit`s
were properly updated.
* src/itree.h (interval_tree_contains): Remove declaration.
* src/itree.c (null_is_sane): New function.
(interval_tree_iter_start): Use it to make sure we haven't garbled null.
(itree_init): Fully initialize `itree_null` here...
(interval_tree_clear): ...instead of here.
(interval_tree_propagate_limit): Remove special code that read NULL->parent.
(interval_tree_remove): Do it explicitly before the call in those two
places where it can happen.
Instead of having one iterator object per buffer, use just a single
global one. There is virtually no benefit to having per-buffer
iterators anyway: if two iterations can be active at the same time,
then there can be cases where those two iterations happen
to operate on the same buffer :-(
* src/itree.h (struct interval_tree): Remove `iter` field.
* src/itree.c (interval_generator_destroy)
(interval_tree_iter_ensure_space): Delete functions.
(iter): New global variable.
(init_itree_null): Rename to `itree_init` and adjust all callers.
Initialize `iter` as well.
(interval_tree_create, interval_tree_init):
Don't initialize `iter` field any more.
(interval_tree_destroy): Don't destroy `iter` field any more.
(interval_tree_insert): Don't bother growing the iterator (it's grown
in `interval_stack_push_flagged` if needed anyway, and in any case
there's no `iter` here to grow any more).
(interval_tree_remove): Tweak assertion to be more precise and
self-evident.
(interval_tree_iter_start): Use the global `iter`.
(interval_generator_create): Make it work with a NULL argument.
This commit basically reverts commit 5b954f8f9. The problem of nested
iterations hasn't been fixed in the mean time, but since the GC can
run arbitrary ELisp code (via `post-gc-hook`), running the GC from
within an itree iteration is already unsafe anyway :-(
* src/alloc.c (mark_overlays): Delete function.
(mark_buffer): Use ITREE_FOREACH.
* src/itree.c (interval_generator_ensure_space, interval_generator_reset):
Inline and then delete functions.
(interval_tree_inherit_offset): Only take the tree's `otick` as arg.
Update all callers.
(struct interval_generator): Remove `tree` field, replace with a copy
of the tree's `otick`.
(interval_stack_push_flagged): The arg should be a real node.
(interval_tree_insert_gap): Prefer checking root==NULL rather than size==0.
Skip loop when tree is empty to avoid pushing&processing the NULL node.
(interval_tree_inherit_offset): Prefer parent==NULL rather than
node==root to avoid accessing the tree object.
* src/buffer.c (delete_all_overlays): Add comment.
* src/itree.c (struct interval_generator): New fields `running`,
`file`, and `line` moved from `interval_tree`.
(interval_stack_push_flagged): Adjust comment to resolve a FIXME.
(interval_tree_clear): Replace assignment with an a
(interval_tree_iter_next): Delete function.
(interval_tree_clear): Don't set `iter_running` here any more.
(interval_generator_create): Set it here instead.
(interval_tree_iter_start): Fetch `iter` once and for all.
(interval_generator_narrow): Mark it as non-static.
(interval_tree_iter_next, interval_tree_iter_narrow):
Delete functions. Inline their old bodies in the callers.
(interval_tree_iter_finish): Take the iter rather than
the whole tree. Adjust all callers.
(interval_generator_next): Move `running `assertion here from
`interval_tree_iter_next`.
* src/buffer.h: Adjust accordingly.
* src/itree.h (struct interval_tree): Remove fields `iter_running`,
`file`, and `line`, moved to `interval_generator`.
(interval_generator_narrow): Replace `interval_tree_iter_narrow`.
These bits really belong in the "workstack" used within
`interval_generator_next`, so move them there.
* src/itree.c (nodeptr_and_flag): New type;
(struct interval_stack): Use it.
(make_nav, nav_nodeptr, nav_flag): New functions.
(interval_tree_insert_gap, interval_tree_delete_gap): Adjust accordingly.
(interval_generator_next): Stash the `visited` bit in the work stack
rather than inside the tree nodes.
(interval_stack_create, interval_stack_destroy, interval_stack_clear)
(interval_stack_ensure_space, interval_stack_push_flagged)
(interval_stack_push, interval_stack_pop): Move before first use.
* src/itree.h (struct interval_node): Remove `visited` field.
* src/pdumper.c (dump_interval_node): Adjust accordingly.
* src/alloc.c (mark_overlay): Add sanity check.
* src/buffer.c (next_overlay_change, previous_overlay_change):
Tweak code to keep the same vars for the bounds.
* src/itree.c (interval_tree_clear, interval_tree_insert)
(interval_tree_remove, interval_tree_insert_fix, interval_tree_remove_fix):
Adjust to the `color` -> `red` change.
(interval_tree_clear): Prefer `true/false` for booleans.
(interval_generator_create): Use an actual `interval_tree_order` value
rather than 0.
(interval_generator_next): Simplify a tiny bit. Add comment.
(interval_generator_narrow): Add sanity check.
* src/itree.h (struct interval_node): Replace `color` field with
boolean `red` field.
(enum interval_tree_order): Remove unused `ITREE_DEFLT_ORDER` value.
* src/pdumper.c (dump_interval_node): Adjust to the
`color` -> `red` change.
When starting an iteration, store __FILE__ and __LINE__ where this
happens in the interval_tree structure.
* src/buffer.h (buffer_overlay_iter_start): New macro adding __FILE
and __LINE__.
(buffer_overlay_iter_start1): Renamed from ..._start.
* src/itree.h (struct interval_tree): Add file and line info.
* src/itree.c: (interval_tree_contains, interval_tree_nodes,
interval_tree_insert_gap): Pass __FILE__ and __LINE__ to iter_start.
(interval_tree_iter_start): Record file and line info in tree.
* lisp/progmodes/hideshow.el
(hs-find-block-beginning-match): New function to be used in
`hs-already-hidden-p'.
(hs-already-hidden-p): Add check if beginning of line is inside a
block.
(hs-toggle-hiding): Don't change to selected-window's buffer when
event arg is absent.
* test/lisp/progmodes/hideshow-tests.el
(hideshow-tests-with-temp-buffer-selected): New helper macro.
(hideshow-tests-make-event-at): New helper function.
(hideshow-already-hidden-p-1): New test.
(hideshow-toggle-hiding-1): New test.
(hideshow-mouse-toggle-hiding-1): New test (bug#52092).
* lisp/emacs-lisp/nadvice.el:
* lisp/emacs-lisp/cl-generic.el: Don't push to
package--builtin-versions "manually", because loaddefs-gen does
this correctly now.
When synthesizing a posn for keyboard events, make sure the `posn-point`
i the same as `point`.
* lisp/subr.el (event--posn-at-point): New function.
(event-start, event-end): Use it.
