This problem made test64 to crash. Description of the bug: when removing entry
from history when removed entry is not the last one it puts one element to
free_entries list, but ignores free entries starting from last_free_element.
Possible solutions:
1. First working: simply populate free_entries list with entries which are still
free, starting from last_free_element.
2. Better (wastes less CPU): after free_entries list size goes to zero (which is
the initial value) continue using last_free_element.
3. Even better (less memory): note that element from the list is *only* removed
before adding another one. So replace free_entries array with one item.
Also renamed last_free_element to last_free_entry: in any case most of the lines
which mention it were altered.
It appears that large portion of non-ShaDa ASCII text files may be parsed as
a ShaDa file because it is mostly recognized as a sequence of unknown entries:
all ASCII non-control characters are recognized as FIXUINT shada objects, so
text like
#!/bin/sh
powerline "$@" 2>&1 | tee -a powerline
(with trailing newline) will be recognized as a correct ShaDa file containing
single unknown entry with type 0x23 (dec 35, '#'), timestamp 0x21 (dec 33, '!')
and length 0x2F (dec 47, '/') without this commit. With it parsing this entry
will fail.
For compatibility the following things are done:
1. Items with type greater then greatest type are ignored when reading and
copied when writing.
2. Registers with unknown name are ignored when reading and blindly copied when
writing.
3. Registers with unknown type are ignored when reading and merged as usual when
writing.
4. Local and global marks with unknown names are ignored when reading. When
writing global marks are blindly copied and local marks are also blindly
copied, but only if file they are attached to fits in the `'N` limit defined
in &shada. Unknown local mark’s timestamp is also taken into account when
calculating which files exactly should fit into this limit.
5. History items with unknown type are ignored when reading and blindly copied
when writing.
6. Unknown keys found in register, local marks, global marks, changes, jumps and
search pattern entries are read to additional_data Dictionary and dumped (of
course, unless any of these elements were not overwritten later). It
obviously works only for values conversible to Object type.
7. Additional elements found in replacement string and history entries are read
to additional_elements Array and dumped (same: only if they were not
overwritten later). Again this works only for elements conversible to Object
type.
8. Additional elements found in variable entries are simply ignored when
reading. When writing *new* variables they will be preserved during merging,
but that’s all. Variable values dumped from current NeoVim session never have
additional elements.
Modifications:
- If file was not written due to write error then writing stops and temporary
file will not be renamed.
- If NeoVim detects that target file is not a ShaDa file then temporary file
will not be renamed.
Some notes:
- Replaced msgpack_unpacker usage with regular xmalloc’ed buffer. Also since
msgpack_unpack_next (as well as msgpack_unpacker_next) is not ever going to
return MSGPACK_UNPACK_EXTRA_BYTES this condition was checked manually.
Function that does return this status is msgpack_unpack, but it is marked as
obsolete.
- Zero type is checked prior to main switch in shada_read_next_item because
otherwise check would be skipped.
- Zeroing entry at the start of shada_read_next_item makes it safer.
- dedent('') does not work.
- v:oldfiles list is only replaced with bang, if it is NULL or empty.
What works:
1. ShaDa file dumping: header, registers, jump list, history, search patterns,
substitute strings, variables.
2. ShaDa file reading: registers, global marks, variables.
Most was not tested.
TODO:
1. Merging.
2. Reading history, local marks, jump and buffer lists.
3. Documentation update.
4. Converting some data from &encoding.
5. Safer variant of dumping viminfo (dump to temporary file then rename).
6. Removing old viminfo code (currently masked with `#if 0` in a ShaDa file for
reference).
Add a new special key that can be used by UIs to toggle the 'paste' option and
use it in the TUI instead of the user's 'pastetoggle' value.
Close#2843#2092
No one has taken a real interest in fixing this, so let's disable it for
now. The issue crops up most on the QB OS X node, but it has
periodically occurred under other nodes too.
Use the existing Vimscript function provider#pythonx#Detect()
to determine whether the Neovim Python module is installed and
Python 2/3 tests can be run.
Always run tests with encoding=utf-8, regardless of user locale
Don't set &encoding after startup in tests
Helped-By: Michael Reed <m.reed@mykolab.com>
This is equivalent to patches 7.4.396, 7.4.445 and 7.4.598.
vim-patch:7.4.396
Problem: When 'clipboard' is "unnamed", :g/pat/d is very slow.
(Praful)
Solution: Only set the clipboard after the last delete. (Christian
Brabandt)
1f285eb49a
vim-patch:7.4.445
Problem: Clipboard may be cleared on startup.
Solution: Set clip_did_set_selection to -1 during startup. (Christian
Brabandt)
1a19d37d90
vim-patch:7.4.598
Problem: ":tabdo windo echo 'hi'" causes "* register not to be
changed.
(Salman Halim)
Solution: Change how clip_did_set_selection is used and add
clipboard_needs_update and global_change_count. (Christian
Brabandt)
af6a579263
Co-Author: @bfredl
os_file_is_readonly() in its current form is equivalent to
!os_file_is_writable(). This does not appear to be a bug, because Vim's
use of check_file_readonly() (which we changed to os_file_is_readonly())
is equivalent to !os_file_is_writable() in every case.
os_file_is_readonly() also fails this test:
returns false if the file is non-read, non-write
A more useful form would define behavior under these cases:
- path is executable (but not writable)
- path is non-existent
- path is directory
But there is no reason for os_file_is_readonly() to exist, so remove it.
This is a port of my original contribution to Vim, added in 7.4.687
(https://github.com/vim/vim/commit/v7-4-687). The TUI code has been
heavily refactored (see esp. 25ceadab37),
so this required some translation, but the logic is the same.
Currently, there are two functions in the UI API that are called when
the mode changes: insert_mode() and normal_mode(). These can be folded
into a single mode_change() entrypoint which can do whatever it wants
based on the mode it is passed, limited to INSERT and NORMAL for now.
A menu item can have separate bindings for each Vim mode.
:emenu checks to see which binding it should execute. But, it assumes
it can only be called from Normal mode, so its mode detection is based
on some guesswork. For instance, it detects if you've just used C-O
and, if so, uses the Insert mode binding.
Now that :emenu can be called from any mode (via vim_command), this
commit has it check the actual mode we're in, and simply use the
binding for that mode if we aren't in Normal mode.
The test is also split in several blocks and heavily modernized. This was
done to prevent the following quoting and escaping problems during migration:
- the vim command `put =...` treats double quotes as the start of a comment so
they have to be escaped with a backslash
- when inserting control characters on the command line they have to be
escaped with <C-V>
The parts one and two of the test are functional identical so they are wrapped
in a local function. The only difference was which letters where used to test
the same feature.
Part six did test a flag in 'cpoptions' that has been removed in neovim. It
has therefore been removed as well.
Reviewed-by: Michael Reed <Pyrohh@users.noreply.github.com>
os.remove() wasn't removing the temporary swap directory which leads to
problems when the test is run a second time.
That's also the reason why the CI never caught this.
os.remove() got replaced by helpers.rmdir().
- lfs.rmdir() only removes empty directories
- os.remove() supercedes lfs.rmdir(); removes files and empty directories
- helpers.rmdir() first removes all files within a directory, then the
directory itself
- Add event loop abstraction module under src/nvim/event. The
src/nvim/event/loop module replaces src/nvim/os/event
- Remove direct dependency on libuv signal/timer API and use the new abstraction
instead.
- Replace all references to uv_default_loop() by &loop.uv, a new global variable
that wraps libuv main event loop but allows the event loop functions to be
reused in other contexts.
Since sleep is a grandchild of nvim, it is not killed after the test ends.
Using a low sleep value allows it to exit automatically after a small interval.
- use eval() and eq() in many places instead of writing to the buffer
- remove has('autocmd') checks and use corresponding code unconditionally as
neovim always has the autocmd feature
- split the test into several it() blocks
Helped-By: Scott Prager <splinterofchaos@gmail.com>
Helped-By: Michael Reed <m.reed@mykolab.com>
Extract the RBuffer class from rstream.c and reimplement it as a ring buffer,
a more efficient version that doesn't need to relocate memory.
The old rbuffer_read/rbuffer_write interfaces are kept for simple
reading/writing, and the RBUFFER_UNTIL_{FULL,EMPTY} macros are introduced to
hide wrapping logic when more control is required(such as passing the buffer
pointer to a library function that writes directly to the pointer)
Also add a basic infrastructure for writing helper C files that are only
compiled in the unit test library, and use this to write unit tests for RBuffer
which contains some macros that can't be accessed directly by luajit.
Helped-by: oni-link <knil.ino@gmail.com>
Reviewed-by: oni-link <knil.ino@gmail.com>
Reviewed-by: Scott Prager <splinterofchaos@gmail.com>
Reviewed-by: Justin M. Keyes <justinkz@gmail.com>
Reviewed-by: Michael Reed <m.reed@mykolab.com>
This is the part of the test that relies on wall clock time and sometimes
fails if the system is under load. The test is repeated up to three times
before a failure is reported to the user.
:undojoin can only be used inside scripts and command chains. So it has to be
tested inside an explicit `source()` call.
Also add a new test case for the different behavior when sourceing normal mode
commands from a script or inserting them interactively.
The legacy test uses `set ul=100` to break the changes into blocks that can be
undone separately. This is needed because the legacy test is sourced from a
file and changes would be grouped into on undo block by default. The lua test
suite does not have this restriction.
Also add a new test case to test this effect of using `set ul=100` in a
sourced script.
This test is real time based as it also tests the `:earlier` and `:later`
commands with time arguments (using `:sleep`). This can sometimes case the
test to fail on systems that are under heavy load or where the time interval
between creating the expected buffer state and the `:earlier` or `:later`
command that tries to jump to it changes.
To be system independent we use nvim's `:sleep` command and `wait()` for it in
the test suit.
The legacy vim test writes to test.out a lot with `:.w >>test.out`. This
does currently not work in the lua test suite so the test is modernized to use
busted's assertions instead of the output file.
This test was treated special in the legacy Makefile but after the conversion
the related code can be removed.
Reviewed-by: Michael Reed <m.reed@mykolab.com>
Reviewed-by: Luke Andrew <luke.github@la.id.au>
Reviewed-by: Justin M. Keyes <justinkz@gmail.com>
Reviewed-by: Florian Walch <florian@fwalch.com>
Remove related dead code and references in the docs.
Helped-By: Michael Reed <m.reed@mykolab.com>
Helped-By: Shougo Matsushita <Shougo.Matsu@gmail.com>
