CVS log for src/lib/libedit/readline.c

Up to [cvs.NetBSD.org] / src / lib / libedit

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

fix insert key (Xose Vazquez Perez)

pass lint.

PR/57376: Jorge Giner: readline file completion does not quote; do the same.

Align types with readline-8.2 (wiz@)

PR/57095: Yilei Yang: Change readline's replace_history_entry to not make a copy
of the string to replace since H_REPLACE already makes a copy (fixes memory leak)

improvements in malloc/free handling.

PR/57016: Ricky Zhou: declare lastidx

PR/57016: Ricky Zhou: Revert to trimming the last newline instead of the
first one so that multi-line commands work again.

PR/56778: Detlev Casanova: Missing rl_initialize call in rl_copy_text

PR/56695: Walter Lozano: Correct declaration of hook functions.

PR/56693: Walter Lozano: Add support for rl_delete_text and rl_set_key

PR/56622: Walter Lozano: Improve readline compatibility by adding
rl_readline_state support.

Add more refreshes from Walter Lozano. The readline example in
http://www.mcld.co.uk/blog/2009/simple-gnu-readline-callback-style-example.html
still does not work, but it is better.

PR/56618: Walter Lozano: Improve libedit compatibility with readline by
implementing:

    rl_copy_text, rl_erase_empty_line, rl_message, rl_on_new_line,
    rl_replace_line, rl_restore_prompt, rl_save_prompt

libedit: fix indentation

No change to the resulting object files.

rl_startup_hook should be called each time. From Carlos Henrique Lima Melara

fix memory issues found by fuzzing (double frees and buffer overflows)

Try to refactor this in order to correct some of the memory issues
reported by Christian Holler.

PR/56370: mirabilos: libedit change from 2017 kills history in gdb

Add more api to make gdb-11 happy (but not gdbtui as usual)

libedit: simplify calls to macro ADD_STRING

The lint comments CONSTCOND and LINTED were not necessary.  It is
simpler to just specify what to free.  GCC optimizes free(NULL) to be a
no-op.

No functional change.

readline: fix lint warning about effective unconst cast

Calling strchr to avoid the syntactical unconst cast is not necessary
here.  A simple pointer assignment is enough.

No functional change.

Add a LINTED comment... Why doesn't NOTREACHED work?

Mostly merge changes from HEAD upto 20200411

add +1 to strlcpy's (Patrick Welche)

Change strncpy to either memcpy (when we know the len), or strlcpy (when
we used to NUL terminate explicitly.

Increment offset when adding an element to history to keep it aligned with
the last element entered (Sandy Li Changqing)

PR/54399: Sören Tempel: Uninitialized memory access in libedit history.
Initialize the buffer using calloc. While here change all malloc(a * sizeof(b))
to calloc(a, sizeof(b)). XXX: should fix realloc similarly.

Sync with HEAD

PR/54281: Jonathan Perkins: NUL terminate rl_line_buffer on modification
to avoid completion leak.

PR/54280: rl_completer_quote_characters should be const for readline compat

PR/54279: Jonathan Perkins: Ignore adjacent start/end prompt ignore.

Follow the man page for EL_GETTC and not require a NULL terminated argument
list: https://reviews.llvm.org/D61191

PR/53983: Jonathan Perkins: Fix types for readline compatibility

PR/53981: Jonathan Perkins: history_list should null-terminate

Synch with HEAD

PR/53856: F. Aragon: editline/libedit not prompting colors in readline mode

Sync with HEAD, resolve a few conflicts

Add a couple more readline compat functions.

Sync with HEAD

Provide more compatibility with readline headers; now python-3.6.5 works
when changing 'readline' -> 'edit' in setup.py.
Revert previous conditional setting of unbuffered.

Only FLUSH if we are ending libedit; DRAIN if we suspend for readline.
This allows pasting multiline buffers (Gerry Swislow)

For applications that don't issue their own prompt (like python)
don't set unbuffered unless they've already printed the prompt.
This avoids printing the prompt before the application has a chance
to process the input line.
From sjg@

PR lib/52547 - read_history (readline.c) should now sets history_length.

Patch from Yen Chi Hsuan in the PR, extracted from Apple's version of
readline.c, then modified by me to be consistent about what the return
value really is.

For readline emulation, don't reset the tty to "sane" (cooked) mode if we
did not start this way. Also set and reset the tty on entry and exit from
readline() since this is what readline does.

PR/51517: Jay West: Tty settings not restored on exit
PR/51518: Jay West: prompt is interleaved with client output

Both these issues are caused by rl_restore_handler not DTRT; fix
it so that it kills the internal libedit state completely. This is
inefficient, but it works.

Also fix:
1. add append_history()/H_NSAVE_FP
2. call the rl_startup_hook before printing the first prompt as documented.
   callint it from rl_initialize breaks python, because the callback ends
   up being invoked before the readline module is installed, and we end up
   dereferencing a NULL pointer.
3. add el_resize_terminal.

With those changes, s/lreadline/ledit/g in python works.

Sync with HEAD

Sync with HEAD

When doing filename autocompletion, append a trailing slash at the end of directory
names. We already do this when there is only one completion option but
in case of of multiple completion options, it wasn't being done.

ok christos@

Sync with HEAD

Make sure we take into account history_base when computing negative history
offsets. (Gerry Swinslow)

Sync with HEAD

pass the stream to the getc function

fix typo

more compatible with readline history functions.

Fix previous to better match readline behavior (Ingo Schwarze)

From Ingo Schwarze:

In libedit, the only way how H_ENTER can fail is memory exhaustion,
too, and of course it is handled gracefully, returning -1 from
history().  So of course, we will continue to handle it gracefully
in add_history() as well, but we are free to decide what to do with
the library state in this case because GNU just dies...

I think the most reasonable course of action is to simply not change
the library state in any way when add_history() fails due to memory
exhaustion, but just return.

If H_ENTER does not fail, we know that the history now contains at
least one entry, so there is no need any longer to check the H_GETSIZE
return value.  And we can of course always set current_history_valid.

While testing these changes, i noticed three problems so closely
related that i'd like to fix them in the same diff.

 1. libedit has the wrong prototype for add_history().
    GNU readline-6.3 defines it as void add_history(const char *).
    Of course, that is very stupid - no way to report problems to
    the caller!  But the whole point of a compatibility mode is
    being compatible, so we should ultimately change this.
    Of course, changing the prototype of a public symbol requires
    a libedit major bump.  I don't want to do that casually.
    Rather, i will take a note and change the prototype the next
    time we need a libedit major bump for more important reasons.
    For now, let's just always return 0.

 2. While *implicitely* pushing an old entry off the history
    increments history_base in GNU readline, testing reveals that
    *explicitly* deleting one does not.  Again, this is not
    documented, but it applies to both remove_history() and
    stifle_history().  So delete history_base manipulation
    from stifle_history(), which also allows to simplify the
    code and delete two automatic variables.

 3. GNU readline add_history(NULL) crashes with a segfault.
    There is nothing wrong with having a public interface
    behave that way.  Many standard interfaces do, including
    strlen(3).  Such crashes can even be useful to catch
    buggy application programs.
    In libedit/readline.c rev. 1.104, Christos made add_history()
    silently ignore this coding error, according to the commit
    message to hide a bug in nslookup(1).  That change was never
    merged to OpenBSD.  I strongly disagree with this change.
    If nslookup(1) is still broken, that program needs to be
    fixed instead.  In any case, delete the bogus check; hiding
    bugs is dangerous.

remove the right history entry (Ingo Schwarze)

From Bastian Maerkisch, via Igno Schwarze:

Even though section "2.3.3 Information About the History List"
of the history(3) info(1) manual only says

  -- Function: int where_history (void)
     Returns the offset of the current history element.

which maybe isn't completely clear, a plausible implementation
is that the offset returned is the same offset that can be used
for history_set_pos(), i.e. that it is 0 for the oldest entry
and increases with time, and that's how the GNU implementation
behaves indeed.

The libedit implementation, on the other hand, returns 1 for the
newest entry and increases going back in time.

GNU readline(3) regards history chronologically, that is, from the
perspective of the dawn of time, so "next" means "newer" and "previous"
means "older".  Libedit, by contrast, uses reverse chronology and
regards history from the perspective of the present, such that "next"
means "longer ago" and "previous" means "not so long ago".

The following patch fixes previous_history() and next_history()
as proposed by Bastian Maerkisch.

But there is a related problem demonstrated by Bastian's regression
tests that his patch did not fix:  next_history() can advance not
only to the newest entry, but beyond it, which core libedit cannot
do.  So that feature must be implemented locally in readline.c.

With that, the last of Bastians tests is fixed, test_movement_direction().

This patch also improves libedit documentation to more clearly state
what "previous" and "next" mean.  GNU readline documentation is
just as unclear, but we can't easily fix that since libedit doesn't
include its own readline.3 manual.

(Ingo Schwarze)

The libedit implementation of history_get() also differs from the GNU
implementation:  libedit goes to the entry with the given number
stored in the HistEvent structure, while GNU subtracts history_base,
then advances that many entries from the oldest one.  If entries were
removed in between, GNU advances further than libedit.

The call sequence H_CURR, H_DELDATA, H_CURR, H_NEXT_EVDATA looks
weird, as if part of that must somehow be redundant.  But actually,
the user interface is so counter-intuitive that every single step
is really required.

 - The first H_CURR is needed to be able to go back after an error.
 - The H_DELDATA is needed to move the cursor.  Even though it takes
   a pointer to ev, that structure is not filled in when the call
   succeeds.  H_DELDATA only moves the cursor, it doesn't tell us
   the new event number.
 - Consequently, the second H_CURR is required to get ev.num filled
   in.  But it doesn't return the data because ev has no field for
   that.
 - So even though the cursor is already positioned correctly,
   H_NEXT_EVDATA is needed as the final step merely to get the data.

(Ingo Schwarze)

In stiffle_history(), trim excessive entries from the history and advance
history_base like the GNU implementation does. (from Bastian Maerkisch)

fix logic (Ingo Schwarze)

From Ingo Schwarze:
 * Replace fcns.c by a shorter and simpler func.h
   and include it only in the one file needing it, map.c.
 * Combine help.h and help.c into a simplified help.h
   and include it only in the one file needing it, map.c.
 * Check the very simple, static files editline.c, historyn.c, and
   tokenizern.c into CVS rather than needlessly generating them.
 * So we no longer autogenerate any C files.  :-)
 * Shorten and simplify makelist by deleting the options -n, -e, -bc,
   and -m; the latter was unused and useless in the first place.
 * Move the declaration of el_func_t from fcns.h to the header
   actually needing it, map.h.  Since that header is already
   included by el.h for unrelated reasons, that makes el_func_t
   just as globally available as before.
 * No longer include the simplified fcns.h into el.h,
   include it directly into the *.c files needing it.

Start removing the WIDECHAR ifdefs; building without it has stopped working
anyway. (Ingo Schwarze)

Make the read_char function always take a wchar_t * argument (Ingo Schwarze)

whitespace and header sorting changes (Ingo Schwarze). No functional changes.

more include file cleanup (Ingo Schwarze)

From Ingo Scharze:
Let "el.h" include everything needed for struct editline,
and don't include that stuff multiple times.  That also improves
consistency, also avoids circular inclusions, and also makes it
easier to follow what is going on, even though not quite as nice.
But it seems like the best we can do...

cleanup chartype.h includes (Ingo Schwarze)

cleanup inclusion of histedit.h (Ingo Schwarze)

attribute unused

OpenBSD readline.c rev. 1.14 2015/02/06 23:21:58 millert
use SIZE_MAX

OpenBSD readline.c rev. 1.13 2015/01/13 08:33:12 reyk
rl_set_keyboard_input_timeout() for readline 4.2 compat

Adjust API to a more modern readline (Ryo Onodera)

- fix types of rl_completion_entry_function and rl_add_defun
- call update pos before completion to refresh the screen
From Thomas Eriksson

Sync lib/libedit with head, requested by christos in #753:

	lib/libedit/Makefile 1.53
	lib/libedit/chartype.h 1.13
	lib/libedit/editline.3 1.83-1.84
	lib/libedit/editrc.5 1.28-1.29
	lib/libedit/eln.c 1.18
	lib/libedit/filecomplete.c 1.33-1.34
	lib/libedit/readline.c 1.112-1.115

Man page improvements, fix overlapping strcpy, improve readline
compatibility, clang build fix.

Fix overlapping strcpy (Gerry Swislow)

set some readline compatibility default key settings.
https://bugzilla.redhat.com/attachment.cgi?id=1001895

src is too big these days to tolerate superfluous apostrophes.  It's
"its", people!

Rebase to HEAD as of a few days ago.

Fix typo in comment (Tobias Stoeckmann)

Rebase.

PR/48957: Federico G. Schwindt: Restore commented out code that broke
rl_callback_handler.

sync with head.

for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.

this commit was splitted into small chunks to avoid
a limitation of cvs.  ("Protocol error: too many arguments")

... if called prior to using_history().  This needed to be worked around
in PHP: http://git.php.net/?p=php-src.git;a=commitdiff;h=31d67bd3

Bug: https://bugzilla.redhat.com/show_bug.cgi?id=1055409

get rid of PATH_MAX.

resync from head

expose rl_catch_signals and explain what we are doing.

resync with head

sync with head

explicitly pass (void *)0 instead of NULL.

Resync to 2012-11-19 00:00:00 UTC

sync with head

Add trailing NULL's to the varargs functions as required. (John Spencer)

PR/46678: Ian Wienand: Add stub implementation for rl_free_line_state()

Pull up following revision(s) (requested by christos in ticket #309):
	lib/libedit/readline.c: revision 1.104
don't crash if add_history is called from an empty line. Called from
nslookup in new bind.
XXX: pullup to 6

don't crash if add_history is called from an empty line. Called from
nslookup in new bind.
XXX: pullup to 6

sync with head.

define the new variable

Add rl_completion_word_break_hook from:
http://gitweb.dragonflybsd.org/dragonfly.git/commitdiff/\
5ec6a45fa762b8cbd0305ca06acb8780335a486a

sync with head

Use C89 functions definitions.
Remove use of __P

Add missing *

re-enable -Wconversion

pass -Wconversion

- fix unused params
- unconditionalize vis.h

KNF return (\1); -> return \1;

kill ptr_t and ioctl_t, add  * sizeof(*foo) to all allocations.

handle systems without getpwent_r

use vis.h from .

unbreak readline history.

setup a callback to be invoked on resize buffers so that readline can
reset rl_line_buffer which unfortunately some applications use it directly.

provide rl_on_newline

From Jess Thrysoee
    - Fix wint_t to Int confusion

rename historyw -> history_w for consistency.
add wide tst code and make it the default.

Fix wide build, test it, but don't turn it on yet.

Wide character support (UTF-8) from Johny Mattsson; currently disabled.

apply apple patches from:
http://opensource.apple.com/source/libedit/libedit-11/patches/

Fix memory leaks in error paths.
From Michael Cook mcook at bbn dot com

Sync with HEAD.

Third (and last) commit. See http://mail-index.netbsd.org/source-changes/2009/05/13/msg221222.html

Fix off by one error reported by:  Caleb Welton cwelton at greenplum dot com

implement RL_PROMPT_{START,END}_IGNORE

more size_t stuff.

pass lint on _LP64.

More fixes for existing portability stuff.

add rl_set_prompt

fix -Wsign-compare issues

- insert a space after the recognized string if it was an exact match
- initialize properly the string used for completion.
From Alex Bligh alex at alex dot org dot uk
- Make char constants consistent

sync with head.

Convert to new 2 clause license

add EL_REFRESH for the benefit of readline

Add rl_forced_update_display() from Gerry Swislow

patches from Axel Thimm

Add rl_completion_matches, fix remove_history

- Add more readline functions, enough for gdb-6.5
- Make el_get varyadic, and implement EL_GETTC.
- XXX: the EL_SETTC api will change in the future.

Change to a 3 clause copyright after permission of the holders.

PR/34062: Tanaka Akira: rl_deprep_term_function is NULL in libedit.
Default to rl_deprep_terminal as suggested; do the same for
rl_prep_term_function

remove if/free block checking known condition

coverity 2762 / run 11

Coverity CID 2743: Not really a memory leak, but make it obvious that we
always free tmp.

Lint comment.

Coverity CID 1666: Plug memory leak.

Coverity CID 1667: Plug memory leak

Coverity CID 1662: Memory leak.

use the tty chars for reprint and eof instead of hard-coded ^R and ^D

PR/32817: Magnus Svensson: write_history and read_history returncode is not
readline compatible.

Partial rl_getc_function support from Jess Thrysoee.

PR/30747: David N. Williams: libedit is missing remove_history()
Added, please test.

Pull up revision 1.52 (requested by lukem in ticket #542):
check for pwd != NULL, fix a missed getpwnam.

Pull up revision 1.50 (requested by lukem in ticket #542):
use getpwent_r

PR/30500: Paul Shupak: Inconsistent definition of tilde_expand().
Provide a layer of indirection between the readline compatibility functions
and our internal implementation, so that we have the freedom to change the
function signature.

Bug reported from Martin Dietze:

    The place to change the completion_append_character is
    usually somewhere in the `rl_completion_entry_function'
    callback which is where one usually can distinguish between
    file- or dir-like entries to append a slash for dirs etc.

    This does no longer work since `fn_complete()' takes the
    `append_character' as argument before the callback is executed,
    so that changes to the variable `rl_completion_append_character'
    have in fact no effect for the current completion.

Fix by adding a function that returns the rl_completion_append_character,
when it gets passed in a filename in readline emulation.

Pull up revision 1.55 (requested by agc in ticket #353):
Use the correct type for the stored callback function

Pull up revision 1.54 (requested by agc in ticket #353):
Sync the alternative readline interface with reality:
+ the rl_callback_handler_install takes a pointer to a void function
  which has one char * argument (it's called that way in the readline
  emulation source, otherwise there's no way to pass the line buffer
  to the function which processes the line when EOL is encountered)
+ provide a prototype for that function signature and use it
Makes the callback readline interface work now.

Use the correct type for the stored callback function

Sync the alternative readline interface with reality:

+ the rl_callback_handler_install takes a pointer to a void function
  which has one char * argument (it's called that way in the readline
  emulation source, otherwise there's no way to pass the line buffer
  to the function which processes the line when EOL is encountered)

+ provide a prototype for that function signature and use it

Makes the callback readline interface work now.

Separate out the filename completion functions from the readline() code.
Pass in loads of parameters instead of relying on shed-loads of global
variables to modify the behaviour.
The filename completion code can now be enabled by code that uses el_gets().
(eg /bin/sh)

check for pwd != NULL, fix a missed getpwnam.

PR/29958: Peter Bex: add rl_variable_bind and rl_attempted_completion_over

use getpwent_r

Always update the position variables before the map function is called.
From Rob Rodgers, thanks!

set UNBUFFERED again after you do the line callback so that the new line
gets refreshed.

cut out the middle-man and use el_insertstr() directly.

Better fix for rl_prep_terminal() from Gerry Swislow.

portability fixes.

Tokenization function enhancements:
* Make tok_init(), tok_end(), tok_reset(), tok_line() and tok_str()
  publically available in <histedit.h>
* Documented the public functions in editline(3)
* Renamed tok_line() -> tok_str()
* Added new tok_line() which takes a "const LineInfo *" instead of
  "const char *" (the former has "cursor" information), and optionally
  return the argv index ("int *cursorc") and offset within that index
  ("int *cursorv").  This means that completion routines can use the
  tokenization code to crack the line and easily find which word the
  cursor is at.  (mmm, context sensitive completion :)
* Fixed TEST/test.c when using "continuation" lines (unmatched quote
  or \ at EOL), and added some more DEBUG messages including highlighting
  where the cursor is (with a `_').

Fix uninitialized variable.

From: Gerry Swislow gerry at certif dot com

1) File name completion should list the files in the current directory
if no text is entered.  The previous version wouldn't list anything if
the text to complete was empty.

2) When listing directories, the entries "." and ".." shouldn't be
shown.

3) The filename completion should be used if the user's
rl_attempted_completion_function doesn't return any matches.  The
previous version didn't do that.

initialize ptr.

Make readline csh-like history work.
From Gerry Swislow <gerry at certif dot com>.

add rl_catch_signals

More libedit readline emulation functions from: Gerry Swislow
<gerry at certif dot com>

don't limit ^c to alpha c, and add VIS_NOSLASH so that vis(3) does not
produce \^c

PR/23107: Nathan Williams: ^D as the first char on the command line does
not DTRT in readline compatibility mode

Implement enough of readline's 4.0 async mode to make gdb happy. This is
not complete yet, but it seems to work...
This required to introduce an unbuffered mode to el_gets(), but that was
a minor change.

Match the stupid function pointer declarations with actual readline's 4.0.
This is gross.

make this compile.

- provide enough hooks to compile gdb-5.3
- fix el_get(e, EL_TERMINAL, (char **))

From michael@moria.de:
Fix realloc case where we could be running out of space if too many matches.

From michael@moria.de:
- use __attribute__((__unused__)) in arguments where appropriate.
- some int -> size_t and char * to const char * conversions.

Consistently spell occurrence with two rs.

null is not 0

include alloca.h for systems that need it.

fix directory descriptor leak [from michael at moria dot de]

remove unused  variables.

vi mode and memory fixes from david laight.

don't crash in memory shortage conditions.

_rl_qsort_string_compare(): Fix casts.  This makes gcc 3.2 happy,
and also allows both LINTED comments to be removed.

Add a couple linted comment and enable WARNS=3

- constify; passes all gcc and lint strict checks.
- add config.h [Jason Evans], to create a portable version of libedit that
  can be easily compiled on other OS's.

fix partial completion - we failed to refresh screen in that case

depoison the pure editline code from readline compatibility hacks.

Standard location of readline headers is /usr/include/readline/, so install
them there.
readline.h of libedit had to move to subdirectory 'readline', due to the way
BSD makefiles work; this is better than potentially fragile Makefile hacks

fix lint problems.

Add support for rl_completion_append_character and rl_special_prefixes.
This addresses lib/10513 by Giles Lean. Tested with PostgreSQL 7.0.2 psql.

history_tokenize(): fix one off-by-one bug
rl_complete_internal(): only replace the completed string with common part
	of possible matches if there is a possible completion

rl_display_match_list():
	* pad entries shorter than 'max' by spaces correctly
	* fix off-by-one error which caused extra newline to be printed
	  if the list fit exactly to a screen
	* fix typo in _rl_qsort_string_compare, which caused the list to not
	  be sorted after all

completion_matches(): fix a off-by-one bug, fix variable name typo
implement displaying of possible completions, add hook to display the list
	on second rl_complete() invocation in row (typically, double <TAB>)

This addresses the completion part of lib/11581 by Richard Earnshaw.

convert to new style guide, which includes:
- ansi prototypes & features (such as stdargs)
- 8 space indents

Fix _rl_compat_sub() to really honour 'globally' flag instead
	of making the substitution always globally - affects
	_history_expand_command() and in turn history_expand()

All praise lint :)

readline(): don't dereference NULL pointer if prompt is NULL
history_tokenize(): avoid stepping too far if backslash is last character
	on the passed string
update some comments

This makes gdb \ escaping work reliably and fixes lib/9511 by Assar Westerlund.

don't try to free() something allocated with alloca() in rl_complete_internal().
(noticed while using completion in gdb and getting inundated with warnings from
the new free())

More trailing white space.

Make this compile under linux

many problems; variables hidden by others, size_t <-> confusion

Adjust to the libedit api changes.

el_gets() takes an int *, not a size_t *.

Const de-poisoning :-( Unfortunately the default gnu readline does not
have full prototypes... With those changes, and a single line change in
gdb/top.c, gdb links with -ledit

PR/4301: Jaromir Dolecek. Add gnu-readline wrapper for editline.