5 * There's been changes in padding/spacing usage in styles. Frames
6 now surround stuff inside borders not with `spacing` pixels, but
7 with `padding_pixels. Spacing is only used to space things within
8 the borders (all the tabs and the client window from each other).
9 Consequently, custom styles may need to be updated to reflect this,
10 by increasing the padding, possibly also altering/removing colour
11 (to set it to background colour).
13 * The use of drawing engine attributes has also changed, and strict
14 ordering of them in the styles is no longer necessary. Additionally,
15 much more attributes are supported, including the name of every
18 * Xinerama support is gone. In addition to being problematic as such,
19 it had bit-rotted, and I will not waste time fixing it.
21 * `ioncore.set` no longer supports the `default_ws_params` parameter
22 for configuring the default layout. It is replaced by
23 `ioncore.deflayout("default", { ... })`, which allows configuring
24 other layouts as well, known by `mod_query.query_workspace` (F9).
25 Some default layouts are configured in `cfg_layout.lua`.
27 * It is now possible to automatically create new workspaces for
28 windows with the `new_group` winprop.
34 There's nothing major in this release, primarily just some minor fixes
35 and tuning to the previous release, that it was time to release.
41 Mostly this release still fixes issues in the big 3ds-20061015 release,
42 but in addition there are some improvements in the query department:
44 * Query activation key now cycles completions (So e.g. the
45 in the context menu activated with META+M, this same key
46 can be used to cycle through the alternatives.) This does
47 not work for queries activated by submap bindings.
49 * Likewise, it is no longer necessary to specify the key to
50 use for cycling for `mod_menu.grabmenu`.
52 * Control+R can now be used for history completion in queries.
53 (Currently matching is done for full string up to 'point', but
54 this may be changed to substring match.)
56 * Note that the parametrisation of WEdln.complete has changed,
57 and the second cycle parameter must be 'next' now instead of
64 Fixes some (expected, but minor) issues in the previous release.
70 * WIonWS and WFloatWS and the corresponding modules are gone, and
71 your custom configuration files will be broken with regard to these.
72 However, a partial backwards compatibility hack exists for layout
75 The F9 and META-F9 bindings now by default create workspaces with
76 a tiled layout of two frames. To create an "empty" workspace,
77 corresponding to the old WFloatWS, use the context menu (META-M)
78 and choose "new-empty-workspace". It is also possible to change
81 * Note that if you restart from an old version to this new version of Ion,
82 transients will stop working as expected for already existing windows.
83 They will work for newly-created windows, however.
85 * There are a few new sets of binding (including one for `WClientWin`!),
86 and some old bindings may not work exactly as expected anymore. In
87 particular, those for switching to full-screen mode.
89 **It is probably best to start from scratch with your custom
92 * `WFrame.set_tabbar` is gone. If you absolutely want to get rid off
93 the tabs, you must change the frame's "mode" with `WFrame.set_mode`.
94 The mode "tiled-alt" has been intended for this, and the corresponding
95 "framed-tiled-alt" style defaults to `bar = "none"`.
97 * The rather popular `detach.lua` script from the scripts repository
98 is obsoleted now, as Ion includes detach functionality in itself.
99 To detach a window, use META-K D in the default bindings. To tile
100 an existing frame from a workspace that doesn't have a tiling yet,
107 Some notable changes in this release include
109 * Lua 5.1 is now required.
111 * Framed transients on by default now. New binding contexts
112 "WFrame.toplevel" and "WMPlex.toplevel" were added to allow
113 for separate sets of bindings for nested transient frames and
114 top-level frames. Some of the bindings in the default binding
115 maps that are likely to be unwanted on transient frames were
116 moved to these contexts. Old custom bindings will continue to
117 work unless they modify the defaults by unbinding some of the
120 * Pressing Mod1+K K in the default bindings now switches to any
121 region with the "activity" flag set (indicated by the a box
122 at a corner of the screen), if there's one, before cycling to
123 previously active region. The same effect can be achieved in
124 your custom bindings with
126 ioncore.activity_goto() or ioncore.goto_previous()
132 This release mostly features minor bugfixes and other improvements. The most
133 visible non-bugfix changes are:
135 * Faster "fontset" filling kludge: only `-misc-fixed-*` fonts are tried. Ion
136 should now load faster in UTF-8 environments, and usually with no less
137 fonts in the fontset than before. (UTF-8 string drawing still does not
138 fully utilise these fonts under XFree86, but it does under XOrg.)
140 * Experimental auto-show-completions support, which is also on by default
141 now. In this mode the Tab key can be used to cycle forward through the
142 completions, and Shift+Tab backwards. Modify the settings seen in the
143 new `mod_query.lua` to get normal Tab-completion, or change the completion
146 * The release scripts do not run automatically autoconf anymore so maybe
147 a few more people would look into the README first.
153 The most visible changes in this release are:
155 * Mod1+space now toggles the scratchpad by default and Mod1+D the dock.
157 * The `ion-(ssh|man|view|edit)` wrapper scripts were removed and instead
158 the programs used can be configured in Ion's configuration files.
160 * `ioncore.exec(_on)` also support the `:cmd` notation for "run cmd in
161 xterm" familiar from the run query. The `::cmd` notation can be used
162 to ask for enter to be pressed even when the program quit succesfully.
164 * Those with custom configuration files should note that many exported toggle
165 functions were changed and renamed, and now accept a string parameter
166 incidating whether to toggle, set or unset the property.
168 - `WClientwin.set_fullsreen` (replaces `WClientWin.toggle_fullscreen`)
169 - `WRegion.set_tagged` (replaces `WRegion.tag/untag/toggle_tag`)
170 - `WFrame.set_tabbar` (replaces `WFrame.toggle_tabbar`)
171 - `WFrame.set_shaded` (replaces `WFrame.toggle_shade`)
172 - `WFloatFrame.set_sticky` (replaces `WFloatFrame.toggle_sticky`)
173 - `WMPlex.l2_set_hidden` (replaces `WMPlex.l2_hide/show`)
174 - `mod_sp.set_shown(_on)` (replaces `mod_sp.toggle(_on)`)
175 - `mod_dock.set_floating_shown_on` (replaces `mod_dock.toggle_floating_on`)
176 - `WRegion.set_activity` (replaces `WRegion.clear_activity` and
177 `WRegion.notify_activity`)
179 For example, `WRegion.set_tagged(_, 'toggle')` should be used in place of
180 `WRegion.toggle_tagged(_)` now.
182 Obviously there are some other changes and fixes too. See the changelog
183 for details as usual.
189 This is mostly a bug fix release, but also features improved
190 `mod_query.query_exec` (F3 key) completion support. Script writers should be
191 aware that As a side effect of one of these bug fixes, many hooks are now
192 called in "protected mode" and can not call any functions that modify the
193 internal state of Ion, except ioncore.defer.
199 The most important changes in this release are:
201 * So-called "placeholders". With the help of these the positions of
202 full-screen windows are remembered in their original frames, and don't
203 just get inserted after currently active window when returning from
204 full-screen mode. Under a session manager placeholders are also used
205 to remember the original order of windows.
207 * Improved `mod_statusbar` and `ion-statusd` communication, and colouring
208 of important/critical meters.
210 * A number of small fixes.
216 This monthly snapshot adds a few new and improved features.
218 * 'Grabbed menus' that have a single cycling key and activate selected
219 entry when all modifiers have been released. See
221 <http://iki.fi/tuomov/repos/ion-scripts-3/scripts/wcirculate.lua>
225 * Potentially blocking status meters are now in a separate ion-statusd
226 program. Please write your additional status meters that do not monitor
227 the state of Ion itself for ion-statusd (and contribute them in the
228 Ion3 scripts repository at <http://iki.fi/tuomov/repos/ion-scripts-3/>).
229 For help on writing such status meters, see e.g. source for
230 `statusd_load` in `ext_statusbar/ion-statusd`.
232 * Floating splits are now supported on plain tiled workspaces as well as
233 on pane workspaces. To create such a split, use the workspace context
234 menu (Mod1+M by default) or write your own bindings.
236 * Line editor now supports history search; Control+Up/Down only scrolls
237 through history entries with matching initial part.
239 * Arbitrary winprop matching criteria is supported. Lua scripts have
240 access to X properties.
242 Of course there are some other minor fixes and improvements as well.
248 This release finally includes a usable (yet still incomplete) version
249 of the `mod_panews` workspace module (that has also bored the name 'autows'
250 and 'rubberws' previously). The final outcome is not exactly what I
251 initially planned, as those plans turned out not to be that workable.
252 Instead, what `mod_panews` does is add to the basic tiled ionws approach
253 overlappable splits and automatically filled panes that are initially
254 empty. Each window is classified and assigned to the pane matching that
255 classification. (The default classifications are T(erminal), V(iewer)
256 and M(iscellaneous).)
258 For a better feel for it, try it out yourself and please give feedback.
259 The overlapping panes may be a bit confusing (this is one place where
260 true translucency might actually be useful and not just eyecandy) at
261 first, but you'll get used to them if you use the feature, and e.g.
262 the Gimp works quite splendidly with the toolboxes in the 'M' pane and
263 image windows in the 'V' pane -- although the simple initial size-based
264 classification heuristics don't always get it right and overriding
265 winprops (setting: `panews_classification`) should eventually added.
267 In addition to the new module, this release adds support for translations
268 of program messages and the manual page, regardless of whether such is
269 of any use in a program like Ion or not. (Currently Finnish and Czech
270 translations are available.) Of course there are some bug fixes and
271 other minor additions as well, and the `./configure` script is also back,
272 the abandonment of libtool being final now.
278 The first thing you'll notice when you start up this release of Ion is
279 that isn't reading your old configuration that. The next thing you
280 should notice is a neat layout-adapting statusbar at the bottom of the
281 screen. That's right, Ion now includes an `ext_statusbar` Lua script that
282 is enabled by default and displays date/time/load/mail count in a
285 So why is it not reading your configuration and save files? Firstly,
286 all of the `.lua` files were renamed to be indicative of their purpose.
287 Secondly, there have been so many changes that your old files would be
290 The `.lua` files are now named as follows:
292 cfg_*.lua Configuration file that the user may wish to edit
293 look_*.lua Drawing engine style file
294 saved_*.lua Save file
295 mod_*.lua Module stub loader
296 ext_*.lua A bigger Lua extension without a C counterpart
297 so that it is not a module
299 The configuration file for `mod_foobar` or `ext_foobar`, if it has one, is,
300 of course, `cfg_foobar.lua`. In the topic of file names, also note that
301 the default installation paths and binary names have changed to include
302 the component '3' to reflect the situation with many binary packages
305 You perhaps noticed above that modules have stub loaders now, so the
306 user has no need to use the `ioncore.load_module` routine. All extensions
307 and additional configuration files can now be loaded with 'dopath'
308 (used to be 'include'). Also, the `menulib`, `querylib`, and `ioncorelib`
309 Lua libraries are gone and instead their contents can be found in the same
310 `mod_whatever` namespace with the corresponding module (the "stub" loader
311 for these modules is a bit more than just a stub...). Some may also
312 want to know that for `mod_*` and `ext_*` only the compiled .lc files are
313 now installed, and not the source `.lua` files, thus removing redundant
314 files and making the installation slightly smaller.
316 There are few other changes to the contents of the configuration files
317 as well, so you're probably best off simply rewriting your modified
318 configuration files based on the new defaults. There quite likely won't
319 be any more _big_ changes to the configuration files before the release
320 of final Ion3, wherever that will be. (Most likely we won't see 3rc:s
321 yet this year.) However, some functions and variables are still likely
322 to be renamed or changed.
324 In addition to being renamed, the layout savefile of this Ion release
325 is incompatible with older releases. This is due to the changes made
326 to the `WIonWS` (and `WAutoWS`) split tree code to make it more modularly
327 extensible with the new kinds of nodes that `WAutoWS` requires.
329 The `mod_autows` module has infact been through many changes since the
330 last release, and I think I have the final form of it finally figured
331 out. However, it is still far from finished and unlikely to be ready
338 The major new features of this snapshot release are:
340 * `WMPlex` support for a sticky status display area to which `WIonWS`s
341 adjust properly. Modified to the dock module to support this method.
342 See the new `dock.lua` to set up the dock in the new embedded or old
343 floating manner (the API for the latter has also changed).
344 * Primitive session management support.
346 There are also some bug fixes and many minor improvements; see the ChangeLog
347 for details. Work on `mod_autows` has also started, but it isn't ready for
351 ### Note on ./configure:
353 ./configure does not exist for the moment as the source autoconf script
354 is not up-to-date. Ion no longer uses libtool/libltdl due to problems
355 with inter-module dependencies, and I have not asked Tom Payne to fix
356 the script yet, as this change may not be final (although that is most
357 likely the case; those without Linux-compatible libdl and a flexible
358 binary format such as ELF will just have to link statically against the
362 ### Notes on session management:
364 Ion loads `mod_sm` automatically when the SESSION_MANAGER environment
365 variable that should be set by the session manager is set, so there's
366 no need load it in `ion.lua`.
368 When session management is in use, all entries in the 'Session' menu
369 (previously 'Exit' menu) actually invoke the session manager to do the
370 task. 'Save' asks the SM save the whole session and 'Exit' (ioncore.shutdown)
371 asks it to shut down the session instead of just causing the WM to quite.
372 To do the latter, use ioncore.resign.
374 Unfortunately, all session managers I have tried (from Debian/unstable),
375 are broken/incomplete in one way or another:
378 xsm: doesn't support any requests from applications. This makes Ion's
379 session menu complete unfunctional. The only way to restart/exit/save
380 state is through xsm's window.
383 gnome-session: This seems the most complete of the all the session managers
384 and works fine until it is requested to shut down the session, when it dumps
385 core and session state is lost if was not explicitly saved previously.
387 A `~/.gnome2/session` file to use with Ion follows:
393 0,RestartCommand=gnome-smproxy --sm-client-id default0
396 1,RestartCommand=/usr/local/ion-3/bin/ion -smclientid default1
399 ksmserver: Only supports a global shutdown request, so that Ion can not be
400 restarted or session state saved in the middle of a session.
406 This is the first development snapshot release of what is to be Ion3.
407 The most visible changes to Ion2 so far are:
409 * Default installation directory is `/usr/local/ion-3` while user
410 configuration files go in `~/.ion3`
412 * The `mod_sp` module was added. It creates an extra toggleable
413 "scratchpad" frame on each screen. Toggle is for now bound by default
414 to Mod1+section, which should be very conveniently located on most
415 Nordic keyboards (left of "1"), but you may have to change it. The
416 scratchpad should be nice for xconsole or similar monitors, xmessage
417 and other popups and just as temporary holding space for windows.
419 * All modules except the drawing engine are now called `mod_something`.
421 * Man-page complection supports a cache of known man-pages for faster
422 completion. See the README for instructions on setting up a cronjob
423 or manually generating the index if you want to use the feature.
424 Also `query_man_path` is no longer used. Instead we try the
425 `ION_MC_MANPATH` and `MANPATH` environment variables and the
428 * Exported functions are now separated into tables (namespaces)
429 instead of cluttering the globals table. Some frequently used
430 configuration functions are imported into the globals table, though.
431 Some functions have also been removed or renamed for simpler and
432 more consistent function set.
434 * New binding configuration mechanism. Dozens of `*_bindings` functions
435 were replaced with a single `[ioncorelib.]defbindings` function that
436 accepts a context parameter. Callbacks are specified as strings
437 (although passing functions still works) to make it easier for
438 external configuration programs to understand the configuration
439 files and perhaps remove some confusion among users who do not
440 care to read a tiny bit of Lua documentation and understand the
441 concept of anonymous functions.
443 * It is now also possible to retrieve a list of bindings with
444 `ioncorelib.getbindings`, for example, for self-documentation.
446 * Single move/resize mode bindings instead of separate for both
447 ionframe and floatframe.
449 There are also quite a few internal changes; see the ChangeLog for
450 details. No documentation tarball is available at the moment as the
451 documentation is out-of-date except for the function reference. If
452 you need the reference, just checkout the documentation from the
453 Subversion repository with
455 svn co http://tao.uab.es/ion/svn/ion-doc/trunk ion-doc-3ds
457 and build the documentation.
463 This release is finally what can be called "Ion 2 release candidate #1".
464 No more new features will be added to "Ion 2" after this release, and the
465 configuration interface has already been frozen for a while. I will wait a
466 couple of weeks for bug reports, and if nothing serious is found, the new
467 stable Ion should finally be released then.
469 The most notable changes since the previous release are:
471 * The dock module is included
472 * An optional autoconf script was added
473 * A few minor bugs were fixed
474 * Some incomplete features were polished, especially focus control on
476 * Changes in X keyboard map are supported now (so e.g. switching to
477 dvorak after Ion has started should update the bindings to match the
478 locations of symbols in the dvorak layout).
481 About the version numbering scheme:
483 Due to a demand of a version numbering scheme more indicative of the status
484 of the project, I was thinking of various different new version numbering
485 schemes for this release: ion-2rc1 (then 2r1, 3d1, etc.), ion-2-20040114-rc1,
486 ion-2.20040114rc1, ion-20040114-2rc1, etc.). It would've been nice if
487 simple lexicographical sorting could be used to order the packages, but
488 in the end I decided to stick with a scheme that is consistent with the
489 'ion-devel' package names:
491 project-branch_and_status-release_date
493 Therefore this release is 'ion-2rc-20040114', and the "stable" one will be
500 This is a big clean-up release. The most noticeable changes are that
501 '-devel' has been removed from path names, user configuration files go
502 in `~/.ion2/` and the main configuration file is 'ion.lua'. (The other
503 `ioncore-*.lua` files have also been renamed.) There have been no notable
504 changes in the configuration files themselves, so your old files will
505 work if you move them to the correct directory and rename the changed
506 files. (`mv ~/.ion-devel ~/.ion2; cd ~/.ion2; mv ioncore.lua ion.lua;`
509 This release also finally contains a working PWM binary. (The ioncore+
510 scripts scheme was replaced with separate binaries statically linked
511 to ioncore.a.) Floating workspaces now support edge snapping and
512 sticky windows, but some PWM features are probably still missing.
514 Some bugs were also fixed and users are now force-fed the manual page
515 the first time Ion is started. Transients can be toggled between
516 top/bottom of the main window with Mod1+K T. The SSH query uses
517 `~/.ssh/known_hosts` for completion instead of a manually defined list.
519 There may be a few additions (e.g. the dock) and bug fixes, of course,
520 before the stable release, but this release should pretty much be what
521 the new "stable" Ion should look like. Please upgrade to it to help
522 weed out the bugs. Note that you need to upgrade Lua to the 5.0.1
530 It's finally time for a new release on Ion's development branch. Most
531 likely this will also be the last "big" release before finally, after
532 almost two years, releasing a new "stable" version of Ion. A few bug
533 fix and code clean-up releases should appear in between, though.
535 The most important additions, changes and non-changes in this release
539 * A little less broken extended character set and string encoding
541 * Extended WM hints fullscreen request support
542 * Hopefully a little clarified configuration file layout
544 * And most important of all:
546 > Configuration files written for the previous release should
547 > still work this time!
549 (However, Ion will complain of old drawing engine styles.)
551 The menu module provides both drop-down menus (in the stock configuration
552 pressing Button3 on a tab should show a context menu) and query-like
553 "in-mplex" menus (F12: main menu at screen level; Mod1+M: the same context
554 menu as above at frame level).
556 What was done enhance support for strings (mostly window titles) in
557 languages that need more than 8-bit character sets is:
559 * Remove UTF8 restriction and support almost arbitrary multibyte
560 encodings instead. (Statefull encodings will not work and combining
561 characters can cause clutter. Both are a sign of bad encoding, IMHO.)
563 * The XCreateFontSet routine that is used to load fonts in the kind
564 of structure the X utf8 and multibyte routines want apparently wants
565 to be able to load glyphs for all character sets specified in the
566 locale and therefore often fails if only single font is specified.
567 Therefore a kludge was added that tries loading more fonts while
568 keeping the fonts' size the same. (The ",*" kludge could load huge
569 fonts). Unfortunately this has a noticeable effect in startup time :(.
571 * There is no longer a system.mk option to enable utf8 (now multibyte)
572 support, but must still specifically be enabled with the -i18n
573 command line switch, mostly thanks to troublesome utf-8 locales.
575 Because Xlib's UTF-8 string drawing code is broken and very unlikely to
576 be fixed, decent text output in an UTF-8 locale is still unlikely and
577 dependent on the fonts loaded in the system. The Xmb routines that are now
578 used always use an iso10646-1 font even if other fonts are loaded and the
579 unicode font does not contain a particular glyph to be drawn. The use of
580 the Xutf8 routines in an utf8 locale can be enabled with the
581 `CF_DE_USE_XUTF8` compile-time option. The advantage of this is that
582 these routines seem to choose the font to draw particular character in
583 most cases more sanely. The downside is that there are other more serious
584 problems with unknown characters.
586 All in all, even if I'd like to support a universal move to UTF-8, thanks
587 to Xlib brokenness I can't really recommend using UTF-8 locales with Ion
588 and the default drawing engine if there's a decent alternative encoding.
589 Most if not all other stateless encodings shouldn't have the problems
590 UTF-8 has. An alternative drawing engine that didn't use the Xlib i18n
591 string drawing routines might solve the troubles with utf-8 support.
595 If you have saved a custom system.mk, you'll need to set LUAC
596 point to the Lua compiler there as some of the share/ files
603 It's been a while since the previous release as I wanted to freeze the
604 Lua configuration/scripting interface for this release and therefore
605 finish work on a few things. Well, on my part the interface is frozen,
606 but I will still accept constructive complaints for a short time and
607 after that the implemented parts of the scripting interface will be
608 frozen. But in all likelihood, if no one has anything else to say, but
609 "it's ok", "it sucks", the interface will no longer change, only possibly
612 What's new for this release then?
614 * Drawing engine module support.
615 * OO-style exported functions. The old `class_fn` functions are now
616 `WClass.fn` (an ugly wrapper is provided in `compat.lua`).
617 * No more screen-specific configuration or savefiles.
618 * Session name (can be specified on command line) instead of display
619 name based savefiles. Also affects query history and not just workspace
621 * Bug fixes, most small and one bigger (almost complete rewrite of
622 the split resizing algorithm).
623 * Some documentation improvements.
625 Converting configuration files.
627 Old `.lua` colour schemes and workspace save files can be automatically
628 converted to a format suitable to be loaded by this latest release.
629 Other files will have to be ported manually. To convert colour schemes,
630 use the script `utils/lookconv.lua`. The usage is
632 lookconv.lua look-old.lua > look-new.lua
634 To convert workspaces savefiles, use the script `utils/saveconv.sh`.
637 saveconf.sh ~/.ion-devel/saves/workspaces-DISPLAY.*.lua \
638 > ~/.ion-devel/SESSIONNAME/workspaces.lua
640 `DISPLAY` here is the actual display part of `$DISPLAY`, probably just `:0`.
641 '*' stands for all screens (if you have only one screen you could do with
642 perhaps just `workspaces-:0.0.lua`). SESSIONNAME is the name of the session
643 where you want to use the converted savefiles. Default session name for
644 `DISPLAY` is `default-session-DISPLAY` with the colon in `DISPLAY` converted
645 to a dash. For most people this is `default-session--0`. In the simplest
646 case the whole command line is therefore
648 saveconv.sh ~/.ion-devel/saves/workspaces-:0.0.lua \
649 > ~/.ion-devel/default-session--0/workspaces.lua
651 The scripts may or may not work and are only provided as a potential
652 convenience that will not be maintained and will be removed eventually.
658 Quite a few bug fixes and one most likely unnoticeable improvement:
662 * Fixed pointer warping on screen change.
663 * A bug in grab handler calling code could crash Ion when leaving
664 keyboard resize mode manually.
665 * Resize display was showing incorrect values for keyboard resize.
669 * Client window last height request bookkeeping code had been lost
670 when configure request policy was changed. This caused transient
671 sizes to be calculated incorrectly.
672 * Return from full screen mode to floatws had been broken.
673 * As the number of dynamic functions has been getting bigger, the
674 functions are now sorted on first use and then binary-searched
675 instead of naive linear searching.
676 * Screen lookup had been broken for windows that are not properly
683 Just some bug fixes and minor behavioural changes and improvements in
690 This release fixes some small problems with the previous release and
691 adds a workaround kludge for the XFree86 textprop bug (it's been fixed
692 but no release with the fix is available) that could cause Opera to
693 crash Ion when UTF8 support was enabled. I also added the winprop
694 needed for galeon's find dialog to default `kludges.lua` and there's
695 some extra documentation and defaults for some systems in system.mk.
701 Most changes in this release centre around making Ion more tolerant
702 to broken configuration files; for details see the ChangeLog. There
703 are also changes in binding configuration as I already mentioned in
704 an earlier poting. Namely `common-frame-bindings.lua` is gone and the
705 bindings previously set there (into variables that the `*ws.lua`
706 files later referenced) were moved to `ioncore-bindings.lua` and are
707 set using the new functions `mplex_bindings` and `genframe_bindings`
708 are were added. Old modified configuration files should still work,
709 however, but if you use `make_active_leaf_fn`, you will need to
710 include `compat.lua` as the `global_bindings` that used this were
711 replaced by bindings in `mplex_bindings` and the function
712 `make_current_clientwin_fn`.
714 There are also some (minor) bug fixes and a few other improvements
715 worth a mention. In particular all regions are now given names of
716 the form `WFoobar<n>` by default and `DEFAULT_MOD.."F9"` was bound to
717 create a new workspace without asking for a name. This binding and
718 `QueryLib.query_workspace` use the workspace type defined in the
719 variable `default_ws_type` instead of being hardcoded to `WIonWS`.
721 All objects passed to Lua now have a unique userdata (a `WWatch`
722 cached in a weak table) so they can be used e.g. as indices to tables.
728 This release unifies some parts of `WScreen` and `WGenFrame`, which makes
729 screen-level queries possible. (Later the queries for small frames might
730 be changed to be shown at screen level.) The statusbar restriction from
731 ion-devel-20030531 was also removed by this some change and a few
732 non-fatal bugs have been fixed.
738 A bug was discovered:
742 * An off-by-one error in `extl_l1_finalize` caused references to some
743 Lua tables (including large completions) never to be released.
745 It could be that this bug was causing some other errors in Lua as well.
746 Sometimes QueryLib bindings failed because some function generators
747 returned nil although they shouldn't. Adding dummy lines (!) in those
748 functions fixed the problem and for some reason this patch also seemed
749 to remove those problems.
752 One more note: as `ioncorelib.lua` and `querylib.lua` are now installed
753 in `$SHAREDIR`, you must remove the old files in `$ETCDIR` if installing
754 over a previous release or else there will be errors. You should also
755 not use any possible old copies you have in `~/.ion-devel/`. (One of the
756 reasons for moving these to `$SHAREDIR` is to stop users from making
757 copies of them as they are not configuration files.)
763 As the subject line says, Ion-devel-20030531 was just released. The short
766 * The license was changed from the Clarified Artistic License to the
767 GNU Library/Lesser General Public License (LGPL).
768 * Screen, viewport and root window renames and other changes (see below).
769 * Some installation directory changes; the ion-* helper programs are now
770 installed either in `$SHAREDIR` or `$EXTRABINDIR` and QueryLib searches
771 for them on the script path (`~/.ion-devel/`, `$ETCDIR`, `$SHAREDIR`,
772 `$EXTRABINDIR`) instead of assuming them being on `$PATH`.
773 * Client windows are now in a separate namespace.
774 * Resize/maximize/shade changes: Shading should work on `WIonFrame`s too
775 and shade mode is automatic when the client area gets too small when
776 resizing. Maximize toggle restores frame to previous size if shaded
777 instead of maximizing. The move/resize mode bindings were changed again
778 to be more consistent and predictable: Left/Right/Up/Down and F/B/P/N
779 grow the frame in the specific direction, Shift+keys shrink and in
780 case of floating frames, `DEFAULT_MOD+keys` move
781 * Returning from full screen mode should work on `WFloatWS`s too.
782 * Lots of minor fixes, export additions and clean-up.
784 As usual, see the (long!) list of new ChangeLog entries at the end of
785 this message for details.
788 I've finally personally switched from the old "stable" 20020207 to using
789 the latest development release at home too and I must say that it is
790 finally starting to look good. There are still things to be written, but
791 unless you need "proven" stability, I see no reason to sticking to that
792 old release anymore. Unless, of course, if you find a problem that I have
793 not encountered or simply have no need for the new features and don't want
794 to port your configuration files.
797 On screens, viewports, root windows and problems with the new
800 The objects previously called "screens" are now called "root windows"
801 and what were called "viewports" are called "screens" to better reflect
802 what the user sees. (When Xinerama is not used there's no difference
803 between a root window and a screen, but when Xinerama is used a root
804 window may be split over multiple screens.)
806 This release also creates so-called virtual root windows for each
807 Xinerama screen when there are more than one. This is to better separate
808 the windows on different screens and thus emulate normal multihead, the
809 main difference being that windows can be moved between screens.
810 Especially the virtual roots are there to keep the windows that are on
811 a floatws on the right screen.
813 Virtual root windows will, however, break a few apps:
815 * Mozilla -remote won't work because of its crappy method for looking
816 up existing windows that can't handle multiple levels of WM windows.
817 Nested workspaces have the same problem. There is, however, a simple
818 solution to this problem: gnome-moz-remote. It seems to use a saner
819 method for looking up an existing Gecko browser and also works with
820 at least Phoenix/MozillaFirebird and Galeon.
822 These two are problems only if you use floatws:s (other problems with
823 which virtual root windows are intended to solve):
825 * The background-setting apps I am aware of will require a following
826 'xrefresh' for the changes to be updated to the (transparent) virtual
827 roots. Background setting apps that support the `_NET_VIRTUAL_ROOTS`
828 property (which, I think, was meant for root window-sized WS
829 backgrounds) could be used to set a separate background for each
830 Xinerama screen, though. I am not aware of any such app.
831 * Some apps' resize and move features will behave erratically on those
832 Xinerama screens not at (0, 0) on the root window. This is again
833 because the apps are assuming there's at most one WM window between
834 them and the root and are requesting windows' positions incorrectly.
835 The ICCCM is quite vague on this and I think _all_ apps that I have
836 tried are doing it wrong but the way e.g. XMMS does it certainly is
837 not an interpretation of the ICCCM. The apps that now behave correctly
838 request position for the outermost WM window. Nested workspaces have
841 There is one more temporary problem:
843 * Status bar modules (and the dock module's statusbar mode) will be
844 *temporarily* broken when virtual roots are _not_ used. This will be
845 fixed later when parts of screen and frame code are unified (it should
846 then be possible to run Ion without any workspace modules and have
847 queries attached to screens). In the meanwhile, if you don't care about
848 the above-mentioned problems and want a statusbar (which probably needs
849 to be ported to this release), you can defined `CF_ALWAYS_VIRTUAL_ROOT`
856 There are quite a few small fixes and minor enhancements invisible to
857 the user in this release. The splitting functions were renamed to be
858 more consistent and there are a couple of enhancements in the Lua code
859 query that should make it much more usable. First, tab-completion can
860 now descend into tables and complete subexpressions. Secondly, the local
861 variable `_` in addition to 'arg[1]' is set to point to the the frame in
862 which the query is executing.
864 I have also written some new documentation that is now available from
871 The most notable change in this release is that the `target_id` system
872 was removed and instead client window status is also saved over
873 restarts in the saves/workspaces-* files. (It shouldn't be too hard
874 to hook a session management module over this.) Thanks to this
875 modification, floatframes can also save their status now. Note
876 that if you restart from older version of Ion to this one, client
877 window layout will be messed up.
879 There are also a few bug fixes and code to save and load line editor
886 There are lots of minor improvements in this release, see the
887 new ChangeLog entries for details.
893 Some minor feature enhancements and a few fixes in this new release:
895 The file system scanning completions are now put in the background
896 by receiving the data from external programs through pipes and select()
897 for data in the main event loop. This way long taking completions don't
898 block Ion from processing other events and NFS problems shouldn't hang
901 Basic window stacking management support code was added. Transients
902 on floatws:s should now be stacked above their parents. If you don't
903 want to change your modified binding configurations at this point and
904 need the functions `floatframe_raise/lower`, include `compat.lua`.
910 This release again fixes some minor problems and enhances a few
911 features. Most of the enhancements are related to UTF8 support and
912 the floatws module. If UTF8 support is enabled, Ion now tries to
913 load the "fixed" font at startup after setting up locales. If locales
914 aren't properly set up, this probably fails and Ion will reset locale
915 back to "POSIX" as this might make fonts loadable although support for
916 non-ASCII characters will be crippled. It might now be safe for
917 package maintainers to enable UTF8 support by default
919 As apps seem to have switched to using `_NET_WM_NAME` for UTF8 titles
920 and filling `WM_NAME` with crap, this property is also preferred over
921 `WM_NAME` if set. Support for some other "extended" WM hints might be
922 added in the future but I have no intentions of moking Ion NetWM-
923 compliant. I might have thought of attempting to do so a few years
924 ago when I last read the specification but it seems that since then
925 they've filled it with lots of pointless bloat that may even be in
926 opposition to Ion's goals. Take multi-parent transients, for example.
928 Full error log is also displayed with xmessage on startup whether
929 it is possible to continue or not.
935 There was a bug in QueryLib written yesterday, therefore this release.
941 I started converting the query code to Lua and discovered some rather
942 silly bugs in the Lua interface so here's a new release with the new
943 and improved query code included. The names of queries have changed
944 (they're all in the table QueryLib defined in `querylib.lua` -- only
945 `query_query` and a few temporary handlers are implemented in query.so)
946 so your modified configuration files will be broken again :(.
952 Be prepared to completely rewrite your configuration files once
953 again: a new version of Ion-devel has been released that uses Lua
954 <http://www.lua.org/> for all configuration. Version 5.0 of the
955 language/library is required. Although scripting possibilities
956 are now much better than before, there is one rather big drawback:
957 Lua is not nearly as error-tolerant as Ion's old configuration
958 parser so a syntax error in a file will cause the file not to be
959 executed and may even cause Ion not to be able to start. More
960 comments and documentation on scripting will follow at a later time.
962 Libtool and libltdl are also now used to implement module support.
963 Hopefully this will make compiling Ion easier on a wider range of
964 platforms. (On the other hand, this could also induce new problems.)
966 Finally, the preferred linking address for Ion's home page is now
967 <http://www.iki.fi/tuomov/ion/>. The pages are still located at the
968 old address, but this redirected address should be more permanent.
974 Just some minor fixes, binary rename and better embedded workspace
975 support in this release. I'm still contemplating whether to convert
976 Ion to C++ and haven't written many things that I intended to as
977 that decision will affect how those are implemented.
979 To create an embedded workspaces in a frame, at the moment you have
980 to modify the appropriate workspace save file (when Ion is not running!)
981 by adding lines such as
983 region "WFloatWS", "testiupotus" {
986 inside the frame definition after the other options (flags, `target_id`;
987 those options will be ignored if after any region definitions).
989 One more thing to note: debugging infos are no longer automatically
990 stripped from the ioncore binary or modules by 'make install' so that
991 I don't have to explain how to get proper backtraces every time a bug
992 is found. If you you are confident that you can send me proper backtraces
993 when you find a bug (or maybe can fix it yourself), you can make the
994 binary and the modules a _lot_ smaller by running 'strip' on them.
1000 The list of new ChangeLog entries is long this time, but I'm finally
1001 starting to get to what I started working on almost two years ago. Yes,
1002 there is *experimental* support for PWM-like "floating frame" workspaces.
1003 A lot of the functionality is still missing -- no need to complain of
1004 such at the time -- but the floatws.so module and basic functionality
1005 is there. Most notably perhaps a menu module is not yet implemented.
1006 (Porting the PWM menu code should a nice little task for whomever
1007 interested... *wink*.) To create some PWM workspaces, load the floatws.so
1008 module (should be loaded by default) and in `query_workspace` (F9)
1009 prefix workspace name with 'WFloatWS:' (or load the module before
1012 This release is also finally fully modularised: In addition to the
1013 ion core binary that can not function alone, there are the modules
1014 ionws.so, floatws.so and query.so. It is also possible statically
1015 compile the modules in the core binary if the system doesn't support
1018 There's also *experimental* UTF8 support that must be specifically
1019 enabled from system.mk. You must have XFree86 (4.x?) and C99 wide
1020 char support available (either libc directly or maybe libutf8+libiconv).
1021 Thanks to bugs (?) in some of the XFree86 Xutf8 functions, your locales
1022 must be properly set up or else X will stop drawing strings at non-ascii
1023 characters instead of ignoring them. See my earlier rants on the mailing
1024 list for reasons on not using the more standard Xmb functions.
1025 To actually see any special characters, you must load the necessary fonts
1026 by specifying a comma-separated list of fonts to the font and `tab_font`
1027 draw.conf options. Multiple font loading does not work when Xft support
1028 is enabled at the time being.
1030 There has been a lot of changes in the config files, again, and I can
1031 promise there's still more to come. See the ChangeLog for details.
1032 If you want to use your old workspaces configurations, move them to
1033 `~/.ion-devel/saves` and replace the strings "WFrame" and "WWorkspace"
1034 with "WIonFrame" and "WIonWS", respectively.
1040 I've made the bugfix release as promised. Get it from the usual place.
1046 There's a new major development release available from Ion web page.
1047 Not many changes are visible to the end user, though, but a lot was
1048 rewritten to be more flexible and simpler; see the ChangeLog for a
1049 full account. The older development release is also still available
1050 because this release can not be considered as stable after major
1051 changes to the code. The most visible changes are
1053 * Changes in binding configuration
1054 * A (mostly) working full screen mode toggle (see below)
1055 * The ability to switch workspaces while dragging tabs (experimental)
1056 * The ability to re-read draw.conf without restarting Ion
1060 Some notes on full screen mode toggle:
1062 * Ion doesn't detect programs trying to leave full screen mode--it
1063 should be possible to devise some method using window properties,
1065 * At least Opera unmaps the window also before changing the size to
1066 enter full screen mode, so Ion doesn't remember the last frame.
1067 * Mozilla sometimes has trouble entering full screen mode but when it
1068 succeeds, Ion remembers the last frame. However, if
1069 `clientwin_toggle_fullscreen` is used to leave the full screen mode
1070 started from Mozilla, Mozilla doesn't know that the mode has been
1073 Clearly there should be some method of communication between Ion and the
1074 programs for full screen toggles to fully work. At the moment I suggest
1075 using `clientwin_toggle_fullscreen` instead of the programs' native toggles
1076 unless the program has some special full screen mode and you really want