The change in #2270 accidentally broke how we access module types.
module<Impl>::TYPE always points to the module superclass and it thus
accesses its empty TYPE field.
This mainly broke legacy action handling.
Ref #2270
Since the forked processes are still our children, we need to wait on
them, otherwise they become zombie processes.
We now fork twice, let the first fork immediately return and wait on it.
This reparents the second fork, which runs the actual code, to the init
process which then collects it.
Ref #770
For some reason when passing some non-const strings to convert, the
convert(T&& arg) method was used instead of the one specialized for
strings.
This caused an error in clang because you can't pass objects with
non-trivial types to varargs functions.
The best solution I found was to just add a specialized function for
non-const strings.
If two WM events arrive withing 25ms of one-another, the second one does
not trigger a bar update.
The module state is still correct, it is just not reflected in the bar.
This somehow caused updates being swallowed in fluxbox, but only after
PR #882 was merged, even though that 25ms restriction existed long
before that.
Fixes#2272
Since 3.5.0, we use m_interval for a modulo operation, this crashes the
bar if the interval is 0. A non-positive interval shouldn't be allowed
anyway, so we now throw an exception in that case.
Fixes#2273
* [Temperature, Ramp] fix wrong icon for temperatures near base and warn temps
* [Temperature, Ramp] fix wrong icon for temperatures near base and warn temps
* Fix minor error
* Added WARN state for cpu module
* Implement WARN state for CPU, Memory modules, working on fs module
* Implement WARN state for fs module
* Simplify WARN state implementation for cpu and memory
* explicitly check percentage in get_by_percentage_with_borders
* Fixed silly error
* implement warn state on battery module, standardize the implementation on other modules
* minor fixes
* fix annoying error
* use more intuitive param name
* Fix percentage with borders bug
* Make requested changes
Hide the effect of warn states when unused
* Backward Compat: use no format instead of fallback label
* Reformat
* Refactor
* Reformat
* Reformat: convert tabs to spaces
* Reformat
It was not quite clear how try_apply_alpha should behave if the given
color was not ALPHA_ONLY. The implementation just returned 'this'.
However, the build relied on it returning the given color.
This broke all bg and fg settings in the entire bar.
To clear this up, we switch this around take the alpha channel of 'this'
and also return 'this' if it isn't ALPHA_ONLY.
Fixes#2255
New config option `speed-unit = B/s` will be used to suffix the upload and download speeds.
* mod::network: udspeed-unit to set network speed unit suffix
* Changed udspeed-unit to speed-unit
Shell commands triggered from action tags used to block polybar until
they finished.
Since we are not actually interested in the output of the commands, it
makes sense to run them completely detached from polybar and have
polybar not block when executing these commands.
Now the spawned child processes no longer get killed when polybar
exits. This is fine because polybar is not responsible for these
processes since they were explicitly started by the user through click
commands.
Ref: #770
Ref: #1680
The intent is for every color to be stored in a rgba instance
The rgba class now stores the color in a 32 bit integer to save space
This also removes the unused class rgb and moves everything else into a
cpp file.
Many functions also had weird template parameters. For example
alpha_channel<unsigned short int> would give a 2 byte number with the
alpha channel byte in both bytes.
color_util::hex would return a hex string with alpha channel if unsigned
short int was given and without if unsigned char was given. Even more
curiously those parameters were passed to *_channel and the result
nevertheless truncated to 8bits.
Only modules can now be action handlers.
This also slightly simplifies the controller because we don't need to
keep track of input handlers, we can just use the module list.
All modules now expose their actions as public static constants
Issues: The menu module no longer closes when an item is clicked (before
it would intercept any executed command and look if it matches one of
its exec commands)
This allows us to identify module by their type and it is also better to
store the module type as part of the module instead of having it
hardcoded in factory.hpp
Action strings now have the form '#MODULE#ACTION'
For example to trigger the action 'toggle' in the 'module/date' module
one would now use '%{A1:#date#toggle:}'
With this action strings can now be uniquely assigned to one module.
Fixes#1172
ramp-0 is used for everything <= base-temperature and ramp-N is used for everything >= warn-temperature
* [Temperature, Ramp] fix wrong icon for temperatures near base and warn temps
* [Temperature, Ramp] fix wrong icon for temperatures near base and warn temps
* Fix minor error
* explicitly check percentage in get_by_percentage_with_borders
* Fixed silly error
Any timer_module based module would sleep for the set interval and then
continue running. Depending on the start time of polybar this
sleep pattern might not be aligned, which causes such modules to always
update in a shifted manner.
Consider the date module as an example. If the update interval is set to
60 seconds and polybar was started at 13:37:37, polybar would update the
clock at 13:38:37, 13:39:37 and so on.
To make matters worse, if a module would perform lengthy checks this
interval might drift over time, causing even more inconsistent updating.
This patch extends the base module with a sleep_until method that calls
the corresponding function on the sleephandler. Additionally the
timer_module is extended to compute the remaining time until the next
interval passes and sleep accordingly.
Closes#2064
Co-developed-by: Dominik Töllner <dominik.toellner@stud.uni-hannover.de>
Co-authored-by: Malte Bargholz <malte@screenri.de>
If an input is enqueued as a response to an input, the new input will be
swallowed because it will likely be enqueued less than 30ms after the
original event.
This is not something that is an issue right now but it is required to
finish #1907 where, in order to close the menu after a click, the menu
module gets an exec action that closes the menu and adds a command to
the event queue.
The setting also isn't too useful since it will just break polybar input
handling if inputs arrive too fast instead of (possibly) slowing down
the bar.
Before the module would just try to evenly distribute desktops
(workspaces) among the viewports.
But since `_NET_DESKTOP_VIEWPORT` actually maps desktops to viewports,
we can use that information to assign workspaces to the right viewport.
Fixes#1849Fixes#1764
A warning implies something went wrong and (possibly) the user should do
something about it. However, warnings are not always used this way.
For example:
* When a fallback value for a `${..}` reference is used, this shouldn't
produce a warning (or notice) since using fallbacks is not something
bad.
* pulse telling you that it uses the default sink because no sink was
specified also does not warrant a warning (even notice may be too
high).
* Whenever polybar shuts down it produces a "Termination signal
received..." warning. Since there isn't a more proper way to shut down
polybar, it should not produce a warning. Same argument for a
`screenchange-reload`
Before it would only reload if the size changed and even that was
reliable since the method relied on the order of the monitor list.
Now if the monitor list differs in any way (pos, dimension, primary,
output, name) a reload is issued
Because of how monitors are removed inside the loop and depending on the
monitor order a cloned monitor may be assigned a width of 0 but is never
actually removed resulting in polybar saying the bar is out of bounds
Fixes#1794
Premature optimization that tried to cache monitors but the cache did
not take into account the parameter values.
The call `get_monitors(..., ..., false, true);` would get all connected
and unconncected monitors a subsequent call
`get_monitors(..., ..., true, false);` would get back the same list of
monitors even though it requested only connected monitors.
Additionally `get_monitors` is never called periodically so the
optimization really didn't help much.
The github module only authenticate by query string, and this method is deprecated:
https://developer.github.com/changes/2019-11-05-deprecated-passwords-and-authorizations-api/#authenticating-using-query-parameters
There is no reason to remove it before the method stop working, so I've made possible to the user choose which authentication method he will use:
* The parameter token remain unchanged.
* If the parameter user is passed then the module will use the not deprecated method, passing user and token on the body of the requisition. Otherwise the module will use the deprecated method.
Co-authored-by: Lucas <araujo.lucasvale@gmail.com>
Fixes#2002
Adds `format-offline` and `label-offline`
* feat(github): offline label & fixes
* Clear label if there are no notifications and empty-notifications = false
* clang-format
Co-authored-by: Patrick Ziegler <p.ziegler96@gmail.com>
This is dependent on a PR to xpp that does the same. Newer compilers
(GCC10 in particular) are stricter about which headers provide the
std exception types.
the `node` and `node_repeat` were never called with the optional
parameter add_space. And its default value is false, so the
corresponding code was never used.
I think in general it is better to just have the user of those functions
call `builder::space` when they need to instead of adding more
complexity to the `node*` functions.
* Add label minlen and alignment.
Fix build
* Update src/drawtypes/label.cpp
Co-Authored-By: infokiller <infokiller@users.noreply.github.com>
* Use existing alignment type.
* Remove redundant max_len handling in label::get.
* Fix shadowing.
* Add label alignment tests.
* Handle minlen/maxlen and alignment in same function.
Also add a test for a test case brought up in the PR discussion.
* Format files with clang-format
* Move builder::get_label_text tests into label tests
builder::get_label_text doesn't really do anything anymore
* builder: remove get_label_text
* label: Clean up label::get()
* Fix comment style.
* Set default label alignment to left.
* Update src/drawtypes/label.cpp
Co-Authored-By: Patrick Ziegler <p.ziegler96@gmail.com>
* Update include/drawtypes/label.hpp
Co-Authored-By: Patrick Ziegler <p.ziegler96@gmail.com>
None of the payload fields seem to be used. They were the only place
where EVENT_SIZE was used and why it had to be a macro (no variable
length arrays).
Since APP_VERSION is different for every commit and almost all file
include settings.hpp, the whole project has to be rebuilt for every
commit. With this, hopefully, this can be greatly reduced and only
changed files need to be rebuilt. This will also help ccache
Some people use text modules instead of the `separator` key in the bar
section to better configure the separator (colors, fonts).
Since we disallowed the same module being used multiple times in #1534,
this will now print an error message.
This should help with this a bit.
Ref #1913
Since all of polybar is built at once, there is no chance that this is
ever linked to an object that was compiled with another `-std=`
Ref: https://stackoverflow.com/a/46857525/5363071
Some WMs like i3 discard position information when unmapping the bar and
because of that the bar would be at the wrong position after being
remapped.
Fixes#1484
Ref: https://github.com/i3/i3/pull/3834
* refactor: Use flat module list if possible
Before if you wanted to iterate over all loaded modules you had to first
iterate over all blocks and then over their modules even if you didn't
care about alignment.
* refactor: setup modules in separate function
* controller: Print error for duplicate modules
You can't use the same name twice inside the module lists
E.g.
modules-left = a b c
modules-center = a
modules-right = b
would print an error.
We only print an error for now because we don't want to break existing
configs. But in the future this should be properly enforced.
A workspace is occupied if it is not active and there is at least one window managed by the WM (`_NET_CLIENT_LIST`) that has set `_NET_WM_DESKTOP` to that workspace.
The behavior when `_NET_WM_DESKTOP` is not set is not yet clear but this is unlikely to happen since most WMs will position windows on some desktop.
Closes#874Fixes#1444Fixes#1033
* Set Desktop OCCUPIED if a window moves there
This covers more of an edge-case. I did this first by accident, it might
vanish later on.
* Replace tracking change of WS with currently used WS
* Untrack occupied workspaces
* Track windows and their desktops in pairs
* Match type of occupied_desktops with current_desktop
Because the index needs to be matched later on, type mismatches would be non-ideal.
* Recreate the occupied desktops everytime and remove duplicates
* Readd support for moving windows to other desktops
* Use less characters to empty the vector
* Rename variable storing the desktops
* Recount windows on every occasion
This alone simplifies the management and the lookup for occupation of a
workspace
* Keep track of number of windows in every workspace
* Add debugging output that shall be removed before merging
* Remove obsolete TODO
* m_client_list should always be diff'd, since the desktop may change
Therefore we update the desktop-count tally every time the client_list
changes. It may just be a desktop-change without a change of
clients.size()...
* Add more logging-spam to understand window/desktop lifecycle
* Lock event-handler to serialize handling of events
* Fix occupied workspace counting and change to bool array
Also, performance improvements when diffing new and old client lists
* Fix crash when all clients are removed
* Conform to linter and styleguide
* Shorten conditional as it is standard enough
Since this only guards against 0-divisions, it can be shortened
without risking too much confusion down the road.
* Guard against multiple threads accessing and modifying data
Fixes#1444
Modification of internal data happens through the handle-method, while
the build-method tries to access the data structures for display. Since
some modifications clear e.g. the m_viewports, references may become
invalid between looping over them an accessing them.
The mutex should guard against this simultanuous access.
* Do not 'adopt_lock', because calls come from very different threads
To my understanding, adopt_lock has some dependency on the mutex-ownership. Since
the lock is once called from the inside (in handle) and once from the outside (in
build), there might be a problem. After brief testing, the segfaults happened fewer
times.
See #1444
* Also listen to _NET_WM_DESKTOP
In order to move a window from one desktop to another, it is sufficient
to set the desktop-property of that window. xmonad fires a lot of events
in the case of moving a window, herbstluftwm only updates the
_NET_WM_DESKTOP-atom of the window.
This change reloads the clientlist in order to correctly set the
desktop state "occupied".
* Describe need and use of mutex
It might be possible to relieve the guard in xworkspaces_module::handle,
but I am unsure about this. Since xmonad emits a lot of events on almost
every minor change, I would let the guard keep its post, avoiding
race-conditions in event-handling.
* Give temporary variables better names
* Clarify purpose of loop
About 80% of this comment are taken from
https://github.com/jaagr/polybar/pull/882#discussion_r255317363
* Remove merge-remainder
* Use a simpler method to list occupied desktops.
Co-authored-by: Jérôme Boulmier <jerome.boulmier@outlook.fr>
* Document m_clients field
The %{PR} tag is introduced for this. It resets all colors as well as
the activation of the underline and overline and font.
This has become necessary because we don't track what raw tags a user
injects into the formatting string and otherwise their raw tags could
bleed through.
This doesn't touch action tags because even before raw action tags
weren't being tracked. Action tags also have the requirement that they
have to be used in pairs, so closing them prematurely could break things
(for example with click actions for the entire bar)
Using a bit vector to track the active attributes does not really give a
significant speed increase, especially for only two attributes
Checking if a syntaxtag or an attribute exists in the map just adds
unnecessary code
Was around since the first commit but no documentation why. It was only
used in the text module and doesn't appear in any public documentation,
so this doesn't break anything.
This is the next step to merge #1237 in stages.
Currently there are barely any restrictions on how the config can be
written. This causes things like config files with DOS line endings to
not be parsed properly (#1366) because polybar splits by `\n` and when
parsing section headers, it can't deal with the `\r` at the end of the
line and thus doesn't recognize any section headers.
With this PR we introduce some rules as to what characters are allowed
in section names and keys.
Note: When talking about spaces I refer to any character for which
`isspace()` returns `true`.
The rules are as follows:
* A section name or a key name cannot contain any spaces as well as any
of there characters:`"'=;#[](){}:.$\%`
* Spaces at the beginning and end of lines are always ignored when
parsing
* Comment lines start with `;` or `#` and last for the whole line. The
whole line will be ignored by the parser. You cannot start a comment at
the end of a line.
* Section headers have the following form `[HEADER_NAME]`
* Key-value lines look like this:
`KEY_NAME{SPACES}={SPACES}VALUE_STRING` where `{SPACES}` represents any
number of spaces. `VALUE_STRING` can contain any characters. If it is
*surrounded* with double quotes (`"`), those quotes will be removed,
this can be used to add spaces to the beginning or end of the value
* Empty lines are lines with only spaces in them
* If the line has any other form, it is a syntax error
This will introduce the following breaking changes because of how
underdefined the config syntax was before:
* `key = ""` will get treated as an empty string instead of the literal
* string `""`
* Any section or key name with forbidden characters will now be syntax
errors.
* Certain strings will be forbidden as section names: `self`, `root`,
* `BAR`. Because they have a special meaning inside references and so a
* section `[root]` can never be referenced.
This replaces the current parser implementation with a new more robust
one that will later be expanded to also check for dependency cycles and
allow for values that contain references mixed with other strings.
This PR also now expands the config paths given over the command line so
that `--config=~/.config/polybar/config` resolves properly.
Closes#1032Closes#1694
* config_parser: Add skeleton with tests
First step in the config_parser develoment. Only tests functions that
are easily testable without many outside dependencies. Integration tests
will follow.
* config_parser: Implement parse_header
* config_parser: Implement get_line_type
* feat(string): Add trim functions with predicate
Not only trimming based on single character matching but based on a
freely specifiable predicate. Will be used to trim all spaces (based on
isspace)
* config_parser: Implement parse_key
* config_parser: Implement parse_line for valid lines
* config_parser: Throw exception on invalid lines
* config_parser: Remove line_no and file_index from parse_line
Cleaner to let the caller catch and fill in the line number and file
path
* string: Clear up misleading description of trim
Before, trim would remove all characters that *didn't* match the
predicate and thus the predicate isspace wouldn't work correctly. But
because we used the inverse (isnospace_pred) it all worked out, but if
the function was used with any other function, it wouldn't have given
the desired output
* config_parser: Implement parse_file
* config_parser: Switch operation to config_parser
This changes the way the config is invoked. Now main.cpp creates a
config_parser object which then returns the singleton config object from
the parse method. Subsequent calls to config::make will return the
already created config object as before
The config_parser does not yet have all the functionality of the old
parser: `inherit` directives are not yet resolved. Other than that all
the old functionality is implemented (creating sectionmap and applying
include-file)
Any sort of dependency detection (except for include-file) are still
missing
* config: Move xrm initialization to constructor
config_parser handles the detection of xrdb references and passes that
info to the config object.
This finally allows us to delete the config::parse_file function because
everything in it has been implemented (except for xrdb detection and
file error handling)
* refactor(config_parser): Cleanup
* config_parser: Set config data after initialization
Looks much cleaner this way
* config_parser: Expand include-file paths
* config_parser: Init xrm if the config uses %{xrdb references
* config_parser: Use same type of maps as in old impl
Polybar has some weird, not yet fixed, inheriting behaviour and it
changes depending on the order in which the config stores its data.
Using the same type of maps ensures that the behaviour stays the same.
* refactor(config_parser): Clearer invalid name error message
* config_parser: Don't allow reserved section names
Sections with the names 'self', 'BAR', 'root' could never be referenced
because those strings have a special meaning inside references
* config_parser: Handle inherit directives
This uses the old copy_inherited function, so this still suffers from
crashes if there are cyclic dependencies.
This also fixes the behaviour where any key that starts with 'inherit'
would be treated as an inherit directive
* config_parser: Clearer dependency cycle error message
* refactor(config_parser): Handle file errors when parsing
This removes the need to check if the file exists separately
* fix(config): expand config file path
Now paths using ~ and environment variables can be used as the config
path
* fix(config): Properly recognize xrdb references
* config_parser: Make messages more informative
* doc(config): Improve commenting
Comments now describe what the config_parser actually does instead of
what it will do.
We also now follow the rule that single line comments inside functions
should use `//` comments
* refactor: Move else on same line as curly braces
* fix(config_parser): Don't duplicate paths in `files`
* refactor(config_parser): Use else if for clarity
* fix(config): Undefined behavior in syntax_error
Before the custom what() method produced undefined behavior because the
returned string became invalid once the function returned.
* refactor(config): descriptive name for useless lines
is_valid could easily be confused as meaning syntactically invalid
without it being clarified in a comment
* refactor(config): Use separate strings instead of key_value
Takes just as much space and is much better to read
* fix(config_parser): TestCase -> TestSuite and fix macro call
Ref: #1644
* config_parser: use const string& in method args
* config_parser: Improve comments
* config_parser: Incorporate review comments
Originally the size function returned the scaled `size` property for
scalable fonts and the non-scaled `pixelsize` property for non-scalable
fonts. This caused lots of issues when that property was 0 (empty bars,
characters not drawn without warning, see references at the bottom).
This behavior was mostly observed on debian where `size` is set to 0 if
`pixelsize` is set.
We now try to use both properties for both types, but prefering `size`
for scalable fonts and `pixelsize` for non-scalable ones.
This behavior doesn't break existing correct behavior but now never
returns 0. It will always try to fall back to the other property or to
some fallback value if both properties are 0.
I originally thought this could also make font patterns more expressive
by being able to specify the size of scalable fonts directly in pixels
like so:
Unifont:size=0:pixelsize=20
or to scale non-scalable fonts by forcing polybar to fall back to the
`size` property (which is always scaled):
Wuncon Siji:pixelsize=0:size=20
But how these two patterns are matched by `fc-match` depends both on the
font and on the distro/fontconfig setup.
Ref #706
Ref #1450
Ref #1257
