| Age | Commit message (Collapse) | Author |
|---|
|
New features:
- Commit ac8d152 ("menuconfig: Add support for custom color schemes
(aka styles)"), contributed by Mitja Horvat (pinkfluid), adds support
for configuring the style (colors and attributes) of menuconfig
interface elements by setting a new MENUCONFIG_STYLE environment
variable.
The format of MENUCONFIG_STYLE is documented in the menuconfig.py
module docstring. This format is now used internally as well to
define the predefined styles.
- Commit a092257 ("Allow macro expansion within symbol names") adds
support for expanding macros within symbol names. This could be used
e.g. to define template Kconfig files that get instantiated by
setting preprocessor variables before including them.
Other changes:
- Commit 6683213 ("Allow user values on 'option env' symbols") makes it
possible to give 'option env' symbols user values. There's no good
reason to disallow it (it's a leftover from an older way of
implementing them), and it simplifies the code a bit.
Note that 'option env' symbols are only supported for backwards
compatibility at this point. Environment variables can be expanded
directly.
|
|
Enforces that all windows are created from styles defined via the new
styling mechanism.
|
|
Mostly motivated by curses.COLORS, which depends on the terminal.
It's probably friendlier to run even if the style gets a bit messed up,
but it's also nice to point out problems.
Missing style templates and missing styles to the left of '=' are
pointed out now too, with the help of some flag hackery.
Prefix the warnings with 'menuconfig warning:' to make it easier to
distinguish them from warnings from Kconfiglib.
|
|
This is consistent with non-existent styles being ignored, and gives
some forwards/backwards compatibility.
I wonder if there should be a warning for both instead.
Piggyback some nits.
|
|
"color values are expected to be in the range 0 to COLORS-1", from
init_color(3ncurses).
|
|
Make the syntax 'fg:123' instead of 'fg:color123'. This is a bit easier
to type. Hex and octal constants are accepted as well.
|
|
- introduce the _parse_styles() function which can take a string
defintion and parse it into a color structure that can be used
by UI elements
- the _parse_styles() function modifies a global dictionary (_style)
which is subsequentially used to style the UI elements
- the _style_to_curses() function takes a string representation
of style colors and converts it to ncurses attributes.
For eaxmple: fg:brightred,bg:black,bold -> returns a color
pair representing brightred and black ORed with A_BOLD
- simplify _init_styles() as now the majority of the heavy lifting
is done by _parse_styles(). _init_styles() initializes the
default theme (either "defualt" or "monochrome")
- terminals without color capabilities are now handled by using
the "monochrome" theme. This theme doesn't use any colors
but rather just bold and standout attribtes.
- add support for the "underline" attribute
- add doctstring at the beginning describing the style syntax
- rename the _style() function to _style_attr() as it was conflicting
with the global _style dictionary
- introduce the StyleError exception; this one is thrown if an error
in the style syntax is encountered
|
|
Some grammar, some copy-paste errors ('string' should be 'name').
|
|
This is handled earlier in _tokenize() now.
|
|
The C implementation supports this (though it's undocumented, and unused
to far).
This can be used e.g. to dynamically instatiate symbols from template
files:
Kconfig.template:
config $(subsys)_LOG
bool "Enable logging for $(subsys)"
depends on $(subsys)_HAS_LOG
... other stuff dependent on $(subsys)
Elsewhere:
subsys = FOO
source "Kconfig.template"
subsys = BAR
source "Kconfig.template"
Pretty sure this can easily be abused, but it should be supported at
least.
|
|
Saves a few % of total parsing time, and probably some evaluation time
too. Microbenchmarking on Python 3 shows that 'is' is about 30% faster
than '==' for comparing integers on my machine.
This is safe even without assuming Python's small-integer optimization
(which caches objects for small integers, guaranteeing that small
integer literals always refer to the same object): We always refer to
symbolic constants using their symbolic names (_T_*, etc.), so they're
guaranteed to always refer to the same integer objects anyway.
This optimization is safe for client code too, unless you get some
unlikely situation where (1) there is no small-integer optimization, (2)
expressions get pickled and unpickled or the like, which would be unsafe
across Kconfiglib versions anyway, and (3) the client code tests
symbolic constants with 'is' instead of '==' (which isn't advertised as
safe, and a bad idea in general when pickling is involved). That's
probably too obscure to worry about.
|
|
Gets rid of a check, doesn't hurt. The check was added before
'option env' was changed to just add a 'default' under the hood.
Note: New Kconfig code doesn't need 'option env'. Environment variables
are now expanded directly, including in the C tools. 'option env' is
only maintained for backwards compatibility.
|
|
Invisible symbols are no longer shown in red outside of show-all mode.
|
|
Adds a small menuconfig tweak that could avoid some user confusion:
Commit 66a2bf5 ("menuconfig: Do not use red for invisible symbols
outside of show-all mode")
|
|
Invisible symbols can show up outside show-all mode if an invisible
symbol has visible symbols in an implicit (indented) submenu. Printing
them in red like in show-all mode is the perfectly consistent thing to
do, but might look confusing/broken.
Only print invisible symbols in red in show-all mode.
Also change the bracket style for all invisible bool/tristate symbols to
- - N_SYM
-M- M_SYM
-*- Y_SYM
That makes it clear that they can't be changed, even when they show up
outside show-all mode, and makes logical sense if that style is taken to
always mean that the symbol value can't be changed.
You lose a quick reference for whether a symbol is a bool or tristate,
but it's not a huge deal.
|
|
New features:
- Commit 84bd863 ("menuconfig: Add a more 'make menuconfig'-like color
scheme") adds an alternative menuconfig color scheme contributed by
contributed by Mitja Horvat (pinkfluid).
The new color scheme can be enabled by setting the environment
variable MENUCONFIG_THEME to 'aquatic'.
The default color scheme now also uses bold text for the selected
entry, like the 'aquatic' theme.
Fixes:
- Commit d1e2a65 ("menuconfig: Always show implicit submenus with
visible nodes") fixes a display issue involving prompt-only
conditions where visible child nodes could disappear outside of
show-all mode.
Other changes:
- Commit 0ed2ba0 ("Improve the running-without-Makefile-patch
documentation") updates the documentation on how to run Kconfiglib on
the Linux kernel without using the Makefile patch.
- Misc. internal menuconfig refactoring.
|
|
Show-all mode does not need to be enabled when jumping to an invisible
symbol with visible children from an implicit submenu, because the
invisible symbol will be shown anyway in that case.
Explicitly check whether the jumped-to node would be shown instead, and
enable show-all mode otherwise.
|
|
Currently, the symbol BAR below (which ends up indented in an implicit
submenu) is shown only if DEP is non-n (or if show-all mode is
enabled):
config FOO
bool "foo" if DEP
default y
config BAR
bool "bar" if FOO
This is bad, because it hides visible symbols from the interface. The
assumption was that an implicit submenu (which is only created if the
parent has a prompt) would never have visible items when the root item
is invisible, but prompt-specific conditions and select/imply can break
that assumption.
Fix it by always showing implicit submenus with visible nodes, along
with the parent node. If the parent node is invisible, show it in red,
like in show-all mode (which happens automatically). That's probably
better than having mysteriously indented nodes when the parent is
invisible.
|
|
- Include an updated list of environment variables, with sample values
- Give a method for listing all referenced environment variables, via
Kconfig.env_vars
- Remove the note that says that Kconfiglib will warn if an unset
environment variable is referenced. It's not true anymore with the
preprocessor, which silently expands unset variables to the empty
string.
Not setting essential environment variables causes obvious errors at
least.
|
|
|
|
Just checking whether the default value expression is a single symbol
correctly handles the string/int/hex cases too, because that case will
always hit for those. D'oh.
|
|
Used in two spots.
|
|
The global should be called _stdscr, and then the names don't clash.
Leftover.
|
|
Add info to module docstring.
|
|
Factor out the call to the iterator.
|
|
Add an alternative, less yellow, more 'make menuconfig'-like color
scheme, contributed by Mitja Horvat (pinkfluid). The new theme is
selected by setting the MENUCONFIG_THEME environment variable to
'aquatic'.
Make it so that custom themes can override just the styles they want to
change from the default theme.
Perhaps some non-hardcoded theming mechanism could be added later, but
keep it simple for now.
Piggyback some related changes:
- Make the selected item bold in the default theme, like the aquatic
theme does.
- Allow the style for the edit box in the jump-to dialog to be
configured separately from the style for edit boxes in pop-up
dialogs. That makes it possible to avoid some color clashes.
- Remove a redundant style setting for the jump-to edit box. The window
itself is already created with the correct style.
|
|
New features:
- Commit 20de53b ("Add a Kconfig.kconfig_filenames attribute") and
commit 932d0f7 ("Add a Kconfig.env_vars attribute that lists env.
variables") make lists of all Kconfig files and referenced
environment variables available.
This could be used e.g. to implement alternative incremental build
schemes. Note that the sync_deps() mechanism already indirectly
catches any relevant changes to files and environment variables
though (anything that changes the configuration output).
- Commit 17d7c1e ("menuconfig: Show all symbols at each menu location
for multi.def. choices") makes it so that if a named choice is
defined in multiple locations, each of which adds options, you can
select any of the options regardless of via menu entry you enter the
choice.
Previously, you could only select the options defined at that menu
entry's location.
Named choices with symbols in more than one location are completely
broken in the C tools, so there's no reference for the behavior
there. This seems less confusing at least.
Suggested by Mitja Horvat (pinkfluid).
Fixes:
- Commit 0b46af4 ("menuconfig: Don't return when selecting a choice
symbol with children") improves menuconfig behavior in an extremely
obscure case.
- Commit 1c2c4ae ("oldconfig: Fix help display command for choices")
makes '?' print the help for choices in oldconfig. The command was
accidentally set to '??' there.
- Commit ddfd5df ("Improve menu structure for promptless choices")
fixes a menuconfig display issue for promptless choices (which
can appear "legitimately" if a named choice is defined in multiple
locations to add extra options).
Cleanup:
- Commit 2320b7a ("oldconfig: Use Kconfig.node_iter() and clean up")
- Some manual tree walking in other places has also been rewritten to
use Kconfig.node_iter().
|
|
Promptless choices can appear "legitimately" if you define a named
choice in multiple locations to add on some symbols (which is broken in
the C tools though).
Prior to this fix, the promptless choice would get flattened, with the
choice symbols appearing in the same menu as the (invisible) choice.
This looks confusing.
Skip flattening promptless choices to fix it.
|
|
Some general cleanup:
- Handle the iteration with the new Kconfig.node_iter() helper.
This makes some function so short that they become pointless. Have
just main() + oldconfig(node) + small helper functions.
- Use _name_and_loc_str(sc) for choices too, so that all locations get
reported for named choices defined in multiple locations
- Rewrite the intro to be less wordy and remove the sample session
(leftover from when oldconfig.py was in examples/, and not that
exciting)
- Print "Updated configuration written to..." instead of
"Configuration saved to..."
- Various other nits
|
|
Kconfig.env_vars is a set() with the names of all environment variables
referenced in the configuration.
Can be used e.g. for custom incremental build implementations, though
sync_deps() already indirectly catches any relevant changes to
environment variables.
|
|
Kconfig.kconfig_filenames is an ordered list of all Kconfig files
included in the configuration, relative to $srctree.
Can be used e.g. for custom incremental build implementations, though
sync_deps() already indirectly catches any relevant changes to files.
Piggyback some ' -> " quote style consistency cleanups.
|
|
Wasn't available when the original version was written.
Also rename _check_undefined_syms() to _check_undef_syms(). That
shortening is used elsewhere.
|
|
- Test in the order AND, OR, NOT, like in other places. This matches
the frequency too.
- Use 'not isinstance(expr, tuple)', which is a bit faster than
isinstance(expr, (Symbol, Choice)).
|
|
'?' should display the help, like for symbols, not '??'. Oversight when
the command was changed.
|
|
If a symbol depends on a choice symbol that immediately precedes it,
it's put into an implicit submenu and not considered a choice symbol.
Don't immediately return to the parent menu when selecting such a
symbol. It makes the children hard to spot.
This is used (or was... not sure if it still is) in a single Kconfig
file in the kernel. It's completely undocumented.
|
|
For named choices defined in multiple locations, show the symbols from
all locations, regardless of via which menu node the choice is entered.
Named choices with symbols in more than one location are broken in the C
tools, so there's no reference for the behavior here. This behavior
probably makes more sense than just showing the choice symbols at the
current location at least, which was the old behavior in Kconfiglib's
menuconfig.
Suggested by Mitja Horvat (pinkfluid).
|
|
Fixes broken-looking menuconfig include path information for items in
the top-level Kconfig file, via commit 64600b3 ("menuconfig: Hide the
include path for top-level items").
|
|
Came out as "Included via " for items in the top-level Kconfig file,
which looks broken.
|
|
Adds commit 13884e9 ("Show include paths in menuconfig symbol
information"), which adds a MenuNode.include_path attribute to
Kconfiglib.
MenuNode.include_path holds the Kconfig include path of the menu node,
which shows how its Kconfig file got 'source'd, starting from the
top-level Kconfig file.
MenuNode.include_path is used to show include paths in the menuconfig
symbol information. It will also be used in genrest.py in Zephyr, which
generates an RST Kconfig reference.
|
|
Add a MenuNode.include_path attribute that holds a tuple of
(filename, linenr) tuples, giving the locations of the 'source'
statements via which the node's Kconfig file was included, starting from
the top-level Kconfig file.
Use MenuNode.include_path to give the include path for symbols and other
items in the help display in the menuconfig interface. This is useful
for figuring out how Kconfig files are organized, and when reorganizing
things.
|
|
Includes two small bug fixes:
- Commit 9b77511 ("Fix file descriptor leak for the top-level Kconfig
file")
- Commit d158fed ("Fix recursive 'source' error reporting")
|
|
This "architecture" has an unorthodox Kconfig setup that would require a
different testing setup.
|
|
The recursive 'source' detection was still fine, but the error reporting
had broken due a missed variable renaming.
The test suite didn't catch it, because a different type of KconfigError
was raised instead, due to a separate error in the test suite (need to
include tests/Krecursive{1,2}, since paths are relative to $srctree).
Fix the variable name and tighten up the tests to check that the
KconfigError message is the one we except. Tighten up the dependency
loop detection tests in the same way too.
|
|
Needs to be close()d. The other Kconfig files are close()d in
_leave_file().
Could drop the reference somehow too, but an explicit close() is best
for PyPy, which doesn't do reference counting.
|
|
New features:
- Commit 7dae988 ("Add a generic node iterator") adds a
Kconfig.node_iter() helper generator for iterating through all menu
nodes in the menu tree, in definition order. Optionally, symbols
defined in multiple locations are only generated once. This generator
is now used internally by write_config() as well.
Suggested by Mitja Horvat (pinkfluid).
- Commit d2c1430 ("Introduce Kconfig.unique_defined_syms and
Kconfig.unique_choices") adds versions of
Kconfig.defined_syms/choices that only include symbols and choices
defined in multiple locations once, while still preserving definition
order.
This is almost always what you want when iterating through (just)
symbols and choices. It removes set() hackery (where it's easy to
accidentally mess up the order) and the like.
Fixes:
- Commit 4ebc2aa ("Flag extra tokens after 'if'/'depends on'/'visible
if' expressions") tightens up syntax checking so that extra trailing
tokens after 'if', 'depends on', and 'visible if' expressions raise a
syntax error, instead of being silently ignored. Oversight.
|
|
Extra trailing tokens after 'if <expr>', 'depends on <expr>', and
'visible if <expr>' now trigger syntax errors instead of being ignored.
Oversight.
This indirectly makes Kconfig.eval_expr() detect extra trailing tokens
as well.
|
|
Suggested by Mitja Horvat (pinkfluid) in
https://github.com/ulfalizer/Kconfiglib/pull/50.
Kconfig.node_iter() iterates through all menu nodes in the menu tree in
Kconfig order. This saves scripts the trouble of implementing their own
tree walking code.
Have node_iter() take a 'unique_syms' flag that can be enabled to only
include symbols defined in multiple locations once. This is often what
you want when generating output (and is used by write_config()). Order
is still preserved.
Piggyback a fix to a syntax error test comment. Parsing has been
tightened up now.
|
|
These are the same as Kconfig.defined_syms and Kconfig.choices, except
duplicates are removed. Kconfig order is still preserved.
This is almost always what you want when iterating through symbols and
choices, as it potentially saves work, avoids generating duplicates when
writing output, and still preserves Kconfig order for readability.
The old attributes will be kept for backwards compatibility (maybe
there's some rare cases where they could be useful too). They're created
internally anyway.
|
|
These are never used at the same time, and Symbol._visited is a good
name for both.
Gets rid of an internal attribute.
|
|
Instead of precalculating a set() to get unique symbols, precalculate a
list with any duplicates from multiple definition locations removed, and
preserve the order of the symbols within it.
This makes it possible to get rid of the Symbol._written shenanigans in
functions that only need to iterate through unique symbols in sorted
order, which is all of them except write_config() (because it needs to
walk the entire menu tree).
|
