Anything at the module/file level:
* _... is protected
* rest is public
Anything at the class level:
* __...__ is public (magic method)
* __... is private
* _... is protected
* rest is public
Anything at the function/method level:
* everything is private
Closes https://github.com/fishman/ctags/issues/216.
Also don't default to CR if OS is neither Windows nor Unix (including OS X).
There's no other GTK backend right now so it doesn't matter much but
still if something else appears, it will most probably not have CR line
endings.
The standard says:
If the first operand is -n, or if any of the operands contain a
backslash ( '\' ) character, the results are implementation-defined.
On OS X it simply prints the "-n" string and everything which follows,
including the newline.
Use printf instead.
When a string is not used as an object property the parser doesn't
need to know its value. Not collecting it into memory lowers memory
consumption and avoids high memory consumption with huge string values.
As discussed in #410, the previous SF redirect is better replaced by
a redirect on www.geany.org which we can control.
This way we can also quickly change the target URL if it changes at
Paypal.
As discussed in SF bug #125, it might be dangerous to store backup
copies in a publicly accessable directory like /tmp with default
permissions, especially on multi-user systems.
So set the file permissions on non-Windows systems to 0600 by default.
Also improve the documentation of the save Actions plugin to reflect this
change.
Since document_compare_by_tab_order_reverse does the exact reverse of
document_compare_by_tab_order the code need not to be duplicated. Instead
document_compare_by_tab_order can be called and the return value be reversed.
Fonts are rendered much smaller on OS X so use bigger sizes
by default. Also, by default they use the Light style which is
extremely thin - use Medium instead. Finally, use better monospace
font than the default Courier (Menlo is derived from Dejavu Sans
Mono and is used by XCode and other editors).
In order to execute the runner script using the terminal application,
the script has to be executable, otherwise it's not executed.
In addition, when the terminal application starts, it sets $HOME
as the working directory. For this reason the working directory
has to be set in the runner script (to avoid additional ifdefs,
this part is shared with other unix systems where the additional
"cd" shouldn't matter).
On OS X the Command key is used for common keybindings instead
of Ctrl. Introduce a new macro, GEANY_PRIMARY_MOD_MASK that
represents the Command key on OS X and Ctrl on other platforms.
For some events, such as mouse key press, GDK_MOD2_MASK is returned
for the Command key by GTK instead of GDK_META_MASK (which is returned
when Command is pressed together with some other key). To hide this
behavior from users, introduce keybindings_get_modifiers() which can be
used instead of gtk_accelerator_get_default_mod_mask() and which
inserts GDK_META_MASK when GDK_MOD2_MASK is found in the mask
on OS X.
The Option initialization didn't match the struct - there was
an extra initialization of -e parameter and missing initialization
of --etags-include a bit later. This means the initializations of
the fields were shifted by one between these two.
It may happen that even though we get GIcon, the actual icon
dosen't exist (typically on Windows and OS X). Check if we
can find the actual icon.
In addition, use "icon name" instead of "stock id" - the latter
doesn't work on OS X / Windows for some reason.
The GIOCondition is always set to G_IO_IN even when input end
is reached (and no other flags are set) so the read_fif_io()
function never returns FALSE which causes an infinite loop.
It is necessary to check also the return value of
g_io_channel_read_line() and return FALSE on EOF or error.
This feature looks like a poorly implemented subset of "Complete word",
but lacks some important features like prefix filtering. It is also
misnamed as it shows not only macros, but also variables and enums.
It also only shows `editor_prefs.autocompletion_max_entries`, but does
so from *each* file, not as a whole.
So drop it altogether, as this feature doesn't really look useful and
the current implementation seem to suffer of too many shortcomings for
it to realistically be actually used.
The ant_glob() function doesn't return a list of strings but rather
a list of waflib.Node.Nod3 objects. These print as paths so build
works but apparently confuse waf which thinks the files have
changed every time the waf command is performed.
Relinking scintilla and geany on every waf invocation is especially
problemmatic during "sudo waf install" where scintilla and geany
binaries get owned by the root user. This isn't a big problem on Linux
but on OS X this prevents subsequent waf calls to update the
binaries and the build fails.
To fix the issue, just convert the waflib.Node.Nod3 objects
to relative (string) paths.
