writer: check return code as gboolean

This makes the check on return code stricter, by using the actual
gboolean value.

codgen: trim extra _async in future variants

Fixes #1208

docs: allow to set doc_struct_name for global functions

Like async_finish, ignored ones or special ones

Useful for Widget's TickCallbackId for example

Some of those are moved to a different struct in some cases
like gdk::Key and would be nice to h...

docs: allow documenting global functions in traits

sys: stop importing time_t,

Nothing seems to be using it

Fixes #1276

docs: work around glib::Variant case

as it's a derive macro and a struct, specify the generated docs as a struct

codegen/function: assert sane GError behavior

This introduces an assertion in generated code for FFI logic
making use of GError.
When the wrap...

Add `#[must_use]` to the autogenerated builder types

They don't do anything on their own and must be built to have any
effect.

Fixes https://github....

gir: fix per emit_by_name change

ObjectExt::emit_by_name now return the correct type directly
instead of Option<Value>

Mark gi-docgen as optional

This should probably be rewritten someday to make use of some other config options

Only the minimum that's known to not regress the general functionnality

This would be used for setting whether the bindings docs would use the gi-docgen parser
or the d...

Only use remove_dir_all if there is a folder to remove

Move auto folder removal into gir directly

Update to system-deps 6

Remove "auto" folder when regenarating non-sys crate

codegen: stop renaming crates inside sys generated code

As per other 2018 edition changes, the renames should be done on Cargo.toml instead
Fixes #998

codgen: don't generate empty impl blocks

Fixes #976

codgen: don't generate a functions mod if it's empty

Noticed that in https://gtk-rs.org/gtk4-rs/git/docs/gsk4/functions/index.html

codgen: generate builders inside a builders module

This is a workaround to make sure the builders are usable from outside
without making the module...

codgen: don't re-export builder types on crate root

they are available nowadays through Type::builder method

Correctly generate opaque types

codgen: don't assume child props are gtk only

libraries such as libhandy have those

Fixes #1096

codgen: fix a logic issue in version_condition_no_doc

If no version is set for another crate name, we should always add a guard
This caused some guar...

revert edition change

per crate min version

we should fallback to comparing against min_cfg_version
if there's no version to compare with

similar to the former commit

When generating a builder pattern; the properties could be
from a super-type interface/class tha...

Instead of comparing the main crate's min_version with the required version of super-types
use t...

codgen: simplify properties getters/setters

Use ObjectExt directly instead of going through ffi

codegen: generate NONE constants as associated constants

So that we can use `Pad::NONE` instead of crate global `NONE_PAD`.

See https://github.com/gtk-r...

Revert #1248

the deprecated guard has no effect on impl

Fix support for nullable paths / OsStrings

And produce "to_glib_extra" code snippet in a single place consistently.

codgen: generate deprecated cfg attribute for Default trait

If the constructor method is deprecated, the builder pattern should have a guard
as well

Switch to 2021 edition

No changes needed.

Add `use glib::translate::*` for generating the equal special function

book: Fix sys crate import to use the 'ffi' convention.

Gir now generates code expecting the crate to be imported as 'ffi'.

book: Add a section on fixing missing macros

Co-authored-by: pentamassiv <[email protected]>
Co-authored-by: Sebastian Dröge <slomo@coaxi...

book: Remove suggestions to add `extern crate` imports

Gir generates edition = "2018" crates now, which do not require them:
https://doc.rust-lang.org/...

book: Fix references to the `girs_dir` config option

analysis: mark to_str functions as renamed

Instead of overriding their original names; this makes the docs work without workarounds
by mark...

Implement support for inline allocated Boxed types

Previously only AutoBoxed was considered boxed, which caused wrong code
generation for special f...