| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
We change the style of documentation pages for constants, so that their
values are displayed in a <dt>/<dd> pair. We add a new method to the
formatter family, format_value() which should print the representation of
a constant's value in the target language; much like repr() in Python.
|
|
|
|
|
| |
This function was failing on g_iconv() which has a blank name in the GIR
file (its name instead coming from the moved-to attribute.)
|
|
|
|
|
|
|
|
|
|
| |
This adds a def in _doc.tmpl that expands to "deprecated" if the node is
deprecated, and to nothing otherwise. We use it to give the deprecated CSS
class to particular elements. The presence of the class on a symbol's main
entry will inform DevDocs that the symbol is deprecated, though it won't
be formatted in a particular way. The presence of the class on an index
entry will cause the index entry to be struck out, so that it's visually
obvious in the index when a symbol is deprecated.
|
|
|
|
|
|
|
| |
For GIR nodes with version annotations, such as "Since:", "Deprecated:",
and "Stability:", we generate stability notes at the top of each node's
documentation. These notes are given the CSS class "versioning-note" so
that we can format them nicely in DevDocs.
|
|
|
|
|
|
|
|
| |
Skip private fields such as "priv" and "parent_instance". Unfortunately
not all private structure and parent instance fields are marked private,
and not all of them are named "priv" or "parent_instance".
This also includes fields that are contained within anonymous unions.
|
|
|
|
|
|
|
|
|
|
|
| |
Code blocks, i.e. |[ ]|, were converted into <code> elements, but <pre>
would be more appropriate.
In addition, add four spaces to the start of each line in a <pre> element
so that Markdown will recognize it as code and not mangle it. As explained
in a FIXME comment, a better solution would be to reimplement
DocstringScanner using Markdown extensions so that such confusion becomes
impossible.
|
|
|
|
|
|
| |
This summary, heavily inspired by the one from pgi-docs, is useful when
browsing the page rather than searching for a specific term. It's also
similar to the summary at the top of gtk-doc pages.
|
|
|
|
|
|
|
| |
In order to generate HTML output that DevDocs can easily scrape and
display, we add a new output format to g-ir-doc-tool (--format=devdocs).
It works similarly to the Mallard output format, but generates very simple
HTML instead. We add a new set of Mako templates to generate this output.
|
|
|
|
|
|
|
|
|
| |
This also removes some "parsing" code for paragraphs, which
is best handled by the actual scanner.
+ Also fixes some pep8 issues.
https://bugzilla.gnome.org/show_bug.cgi?id=745298
|
| |
|
|
|
|
|
|
|
|
| |
On OS X, the output of sed as reported back to Python has a newline at
the end. It seems like a good idea to strip whitespace from the symbol
and identifier filter commands anyhow, so strip() the result.
https://bugzilla.gnome.org/show_bug.cgi?id=752468
|
|
|
|
|
|
|
|
|
| |
For generating other output formats such as DevDocs-ready HTML, we add an
output format option (-f). The default output format is "mallard" which
leaves the existing behaviour unchanged.
We can fold the existing --write-sections-file option into the new output
format option, as a new "sections" format.
|
| |
|
|
|
|
|
| |
Added in 9535fc48. After the switch it calls _g_ir_node_build_typelib() again
handling the other types.
|
|\
| |
| |
| |
| | |
autotools: add various compiler warning flags and enable -Werror for CI
See merge request GNOME/gobject-introspection!56
|
| |
| |
| |
| | |
This lets us enable -Wdeprecated-declarations
|
| |
| |
| |
| |
| | |
The ffi decls were added in cbdd9ee09e367e4dd to work around broken
ffi headers. Let's assume this is fixed now.
|
| |
| |
| |
| |
| |
| |
| | |
In case the surrounding code handles missing cases break, otherwise add
a g_assert_not_reached().
The generated parser code triggers this as well, so disable it there only.
|
| |
| |
| |
| | |
hashv is unsigned, no need to check if >= 0
|
| |
| |
| |
| | |
and fix a missplaced break
|
| |
| |
| |
| |
| | |
Except for the generated lexer code which triggers it with the
version of flex used in our CI.
|
| | |
|
| |
| |
| |
| | |
The error arg was used but the result never checked.
|
| |
| |
| |
| |
| |
| |
| |
| | |
Where it's easy add dummy args to match the cast; where the target is a subset just
prevent the warning with a cast to void*.
Provide a real copy function for the boxed type code in regress_foo.
This code is never executed afaics, but why not.
|
| |
| |
| |
| | |
And use G_GNUC_PRINTF for the suggested function.
|
| | |
|
| | |
|
| |
| |
| |
| |
| | |
Except for the Python module because nothing in the CPython API is marked const
and we'd have to cast everywhere.
|
|/
|
|
|
|
|
|
|
| |
* This adds a new dependency on autoconf-archive.
* Pass WARN_CFLAGS/WARN_LDFLAGS/WARN_SCANNERFLAGS to everything we control the source of
* Disables all warning flags which make the build error out for now, we'll re-enable them
in followup commits.
* AX_COMPILER_FLAGS is used with the release flag always on so we don't get
-Werror by default.
|
|
|
|
|
|
|
| |
I made this edit to Simon's commit to resolve a merge conflict, but I
messed it up. Revert to what it was before.
See #113.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Functional changes:
* Consistently check that return code from pkg-config is zero.
* Use shell word splitting rules to process pkg-config output to match
behaviour obtained by running `cc program.cc $(pkg-config --cflags ...)`.
Fixes issue #171 .
* Use user preferred encoding to process output from pkg-config on
Python 3. Python 2 behaviour defaults to using ascii encoding as before.
edit creiter: still ignore pkg-config errors by default for now as we
depend on it when glib is a subproject.
|
|
|
|
| |
as was pointed out in !12
|
|\
| |
| |
| |
| |
| |
| | |
Resolve "Memory leaks in GI regress and marshalling tests property setters"
Closes #113
See merge request GNOME/gobject-introspection!55
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Clear strv, boxed_struct, variant, boxed_glist, and object properties
during destruction. This fixes trivial memory leaks found in the PyGObject
tests suite.
https://bugzilla.gnome.org/show_bug.cgi?id=735013
Closes #113
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Clear the string property of RegressTestObj prior to setting within
the property setter.
Make sure the linked list, hash table, and string properties are cleared
during destruction.
This fixes trivial memory leaks found with the PyGObject test suite.
https://bugzilla.gnome.org/show_bug.cgi?id=735013
Closes #113
|
|/
|
|
|
|
|
|
| |
Cleanup which replaces manual loop implementation with API calls
when setting the linked list property.
https://bugzilla.gnome.org/show_bug.cgi?id=735013
Closes #113
|
|
|
|
| |
autotools is picky on Windows and errors out otherwise
|
|
|
|
|
|
| |
--libs is a subset of --ldflags which is already passed.
This works around a msys2 bug in python-config:
https://github.com/Alexpux/MINGW-packages/pull/4118
|
| |
|
|
|
|
| |
It's DWORD and gcc complained about defaulting to int.
|
|
|
|
| |
It was never defined because G_OS_WIN32 was used before the glib include.
|
|
|
|
|
|
|
|
|
| |
From what I understand this was to support a g-i built with mingw
using a different msvcrt to work with the official CPython built with
msvc.
Now that we can build with meson there is no need for this anymore as far
as I can see. Just remove the code. Please file a bug if this affects you.
|
|
|
|
|
|
|
|
|
| |
This reverts f345916405d94829696985 and related. The commit states that
both versions are about the same in performance, but the C version is more code and
harder to maintain. It also states that the behaviour re invalid control
characters is better with the C version which produces entities. But those
make any Python xml parser fail, which given that most of our tooling is Python,
doesn't seem better to me, see #135.
|
| |
|
|
|
|
|
| |
There was a stray test() function already there. Move it to a unittest
file and run it with the other tests.
|
|\
| |
| |
| |
| | |
transformer: don't shell out for filter commands and make the tests work on Windows
See merge request GNOME/gobject-introspection!54
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Windows
Instead of passing the filter commands to the shell, split them with the shlex module.
This gives us more control and a more similar behaviour on unix/win32 systems.
I could only find two users in Debian which just call python with some script, so
this seems safe to me.
Adjust the tests accordingly and use the running python executable to test the
functionality as we can be sure that's available on all platforms. Also add a
new test for the symbol filter commadn which wasn't tested before.
|
|/
|
|
|
|
|
| |
Use the test executable to generate files in a temporary location
and use difflib to compare their content.
This removes the requirement on "diff" for the test.
|
|\
| |
| |
| |
| | |
tests: Fix function incompatibility
See merge request GNOME/gobject-introspection!53
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Casting a function type to a type with more parameters is invalid and
triggers a warning on some compilers:
installed-tests/gimarshallingtests.c:
In function ‘gi_marshalling_tests_interface_get_type’: error: cast
between incompatible function types from ‘void (*)(void *)’ to
‘void (*)(void *, void *)’ [-Werror=cast-function-type]
Add those additional parameters to shut up the warnings.
https://gitlab.gnome.org/GNOME/gobject-introspection/merge_requests/53
|