New API in Cairo 1.12 is not wrapped

Submitted by Povilas Kanapickas

Assigned to Murray Cumming

Link to original bug (#74914)

Description

Created attachment 93975 Remove no longer useful .cvsignore files

I've wrapped most of the new Cairo 1.12 API (patches attached). The new wrappers are not ready for committing yet: I've only made sure that the library compiles, the new wrappers have not been tested.

Patch 93975, "Remove no longer useful .cvsignore files":
0001-Remove-no-longer-useful-.cvsignore-files.patch

Version: CVS HEAD

Povilas Kanapickas submitted a patch:

Patch 93976, "Wrap FtFontFace::{get,set,unset}_synthesize":
0002-Wrap-FtFontFace-get-set-unset-_synthesize.patch

Povilas Kanapickas submitted a patch:

Patch 93977, "Wrap cairo script device and script surface APIs":
0003-Wrap-cairo-script-device-and-script-surface-APIs.patch

Povilas Kanapickas submitted a patch:

Patch 93978, "Wrap XCB device and XCB surface APIs":
0004-Wrap-XCB-device-and-XCB-surface-APIs.patch

Povilas Kanapickas submitted a patch:

Patch 93979, "Wrap Xlib device API":
0005-Wrap-Xlib-device-API.patch

Povilas Kanapickas submitted a patch:

Patch 93980, "Wrap new Surface member functions":
0006-Wrap-new-Surface-member-functions.patch

Povilas Kanapickas submitted a patch:

Patch 93981, "Wrap the mesh surface API":
0007-Wrap-mesh-surface-API.patch

Murray Cumming said:

Thanks. In an email, you said: " There is one place where I'm a bit unsure how to wrap the API: cairo_surface_map_to_image requires that the returned surface is not disposed through cairo_surface_destroy, but through cairo_surface_unmap_image instead. This means we can't use RefPtr as usual. Returning a plain pointer doesn't sound as a good idea either, so I created a small wrapper modelling an unique pointer with a custom deleter (see MappedImagePtr). "

Which patches does this affect? I'd like to just apply the other patches as they seem straightforward.

Povilas Kanapickas uploaded an attachment:

Seems that I haven't committed everything. One function is enabled only in C++11 mode so that otherwise noncopyable Path can be returned using move semantics.

Attachment 94466, "Wrap the mesh surface API":
0001-Wrap-mesh-surface-API.patch

Povilas Kanapickas said:

Which patches does this affect? I'd like to just apply the other patches as they seem straightforward.

This affects the "Wrap new Surface member functions" patch. Others may indeed be worth to commit, but I think writing a toy program to test the wrappers is still a good idea.

Regards, Povilas

Murray Cumming commented on patch 93975:

Pushed.

Murray Cumming commented on patch 93976:

Pushed.

Murray Cumming commented on patch 93977:

Pushed.

Murray Cumming commented on patch 93978:

Review of attachment 93978:

Just one comment for this patch.

::: cairomm/xcb_device.h @@ +54,5 @@

• • Returns the connection for the XCB device.
• • @since 1.12
• */
• xcb_connection_t* get_connection() const;

I think this needs some comment about why it makes sense for it to be const, if it does. I don't know.

Murray Cumming said:

(In reply to comment 7)

cairo_surface_map_to_image requires that the returned surface is not disposed through cairo_surface_destroy, but through cairo_surface_unmap_image instead. This means we can't use RefPtr as usual. Returning a plain pointer doesn't sound as a good idea either, so I created a small wrapper modelling an unique pointer with a custom deleter (see MappedImagePtr).

Yes, I see that now. The C API is a bit awkward. I would have tried the same thing, but I wonder if it will be simple enough to make it worth wrapping this function.

I'm a bit surprised that you don't need a copy constructor and operator= for MappedImagePtr even just to use it as a return type. And without letting people copy it (with underlying reference-counting), you'd force people to do the (implicit) unmap in the same scope rather than at some later dynamically-chosen time.