(#251): Surface::create_similar() and friends should return a Result #287

federicomenaquintero · 2019-10-09T23:49:58Z

Cairo's surface creation functions never return NULL; instead they
always return a surface, but it may be in an error state. In #141 we
started making the binding functions return Result for this; some
returned the plain FooSurface type, some others Option.

This makes the following functions return Result<FooSurface, Status>

Surface::create_similar()
Surface::create_similar_image()
RecordingSurface::create()
Device.surface_create()
Device.surface_create_for_target()

The foundation for all of this is that Surface::from_raw_full() now
also returns Result<FooSurface, Status>. This is to make things
consistent with ImageSurface::from_raw_full(). Analogously, we now
have RecordingSurface::from_raw_full() that also returns Result.

Fixes #251

federicomenaquintero · 2019-10-10T00:07:11Z

Bleh, forgot to check the build with features. I'll fix this tomorrow.

…esult Cairo's surface creation functions never return NULL; instead they always return a surface, but it may be in an error state. In gtk-rs#141 we started making the binding functions return Result for this; some returned the plain FooSurface type, some others Option<FooSurface>. This makes the following functions return Result<FooSurface, Status> Surface::create_similar() Surface::create_similar_image() Device.surface_create() Device.surface_create_for_target() PdfSurface::new() RecordingSurface::create() RecordingSurface::from_raw_full() SvgSurface::new() XCBSurface::create() XCBSurface::create_for_bitmap() XCBSurface::create_with_xrender_format() From macro for_stream_constructors!: *::for_stream() *::for_raw_stream() The foundation for all of this is that Surface::from_raw_full() now also returns Result<FooSurface, Status>. This is to make things consistent with ImageSurface::from_raw_full(). Analogously, we now have RecordingSurface::from_raw_full() that also returns Result. Fixes gtk-rs#251

These may fail due to the caller passing a too-big size.

Most of cairo_pdf_surface_*() can set the surface to an error state, due to their internal call to _extract_pdf_surface(), which checks a few conditions with the surface itself and the paginated surface.

@EPashkin

Thanks to @EPashkin for the suggestion.

federicomenaquintero · 2019-10-10T15:42:06Z

Fixed the features.

Also, went along and fixed all the PdfSurface functions that may set &self in an error state.

EPashkin · 2019-10-10T15:48:39Z

src/pdf.rs

@@ -170,7 +176,7 @@ impl FromGlibPtrFull<*mut ffi::cairo_surface_t> for PdfSurface {
    #[inline]
    unsafe fn from_glib_full(ptr: *mut ffi::cairo_surface_t) -> PdfSurface {
        Self {
-            inner: Surface::from_raw_full(ptr),
+            inner: Surface::from_raw_full(ptr).unwrap(),


Maybe expect here

Uuh, I think this needs a treatment similar to ImageSurface, with TryFrom.

EPashkin · 2019-10-10T15:54:31Z

AppVeyor's errors are legit

sdroege · 2019-10-10T18:08:49Z

src/pdf.rs

        unsafe {
            ffi::cairo_pdf_surface_restrict_to_version(self.inner.to_raw_none(), version.into());
        }
+        self.status().to_result(())


What does this mean exactly? Is the surface still usable afterwards or is it completely broken? :)

Most cairo_pdf_surface_*() functions start by calling _extract_pdf_surface(), which can set the surface in an error state.

Afterwards... not all of cairo_surface_*() check for the surface->status. It's not completely unusable if in error, but some functions will still work ¯\(ツ)/¯

sdroege · 2019-10-10T18:12:50Z

Looks good to me generally, thanks for the work :)

federicomenaquintero · 2019-10-10T21:54:43Z

Let me see if I can fix the TryFrom thing - to make it consistent with other concrete surfaces that wrap the abstract one.

sdroege · 2019-10-22T07:27:20Z

Any updates here?

sdroege · 2019-11-17T21:35:37Z

@federicomenaquintero we'd like to get a release out next weekend. Any chance this can be ready by then? :)

federicomenaquintero · 2019-11-18T17:49:10Z

I'll try to work on this this week. Need to remind myself what's left to be done :)

Also, do the from_glib_*() in the same fashion as ImageSurface.

sdroege · 2019-11-20T11:17:23Z

@federicomenaquintero Thanks, let us know when it's time for another review or if you're stuck somewhere :)

federicomenaquintero · 2019-11-20T17:31:16Z

I think it's ready.

The next step is probably to take all the functions that now all look alike - FooSurface::try_from(Surface); the to_glib ones - and make a macro for them or something.

federicomenaquintero · 2019-11-20T17:54:05Z

Argh, QuartzSurface. I'll deal with it.

… ones

… others

This makes them consistent, and should make it possible to generate them with a macro.

This includes implementations of: struct DerivedSurface(Surface) TryFrom<Surface> DerivedSurface.from_raw_full() ToGlibPtr FromGlibPtrNone FromGlibPtrBorrow FromGlibPtrFull gvalue_impl! Deref Clone Display

federicomenaquintero · 2019-11-20T20:30:29Z

Now all derived surfaces are done with a macro. This even buys us some methods that were missing from the old implementations.

sdroege · 2019-11-20T21:17:11Z

Looks good to me, thanks! This cleans up everything quite nicely.

The CI failure in the examples is expected, so this is IMHO good to go once all jobs have failed only when building the examples and not before that @GuillaumeGomez

sdroege · 2019-11-21T11:29:50Z

@GuillaumeGomez Please trigger CI once gtk-rs/examples#271 is merged and then this can go in IMHO

sdroege · 2019-11-21T12:31:22Z

@GuillaumeGomez @federicomenaquintero minimum rust version in the CI needs to be updated to 1.39 or it will fail :) Can one of you push a fix for that to this branch?

sdroege · 2019-11-21T17:14:55Z

Now only misses a cargo fmt run. @GuillaumeGomez? :)

federicomenaquintero · 2019-11-21T19:18:37Z

Uuuuh, what changed that requires Rust 1.39? I mean, I need to update things in Suse, anyway, but just to know.

federicomenaquintero · 2019-11-21T19:18:49Z

Thanks for the merge!

GuillaumeGomez · 2019-11-21T19:21:12Z

async/await in other crates. That wouldn't make sense to have cairo handle its version on its own.

federicomenaquintero · 2019-11-21T19:21:39Z

Got it. Sounds good, thanks!

federicomenaquintero added 4 commits October 10, 2019 10:02

xcb: return Result from the functions that take a size

b354658

These may fail due to the caller passing a too-big size.

pdf: Return Result from most functions

a4e7c3f

Most of cairo_pdf_surface_*() can set the surface to an error state, due to their internal call to _extract_pdf_surface(), which checks a few conditions with the surface itself and the paginated surface.

Add a Status.to_result() and use it throughout instead of match{}

d40cd7c

Thanks to @EPashkin for the suggestion.

federicomenaquintero force-pushed the surface-create-similar-result branch from f4997e6 to d40cd7c Compare October 10, 2019 15:40

EPashkin reviewed Oct 10, 2019

View reviewed changes

Use the full Status type name in the for_stream_constructors! macro

7f2e839

sdroege reviewed Oct 10, 2019

View reviewed changes

federicomenaquintero added 3 commits November 19, 2019 17:52

Add PdfSurface::try_from(Surface)

2b7ed5b

Also, do the from_glib_*() in the same fashion as ImageSurface.

Add PsSurface::try_from(Surface)

2beea0a

Also, do the from_glib_*() in the same fashion as ImageSurface.

Add SvgSurface::try_from(Surface)

a7fe8f9

Also, do the from_glib_*() in the same fashion as ImageSurface.

federicomenaquintero added 5 commits November 20, 2019 13:57

QuartzSurface: do from_raw_full() in the same fashion as the previous…

b4a8778

… ones

RecordingSurface: do the from_glib impls in a way consistent with the…

dfc78b6

… others

Win32Surface: do from_raw_full() consistently with the others

248da91

Make all derived surfaces newtypes intead of having an "inner" field

f9a32bf

This makes them consistent, and should make it possible to generate them with a macro.

Generate all derived surfaces with a declare_surface! macro

275cc3c

This includes implementations of: struct DerivedSurface(Surface) TryFrom<Surface> DerivedSurface.from_raw_full() ToGlibPtr FromGlibPtrNone FromGlibPtrBorrow FromGlibPtrFull gvalue_impl! Deref Clone Display

Also do XCBSurface with declare_surface!

5fe7cea

Update minimum rust version to 1.39

6ea6f03

fmt

1794024

GuillaumeGomez merged commit e72b056 into gtk-rs:master Nov 21, 2019

federicomenaquintero deleted the surface-create-similar-result branch November 21, 2019 19:21

Uh oh!

(#251): Surface::create_similar() and friends should return a Result #287

(#251): Surface::create_similar() and friends should return a Result #287

Uh oh!

Conversation

federicomenaquintero commented Oct 9, 2019

Uh oh!

federicomenaquintero commented Oct 10, 2019

Uh oh!

federicomenaquintero commented Oct 10, 2019

Uh oh!

EPashkin Oct 10, 2019

Choose a reason for hiding this comment

Uh oh!

federicomenaquintero Oct 10, 2019

Choose a reason for hiding this comment

Uh oh!

EPashkin commented Oct 10, 2019

Uh oh!

sdroege Oct 10, 2019

Choose a reason for hiding this comment

Uh oh!

federicomenaquintero Oct 10, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sdroege commented Oct 10, 2019

Uh oh!

federicomenaquintero commented Oct 10, 2019

Uh oh!

sdroege commented Oct 22, 2019

Uh oh!

sdroege commented Nov 17, 2019

Uh oh!

federicomenaquintero commented Nov 18, 2019

Uh oh!

sdroege commented Nov 20, 2019

Uh oh!

federicomenaquintero commented Nov 20, 2019

Uh oh!

federicomenaquintero commented Nov 20, 2019

Uh oh!

federicomenaquintero commented Nov 20, 2019

Uh oh!

sdroege commented Nov 20, 2019

Uh oh!

sdroege commented Nov 21, 2019

Uh oh!

sdroege commented Nov 21, 2019

Uh oh!

sdroege commented Nov 21, 2019

Uh oh!

federicomenaquintero commented Nov 21, 2019

Uh oh!

federicomenaquintero commented Nov 21, 2019

Uh oh!

GuillaumeGomez commented Nov 21, 2019

Uh oh!

federicomenaquintero commented Nov 21, 2019

Uh oh!

Uh oh!

federicomenaquintero Oct 10, 2019 •

edited

Loading