Merge lp:~vanvugt/mir/revive-trusty into lp:mir

[Daniel van Vugt]

Revive support for building on trusty and generally improve portability
of the source code (LP: #1418962)

In addition to 3v1n0's initial fixes, this proposal makes lp:mir buildable without any hacks. Just using standard trusty packages and gcc.

[Chris Halse Rogers (raof)]

The test changes are making a non-trivial change to the expectations; presumably that's because the UnorderedElements matcher is a newly added feature?

review: Needs Information

[PS Jenkins bot (ps-jenkins)]

PASSED: Continuous integration, rev:2319
http://jenkins.qa.ubuntu.com/job/mir-ci/2980/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-vivid-i386-build/1306
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-vivid-amd64-build/1305
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-vivid-touch/1262
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-vivid-amd64-ci/977
        deb: http://jenkins.qa.ubuntu.com/job/mir-vivid-amd64-ci/977/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-vivid-armhf/1262
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-vivid-armhf/1262/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/4299
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18075

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-ci/2980/rebuild

review: Approve (continuous-integration)

[Daniel van Vugt (vanvugt)]

Yep, apparently UnorderedElementsAre is quite new (post-trusty). ElementsAre has been around longer.

[Chris Halse Rogers (raof)]

Rather than adding an unwanted constraint, could you implement UnorderedElementsAre when it doesn't already exist?

[Alexandros Frantzis (afrantzis)]

First of all we need to make an official decision about whether we want to support trusty, or, more generally, what's the oldest release we want to support at any point in time (perhaps there is a decision and I am not aware of it?).

Also, as discussed before, unless we have a trusty build as part of CI, we will keep on breaking compatibility.

> Rather than adding an unwanted constraint, could you implement UnorderedElementsAre when it doesn't already exist?

+1, Rectangles wasn't meant to guarantee a particular ordering of the contained rectangles (although the current implementation does retain input order, but that's a detail we should depend on).

review: Needs Fixing

[Alan Griffiths (alan-griffiths)]

> Rather than adding an unwanted constraint, could you implement
> UnorderedElementsAre when it doesn't already exist?

+1

[Alan Griffiths (alan-griffiths)]

> > Rather than adding an unwanted constraint, could you implement
> > UnorderedElementsAre when it doesn't already exist?
>
> +1

review: Needs Fixing

[Daniel van Vugt (vanvugt)]

The test case is a trivial problem. I'd rather delete it than argue, but it can also be rewritten in other ways...

Yes, trusty is important. We need to give users like Trevinho the option of using the the latest LTS release and still be able to work with the Mir source. Even if we didn't care about trusty, the improved portability is worth having (until it becomes too painful as the Android headers have).

[lp:~vanvugt/mir/revive-trusty]

2320. By Daniel van Vugt on 2015-02-17

Merge latest trunk

2321. By Daniel van Vugt on 2015-02-17

Rewrite test case in a way that doesn't assume ordering.

2322. By Daniel van Vugt on 2015-02-17

Rewrite in the horizontal style

[PS Jenkins bot (ps-jenkins)]

PASSED: Continuous integration, rev:2322
http://jenkins.qa.ubuntu.com/job/mir-ci/2987/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-vivid-i386-build/1316
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-vivid-amd64-build/1315
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-vivid-touch/1272
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-vivid-amd64-ci/984
        deb: http://jenkins.qa.ubuntu.com/job/mir-vivid-amd64-ci/984/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-vivid-armhf/1272
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-vivid-armhf/1272/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/4309
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18099

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-ci/2987/rebuild

review: Approve (continuous-integration)

[Alexandros Frantzis (afrantzis)]

120 + static std::vector<Rectangle> unique_elements(Rectangles const& rects)

No need for static.

156 + EXPECT_THAT(unique_elements(rectangles),
157 + AllOf(SizeIs(2), Contains(rect[0]), Contains(rect[2])));

The test, as written before, checked exactly the post-conditions that were required from the Rectangles class. That is:

1. Rectangles can contain multiple equal rectangles.
2. Rectangles does not guarantee ordering of the rectangles.

The new version of the test ignores (1). Granted, we do check (1) in a dedicated test, but in the dedicated test we are dealing with a Rectangles object containing multiple rectangles of a single kind only. The changed test checks behavior in the presence of different kinds of rectangles.

It seems wrong to make a test less expressive only to work around backward compatibility issues.

review: Needs Fixing

[Alan Griffiths (alan-griffiths)]

> It seems wrong to make a test less expressive only to work around backward
> compatibility issues.

+1

Is it even useful? Who needs to be able to build /the tests/ on trusty?

review: Needs Fixing

[Robert Carr (robertcarr)]

We can settle the issue of whether this is required out of MP (as it'so pened on the mailing list now).

If required this seems ok to me. (though maybe its worth encapsulating Alf's comment in to a source code comment.

review: Abstain

[Daniel van Vugt (vanvugt)]

I think the needs fixing comments are debatable. But easier to satisfy than argue.

[lp:~vanvugt/mir/revive-trusty]

2323. By Daniel van Vugt on 2015-02-18

Merge latest trunk

2324. By Daniel van Vugt on 2015-02-18

Don't mention "static".
Also restore support for testing multiple identical rectangles.

[PS Jenkins bot (ps-jenkins)]

PASSED: Continuous integration, rev:2324
http://jenkins.qa.ubuntu.com/job/mir-ci/3003/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-vivid-i386-build/1336
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-clang-vivid-amd64-build/1335
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-vivid-touch/1292
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-vivid-amd64-ci/1000
        deb: http://jenkins.qa.ubuntu.com/job/mir-vivid-amd64-ci/1000/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-vivid-armhf/1292
        deb: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-vivid-armhf/1292/artifact/work/output/*zip*/output.zip
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/4325
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18140

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-ci/3003/rebuild

review: Approve (continuous-integration)

[Alan Griffiths (alan-griffiths)]

The original version of the tests is easier to follow so this change needs more justification than has been presented to date.

review: Disapprove

[Robert Carr (robertcarr)]

If someone who hasn't reviewed this yet marks resubmit it could have one of each review status.

Unmerged revisions

2324. By Daniel van Vugt on 2015-02-18

Don't mention "static".
Also restore support for testing multiple identical rectangles.

2323. By Daniel van Vugt on 2015-02-18

Merge latest trunk

2322. By Daniel van Vugt on 2015-02-17

Rewrite in the horizontal style

2321. By Daniel van Vugt on 2015-02-17

Rewrite test case in a way that doesn't assume ordering.

2320. By Daniel van Vugt on 2015-02-17

Merge latest trunk

2319. By Daniel van Vugt on 2015-02-16

Update docs

2318. By Daniel van Vugt on 2015-02-16

Fully revive support for building on trusty (even with gcc). The only
caveat is that Android support is not available. So you need to build
with:
  cmake -DMIR_PLATFORM=mesa

2317. By Daniel van Vugt on 2015-02-16

Drop Android hacks. Android builds on trusty are too messy to support now.

2316. By Daniel van Vugt on 2015-02-16

Merge latest trunk

2315. By Daniel van Vugt on 2015-02-12

Start reviving trusty support. But it's getting messy.

=== modified file 'debian/control'
--- debian/control	2015-02-18 03:04:18 +0000
+++ debian/control	2015-02-18 03:31:48 +0000
@@ -37,6 +37,7 @@
                umockdev (>= 0.8.7),
                libudev-dev,
                google-mock (>= 1.6.0+svn437),
+               libgtest-dev,
                valgrind [!arm64],
                libglib2.0-dev,
 Standards-Version: 3.9.4

=== modified file 'doc/building_source_for_pc.md'
--- doc/building_source_for_pc.md	2013-08-28 03:41:48 +0000
+++ doc/building_source_for_pc.md	2015-02-18 03:31:48 +0000
@@ -37,10 +37,19 @@
 
     $ cmake-gui ..
 
-The next step is to build the source and run the tests:
-
-    $ make (-j8)
-    $ ctest
+The next step is to build the source:
+
+    $ make
+
+If you get build failures in Android-related files (e.g. on trusty), you
+can reconfigure the build to skip those and only build for desktop:
+
+    $ cmake .. -DMIR_PLATFORM=mesa
+    $ make
+
+Run the tests:
+
+    $ make test   # or ctest
 
 To install Mir just use the normal make install command:
 

=== modified file 'src/client/mir_screencast.h'
--- src/client/mir_screencast.h	2015-02-13 06:12:34 +0000
+++ src/client/mir_screencast.h	2015-02-18 03:31:48 +0000
@@ -26,6 +26,7 @@
 #include "mir/geometry/rectangle.h"
 
 #include <EGL/eglplatform.h>
+#include <memory>
 
 namespace mir
 {

=== modified file 'src/platforms/mesa/server/cursor.cpp'
--- src/platforms/mesa/server/cursor.cpp	2015-02-17 02:35:11 +0000
+++ src/platforms/mesa/server/cursor.cpp	2015-02-18 03:31:48 +0000
@@ -32,6 +32,11 @@
 #include <stdexcept>
 #include <vector>
 
+// Detect older versions of gbm.h
+#ifndef GBM_BO_IMPORT_FD  // Unrelated, but the only detectable macro change
+#define GBM_BO_USE_CURSOR GBM_BO_USE_CURSOR_64X64
+#endif
+
 namespace mg = mir::graphics;
 namespace mgm = mg::mesa;
 namespace geom = mir::geometry;

=== modified file 'src/server/symbols.map'
--- src/server/symbols.map	2015-02-13 06:12:34 +0000
+++ src/server/symbols.map	2015-02-18 03:31:48 +0000
@@ -765,3 +765,11 @@
   };
  local: *;
 };
+
+MIR_SERVER_29.1 {
+ global:
+  extern "C++" {
+    vtable?for?mir::DefaultServerConfiguration;
+  };
+} MIR_SERVER_29;
+

=== modified file 'tests/acceptance-tests/test_client_surface_visibility.cpp'
--- tests/acceptance-tests/test_client_surface_visibility.cpp	2015-02-13 06:12:34 +0000
+++ tests/acceptance-tests/test_client_surface_visibility.cpp	2015-02-18 03:31:48 +0000
@@ -50,7 +50,7 @@
     StoringShell(
         std::shared_ptr<msh::Shell> const& wrapped,
         std::shared_ptr<ms::SurfaceCoordinator> const surface_coordinator) :
-        msh::ShellWrapper{wrapped},
+        msh::ShellWrapper(wrapped),
         surface_coordinator{surface_coordinator}
     {}
 

=== modified file 'tests/acceptance-tests/throwback/test_focus_selection.cpp'
--- tests/acceptance-tests/throwback/test_focus_selection.cpp	2015-02-13 06:12:34 +0000
+++ tests/acceptance-tests/throwback/test_focus_selection.cpp	2015-02-18 03:31:48 +0000
@@ -49,7 +49,7 @@
 struct MockShell : msh::ShellWrapper
 {
     explicit MockShell(std::shared_ptr<msh::Shell> const& wrapped) :
-        msh::ShellWrapper{wrapped}
+        msh::ShellWrapper(wrapped)
     {
         ON_CALL(*this, open_session(_, _, _)).
             WillByDefault(Invoke(this, &MockShell::unmocked_open_session));

=== modified file 'tests/unit-tests/geometry/test-rectangles.cpp'
--- tests/unit-tests/geometry/test-rectangles.cpp	2015-02-13 06:12:34 +0000
+++ tests/unit-tests/geometry/test-rectangles.cpp	2015-02-18 03:31:48 +0000
@@ -23,6 +23,7 @@
 
 #include <iterator>
 #include <algorithm>
+#include <initializer_list>
 
 using namespace mir::geometry;
 using namespace testing;
@@ -32,15 +33,24 @@
 struct TestRectangles : Test
 {
     Rectangles rectangles;
-
-    auto contents_of(Rectangles const& rects) ->
-        std::vector<Rectangle>
-    {
-        return {std::begin(rects), std::end(rects)};
-    }
 };
+
+bool equivalent_unordered(Rectangles const& rects,
+                          std::initializer_list<Rectangle> const& expect)
+{
+    std::vector<Rectangle> check(rects.begin(), rects.end());
+    for (auto const& e : expect)
+    {
+        auto it = std::find(check.begin(), check.end(), e);
+        if (it == check.end())
+            return false;
+        check.erase(it);
+    }
+    return check.empty();
 }
 
+} // namespace
+
 TEST_F(TestRectangles, rectangles_empty)
 {
     EXPECT_EQ(0, std::distance(rectangles.begin(), rectangles.end()));
@@ -240,23 +250,23 @@
 
     rectangles = Rectangles{rect[0], rect[1], rect[2]};
 
-    EXPECT_THAT(contents_of(rectangles), UnorderedElementsAre(rect[0], rect[1], rect[2]));
+    EXPECT_TRUE(equivalent_unordered(rectangles, {rect[0], rect[1], rect[2]}));
     EXPECT_THAT(rectangles.bounding_rectangle(), Eq(Rectangle{{0,0}, {900,700}}));
 
     rectangles.remove(rect[1]);
 
-    EXPECT_THAT(contents_of(rectangles), UnorderedElementsAre(rect[0], rect[2]));
+    EXPECT_TRUE(equivalent_unordered(rectangles, {rect[0], rect[2]}));
     EXPECT_THAT(rectangles.bounding_rectangle(), Eq(Rectangle{{0,0}, {900,600}}));
 
     rectangles.add(rect[2]);
 
-    EXPECT_THAT(contents_of(rectangles), UnorderedElementsAre(rect[0], rect[2], rect[2]));
+    EXPECT_TRUE(equivalent_unordered(rectangles, {rect[0], rect[2], rect[2]}));
     EXPECT_THAT(rectangles.bounding_rectangle(), Eq(Rectangle{{0,0}, {900,600}}));
 
     rectangles.add(rect[1]);
     rectangles.remove(rect[2]);
 
-    EXPECT_THAT(contents_of(rectangles), UnorderedElementsAre(rect[0], rect[1], rect[2]));
+    EXPECT_TRUE(equivalent_unordered(rectangles, {rect[0], rect[1], rect[2]}));
     EXPECT_THAT(rectangles.bounding_rectangle(), Eq(Rectangle{{0,0}, {900,700}}));
 }
 

=== modified file 'tests/unit-tests/graphics/mesa/test_cursor.cpp'
--- tests/unit-tests/graphics/mesa/test_cursor.cpp	2015-01-22 09:00:14 +0000
+++ tests/unit-tests/graphics/mesa/test_cursor.cpp	2015-02-18 03:31:48 +0000
@@ -35,6 +35,11 @@
 
 #include <string.h>
 
+// Detect older versions of gbm.h
+#ifndef GBM_BO_IMPORT_FD  // Unrelated, but the only detectable macro change
+#define GBM_BO_USE_CURSOR GBM_BO_USE_CURSOR_64X64
+#endif
+
 namespace mg = mir::graphics;
 namespace mgm = mir::graphics::mesa;
 namespace geom = mir::geometry;