test(linux): Add unit tests for mcompile by Markus-SWAG · Pull Request #15892 · keymanapp/keyman

Markus-SWAG · 2026-04-28T15:40:47Z

This is a first try to add unit tests to mcompile.
There is a new file converter.cpp containing a minimal main. This allows the tests to have an own main provided by googletest.
Since gsettings are used, there is the need for new include files and new library dependencies.
The separation of converter.cpp required some refactoring of the header files.

keymanapp-test-bot · 2026-04-28T15:40:57Z

User Test Results

Test specification and instructions

User tests are not required

This is temporary, when the transition to xkblibcommon is done, wayland will be available.

ermshiperete

:+1 Looks good.

As we talked in the huddle, I'm not yet approving it so that it doesn't get accidentally merged while the tests won't work on the noble ba.

darcywong00 · 2026-05-01T02:04:36Z

We have a convention for pull request titles

Reference

https://github.com/keymanapp/keyman/wiki/Pull-Request-and-Commit-workflow-notes#title

I suggest changing the title to something like test(linux): Add unit tests for mcompile

This way, the (linux) scope ensures this pull request title automatically gets added to the "Keyman for Linux changlog at
https://help.keyman.com/products/linux/version-history/

- adding header files - renaming files and methods - restructuring meson build

Co-authored-by: Eberhard Beilharz <ermshiperete@users.noreply.github.com>

…-keyboard-support-unit-tests' into feat/linux/mnemonic-keyboard-support-unit-tests

ermshiperete · 2026-05-07T10:19:19Z

+      EXPECT_EQ(keycodes.size(), expected_keysyms.size()) << "Keycodes and expected keysyms vectors must be of the same size.";
+      for (guint k = 0; k < keycodes.size() && k < expected_keysyms.size(); k++) {


I think it doesn't make sense to continue if the the two vectors have different size, so we could use ASSERT_EQ here. That would also simplify the for loop slightly. Also, since we're using a std vector it doesn't really make sense to use a Gnome type for k - doesn't really matter but will reduce dependency. (I'm not sure - is there a uint or should it be unsigned int).

Suggested change

EXPECT_EQ(keycodes.size(), expected_keysyms.size()) << "Keycodes and expected keysyms vectors must be of the same size.";

for (guint k = 0; k < keycodes.size() && k < expected_keysyms.size(); k++) {

ASSERT_EQ(keycodes.size(), expected_keysyms.size()) << "Keycodes and expected keysyms vectors must be of the same size.";

for (uint k = 0; k < keycodes.size(); k++) {

When ASSERT_* fails, doesn't it prevent all further instantiations from running?

Maybe, but my understanding/assumption is that it won't execute the following checks but continue with the next instantiation.

- formatting - improved constructors

ermshiperete

LGTM

ermshiperete · 2026-05-08T08:23:36Z

+    AltGrDe,
+    GetKeyValUnderlyingFromKeyCodeUnderlyingTest,
+    testing::ValuesIn(KeyboardTestParameters(
+                          {u'æ',      u'\xfffe', u'¢',      u'ð',      u'\xfffe', u'\xfffe', u'\xfffe', u'\xfffe', u'\xfffe', u'\xffff',


Why is it that we have u'\xffff' for some values and u'\xfffe' or u'\000' for others?

ermshiperete

Oops, forgot to check devin.ai! That found a few issues that might cause problems in the future.

Co-authored-by: Eberhard Beilharz <ermshiperete@users.noreply.github.com>

ermshiperete · 2026-05-12T13:58:13Z

+    if (test_display) {
+      gdk_display_close(test_display);
+      test_display = nullptr;
+    }


From devin.ai:

gdk_display_close in TearDown breaks subsequent parameterized test instances

The KeyboardConversionTest fixture calls gdk_display_close(test_display) in TearDown() (keymap.tests.cpp:140). This fixture is used with testing::WithParamInterface, meaning GTest creates a new fixture instance per parameter value, calling SetUp/TearDown for each. However, gdk_init() in SetUp (keymap.tests.cpp:89) is a one-time initialization — subsequent calls are no-ops and will not reopen the display. After the first test's TearDown closes the display, all subsequent test instances will call gdk_display_get_default() and get a closed display. The GDK documentation explicitly states gdk_display_close "is not a reversible operation" and "should not be called normally." This could cause the ASSERT_NE(test_display, nullptr) or ASSERT_NE(test_keymap, nullptr) checks to fail for every test after the first, or cause undefined behavior when using the keymap from a closed display.

ermshiperete · 2026-05-12T14:00:14Z

+    include_directories: [common_include_dir],
+    include_directories: [common_include_dir],


Suggested change

include_directories: [common_include_dir],

include_directories: [common_include_dir],

include_directories: [common_include_dir],

ermshiperete · 2026-05-12T14:02:43Z

+  if (KMX_DoConvert(kmxfile, bDeadkeyConversion, argc, (gchar**)argv)) {
+    if(!KMX_SaveKeyboard(kmxfile, outfile)) {
+      KMX_LogError(L"Failed to save keyboard (%d)\n", errno);
+      delete[] kmxfile;
+      return 3;
+    }
+  }
+
+  delete[] kmxfile;
+
+  return 0;


From devin.ai:

Program returns success (0) when KMX_DoConvert fails — pre-existing issue

In main.cpp:78-86, when KMX_DoConvert returns FALSE, the code falls through to delete[] kmxfile; return 0;, returning success even though conversion failed. This is the same behavior as the old code in mcompile.cpp and is consistent with the comment at mcompile.cpp:8-9 that the program 'deliberately leaks memory as it has a very short life cycle.' However, a non-zero exit code would be more appropriate for a failed conversion.

ermshiperete · 2026-05-12T14:03:48Z

+ * Mnemonic layout support for Linux
+ */
+
+#include "mcompile.h"


From devin.ai:

main.cpp relies on transitive includes for strcmp, printf, and errno

main.cpp uses strcmp (line 24), printf (lines 29, 44), and errno (lines 73, 80) but only directly includes mcompile.h. These symbols are available through transitive includes: mcompile.h → <gdk/gdk.h> → <glib.h> → various C standard headers. This works but is fragile — a change in GDK's internal includes could break compilation. The old code had the same issue. Consider adding explicit #include <cstring>, #include <cstdio>, and #include <cerrno>.

ermshiperete · 2026-05-12T14:17:20Z

+    }
+  }
+
+  delete[] kmxfile;


From devin.ai - took me a while to understand this, but I think the analysis is correct. And I guess the other two delete[] lines will have to be changed as well:

delete[] on LPKMX_KEYBOARD is technically UB but an improvement over the old delete

In main.cpp:74,81,86, delete[] kmxfile is used where kmxfile is LPKMX_KEYBOARD (i.e., KMX_KEYBOARD*), but the underlying memory was allocated as new KMX_BYTE[size] in KMX_LoadKeyboard (mc_kmxfile.cpp:439-441). Using delete[] with a different type than new[] is technically undefined behavior per the C++ standard. However, both KMX_BYTE and KMX_KEYBOARD have trivial destructors, so in practice this is benign. The old code used delete kmxfile (scalar delete on array allocation), which was even more wrong. The PR improves this, though the correct fix would be to cast back to PKMX_BYTE before deleting: delete[] reinterpret_cast<PKMX_BYTE>(kmxfile);.

Suggested change

delete[] kmxfile;

delete[] reinterpret_cast<PKMX_BYTE>(kmxfile);

ermshiperete · 2026-05-12T14:27:26Z

+    }
+  }
+
+  delete[] kmxfile;


We could consider a further refactoring (in a separate PR):
add a field (buf) to the end of KMX_KEYBOARD that holds the original data. That would be cleaner. Then we could do delete kmxfile. The d'tor of KMX_KEYBOARD could then delete[] buf.

Also, the variable name kmxfile should probably be kmxkeyboard.

ermshiperete

Always more...

ermshiperete · 2026-05-21T10:21:04Z

+  KMX_WCHAR expected_char;
+  std::string layout;
+  guint shiftstate;
+  KMX_WCHAR expected_deadkey_value;


We don't need/use this variable, and KeyboardTestDataValues (that uses it) has it's own variable with the same name.

Suggested change

KMX_WCHAR expected_deadkey_value;

adding unit tests for Linux mcompile

99baacb

Markus-SWAG requested review from SabineSIL, ermshiperete and mcdurdin April 28, 2026 15:40

keymanapp-test-bot Bot added the user-test-missing User tests have not yet been defined for the PR label Apr 28, 2026

keymanapp-test-bot Bot added this to the A19S28 milestone Apr 28, 2026

github-actions Bot added the linux/ label Apr 28, 2026

github-project-automation Bot added this to Keyman Apr 28, 2026

github-project-automation Bot moved this to Todo in Keyman Apr 28, 2026

Markus-SWAG added 3 commits April 28, 2026 18:50

adding dependency to libglibmm-2.4-dev

77acb6f

adding dependency to libglibmm-2.4-dev typo

c040314

Prevent running unit tests on Wayland

0ed8eb1

This is temporary, when the transition to xkblibcommon is done, wayland will be available.

keymanapp-test-bot Bot removed the user-test-missing User tests have not yet been defined for the PR label Apr 28, 2026

ermshiperete reviewed Apr 30, 2026

View reviewed changes

mcdurdin reviewed May 4, 2026

View reviewed changes

incorporating review results for adding unit tests

f309d90

- adding header files - renaming files and methods - restructuring meson build

Markus-SWAG changed the title ~~adding unit tests for Linux mcompile~~ test(linux): Add unit tests for mcompile May 5, 2026

ermshiperete reviewed May 6, 2026

View reviewed changes

parametrized unit tests for mcompile

8bd47e4

github-actions Bot added the test Any acceptance test issue label May 6, 2026

Markus-SWAG and others added 8 commits May 6, 2026 17:51

Update linux/mcompile/keymap/deadkey.h

d5accca

Co-authored-by: Eberhard Beilharz <ermshiperete@users.noreply.github.com>

renamed files according test standards

6e38fc0

Merge remote-tracking branch 'refs/remotes/origin/feat/linux/mnemonic…

c05bba9

…-keyboard-support-unit-tests' into feat/linux/mnemonic-keyboard-support-unit-tests

substituting SIL International by SIL Global in all file headers

bb87e1d

erasing year from the file headers

0646808

added delete kmxfile as before

e3dd7f1

added 1 more file header

8ba6a13

fixing typo

9cf4e78

adding test suite for KMX_get_KeyVal_From_KeyCode

471c6c4

ermshiperete reviewed May 7, 2026

View reviewed changes

Markus-SWAG added 2 commits May 7, 2026 17:37

incorporating review results

61f1a2b

- formatting - improved constructors

reformatting test parameters

8803089

ermshiperete approved these changes May 8, 2026

View reviewed changes

ermshiperete requested changes May 8, 2026

View reviewed changes

keyman-server modified the milestones: A19S28, A19S29 May 11, 2026

Markus-SWAG and others added 6 commits May 11, 2026 16:15

German Unit test now alos successfully finishing

c9c45a0

Update linux/mcompile/keymap/test/meson.build

0ccc9af

Co-authored-by: Eberhard Beilharz <ermshiperete@users.noreply.github.com>

Update linux/mcompile/keymap/mc_kmxfile.cpp

8cf0c41

Co-authored-by: Eberhard Beilharz <ermshiperete@users.noreply.github.com>

Update linux/mcompile/keymap/deadkey.h

cc9e0be

Co-authored-by: Eberhard Beilharz <ermshiperete@users.noreply.github.com>

Update linux/mcompile/keymap/main.cpp

7b367e9

Co-authored-by: Eberhard Beilharz <ermshiperete@users.noreply.github.com>

changes after review

28b0c96

ermshiperete reviewed May 12, 2026

View reviewed changes

ermshiperete self-requested a review May 12, 2026 14:30

ermshiperete reviewed May 12, 2026

View reviewed changes

improving tests, check on deadkey variable

fcea563

ermshiperete reviewed May 21, 2026

View reviewed changes

		EXPECT_EQ(keycodes.size(), expected_keysyms.size()) << "Keycodes and expected keysyms vectors must be of the same size.";
		for (guint k = 0; k < keycodes.size() && k < expected_keysyms.size(); k++) {

		include_directories: [common_include_dir],
		include_directories: [common_include_dir],

	delete[] kmxfile;
	delete[] reinterpret_cast<PKMX_BYTE>(kmxfile);

Uh oh!

Conversation

Markus-SWAG commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

keymanapp-test-bot Bot commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

User Test Results

Uh oh!

ermshiperete left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

darcywong00 commented May 1, 2026

Reference

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ermshiperete left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ermshiperete left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ermshiperete left a comment

Markus-SWAG commented Apr 28, 2026 •

edited

Loading

keymanapp-test-bot Bot commented Apr 28, 2026 •

edited

Loading