@hsutter

I am also wondering whether there should be an ‘exception’ regarding abstract classes. If a class has pure virtual functions, is it still necessary to declare copy operations at all? Technically, it is not necessary, as slicing cannot happen.

I.e., I prefer

class Serializable {
public:
    virtual ~Serializable() = default;
    virtual void Serialize(vector<byte>& buffer) = 0;
};

to

class Serializable {
public:
    virtual ~Serializable() = default;
    virtual void Serialize(vector<byte>& buffer) = 0;

protected:
    Serializable(const Serializable&) = default;
    Serializable& operator=(const Serializable&) = default;
};

The issue as I see it as that robot currently assumes that nothing will log to the console (despite providing multiple ways to do so) while a test is running, and that it is ok to leave cursor part way through a line and assume it will not move.

I can think of 4 different ways this could be solved (I prefer option 3)

Option 1 - Keep track of whether anything logs to console during the test run

The first time (in a test) anything logs to console, print a newline char before the output, so the new message goes on a new line.

When the test finishes, check whether anything has logged to the console. It anything has written to the console, output the whole line again, so the console output looks like this:

==============================================================================
Suite                                                                         
==============================================================================
Test1 :: some doc                                                     
hi
Test1 :: some doc                                                     | PASS |
------------------------------------------------------------------------------
Suite                                                                 | PASS |
1 test, 1 passed, 0 failed
==============================================================================

Option 2 - Only output full lines

Don't log the test line until the test finishes, and then log the whole line (incl newline) in one go. The cursor is always left at the start of a new line.

==============================================================================
Suite                                                                         
==============================================================================
hi
Test1 :: some doc                                                     | PASS |
------------------------------------------------------------------------------
Suite                                                                 | PASS |
1 test, 1 passed, 0 failed
==============================================================================

Option 3 - Have the existing robot output be a listener that can be turned off

Have the output (option 1 or 2) being done by a listener that is enabled by default, but turned off using the --console switch. Users could then develop their own listener based on that one if they want it to behave differently.

Option 4 - No change

Don't change anything - users who want this behaviour can implement using --console none](https://robotframework.org/robotframework/latest/RobotFrameworkUserGuide.html#console-output-type), and their own listener that uses logger.console

Similarly to ghdl-ls, this PR adds a ghdl-dom entrypoint when pyGHDL is installed, so that user can analise their sources directly:

$ ghdl-dom path/to/a/file
# or
$ ghdl-dom path/to/a/file path/to/anothe/file ...

For instance:

# ghdl-dom libraries/ieee/math_real.vhdl libraries/ieee2008/numeric_bit-body.vhdl
Unknown expression kind 'String_Literal8'(11) in expression '1294'.
Unknown expression kind 'Aggregate'(185) in expression '1298'.

This PR adds a pyproject.toml file for specifying the line-length used by black, which is increased from 88 to 120.

Commented the great PR!

Should this kind of documentation rather go to the documentation of Set Test Variable itself?

Notice that listeners do get all logged messages if they implement log_message method. A problem with messages logged using print() is that they are logged only after the keyword finishes, not when print() is actually called. That's due to our somewhat naïve stdout/stderr intercepting implementation what could possibly be enhanced.

I don't think the implementation above (and in PR #3988) is a good idea for these two important reasons:

• Printed messages should not be written to the console by default because most of the time they just add noise. Being able to enable that could be considered.
• The implementation only handles messages logged using print(), not messages logged using robot.api.logger or logging.

This PR has multiple problems:

• I don't think writing all output to the console by default is a good idea. Having a command line option to enable something like that could be considered, though. We can discuss that further in #3976.
• This feature has no tests.
• Plenty of existing tests are failing.

I don't think there's anything we really can to to fix this. The problem is that we have a single stdout stream and write to it from multiple sources that cannot easily cooperate. It would probably be easiest to just create a separate GUI runner that could show execution in one window/pane and logged messages in another. That should be implemented as a separate project outside the core, though. Or do you have ideas how this could be handled?

This PR adds a job matrix pytest script for testing DOM on MSYS2. By using the feature introduced in #1794, all the *.vhdl sources in the repository are processed. It needs ~4-6~ 8-12 min, however, it's still faster than some other jobs in the same workflow.

My understading is that these jobs cannot fail. When the exceptions are handled, the exit code is 0. When assertion failure happen (#1793) it seems that the exit code is 0 too. As a result, they are not very useful as an automated test yet, but just for human inspection. When the missing features are completed, we might improve DOM.py for it to actually fail.

Codecov Report

Merging #4007 (b15d0ab) into master (c53f34e) will decrease coverage by 1.79%. The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master    #4007      +/-   ##
==========================================
- Coverage   75.11%   73.33%   -1.78%     
==========================================
  Files         222      219       -3     
  Lines       18840    18292     -548     
  Branches     3083     2881     -202     
==========================================
- Hits        14150    13412     -738     
- Misses       4163     4307     +144     
- Partials      527      573      +46     

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update fc57664...b15d0ab. Read the comment docs.

Please add a newline here.

Yes, we are picky about whitespace usage. The reason is that otherwise there are unnecessary whitespace in diffs in the future.

Check Test Tags could be used here as well.

Extra empty lines at the end of the file should be removed. There should, however, be a new line at the end of the file.

We have a separate Check Test Tags keyword that could be used to replace all the above like

Check Test Tags    ${TESTNAME}    Tag0    Tag1    Tag2    Tag3

Indentation looks strange. Names should be aligned like

keywords_to_execute = ('BuiltIn.Import Library',
                       'BuiltIn.Set Library Search Order',
                       'BuiltIn.Set Tags',
                       'BuiltIn.Remove Tags')

Did I already mention we are picky about whitespace? :smile:

This is working as desired now:

========== 17 passed, 1140 xfailed, 1607 xpassed in 465.68s (0:07:45) ==========

Instead of importing and calling DOM, it is called as an external script/tool. That "fixes" the issue with other tests being broken.

I added options -v and -s to the pytest call because the captured output is not shown otherwise. When the remaining features are added and we remove the xfail marks, we might remove those verbosity options too.

Since this is now a pyunit test, it's executed on all jobs, not only on Windows.

The aforementioned problem was fixed in the above commit.

Here is the test case:

*** Test Cases ***
Testing 2
   IF  1 == 1
       Log  hello
   ELSE
       Log  none
   END
   Keyword

*** Keywords ***
Keyword
   IF  1 == 1
       Fail
   ELSE
       Log  none
   END

Then do rebot --log NONE --report NONE --xunit NONE --consolecolors off --NoStatusRC --removekeywords all --output output_invalid.xml output.xml If you do: rebot --log NONE --report NONE --xunit NONE --consolecolors off --NoStatusRC --removekeywords all --output output_invalid.xml output_invalid.xml Boom =>

[ ERROR ] Reading XML source 'output_invalid.xml' failed: Incompatible child element 'doc' for 'if'.

Try --help for usage information.

The defaults attribute was intentionally removed in 2.9 (see 9b2a4f49c61dc81c51db1a3f30aeb90e4c700ecd). Looks like the docs need to be updated to reflect this.

I merged #4010 and it fixes the crash. A problem now is that the IF structure is completely removed from the log file. Considering that all keyword data is removed, we could consider this fine but I'd like to have something left there. Anyway, this is a lot smaller problem than the crash.

Please have a look at TODO 4. Other stuff is done so far and ready for merge.

@pekkaklarck @mkorpela Hi, guys! I try to fix this problem. Tests before fix: After fix:

Looks good! The updated test makes sure there's no crash, but it would be good to also actually validate that IF works correctly. I can add a test for that separately. I also want to anyway check how this looks in the log file.

Thanks for review @mkorpela and even bigger thanks for the PR @asonkeri!