Make tests stop build on error

[fredrik-sk]

Update script to make the automatic tests catch when testfiles should be updated.
(It just happend again.)

[lategoodbye]

I like this change, but i want to add normalized output into test before merging this.

[fredrik-sk]

I'm not sure what normalized outputs means.
I would like this to be merged ASAP so noone can issue a pull request and forgot to update tests.

If you modify the output, you will have to touch all "golden files".

[lategoodbye]

I'm not sure what normalized outputs means.

mbus_parse_hex supports a option -n. This supports parsing the M-Bus frame the new (normalized value) way found in mbus-protocol-aux. The normalized output is better for machine processing, while the default output (mbus-protocol) is more intended for humans.

I would like this to be merged ASAP so noone can issue a pull request and forgot to update tests.

I understand but your check would only handle half of the truth.

If you modify the output, you will have to touch all "golden files".

No, we need to add all the normalized output files.

[lategoodbye]

Here as an example for allmess_cf50.hex

default output:

    <DataRecord id="4">
        <Function>Instantaneous value</Function>
        <StorageNumber>0</StorageNumber>
        <Unit>Flow temperature (1e-1 deg C)</Unit>
        <Value>1288</Value>
    </DataRecord>

normalized output:

    <DataRecord id="4">
        <Function>Instantaneous value</Function>
        <StorageNumber>0</StorageNumber>
        <Unit>°C</Unit>
        <Quantity>Flow temperature</Quantity>
        <Value>128.800000</Value>
    </DataRecord>

[lategoodbye]

The necessary changes for generate-xml.sh are available in #168

[fredrik-sk]

Updated/rebased onto rscada/generate-norm-xml. Found one or two things that have bothered me. Please review!

[fredrik-sk]

Pushed a new version

[fredrik-sk]

@lategoodbye : Can you review this PR please?

[lategoodbye]

The fact that we expect parse errors is bad. I suggest to move the relevant hex files into unsupported-frames.

[fredrik-sk]

There can be a point in doing negative tests.
I see that we have two more directories, unsupported-frames and error-frames.

Should we do execute the code for these directories as well?

Since the files are named: invalid_length.hex and invalid_length2.hex, maybe error-frames/ is more appropriate?

[lategoodbye]

Yes, negative tests are a good point. The only one we currently have are invalid_length*.hex
If you want to integrate this in the pull request make two different directories like test-frames-good and test-frames-bad. But if you go this way, this logic should good out of generate-xml.sh

error-frames are valid errors messages send by the slave. The mentioned hex files doesn't fit in there. unsupported-frames are for currently unsupported frames so executing code on these directories doesn't make sense to me.

[fredrik-sk]

I added a simple if statement for the names of the directories that should contain errors.
It seem simplest to keep this information in this script, otherwise we would have some other script know that error-frames should have 30 parse errors, and pass that information to generate-xml.sh.

See 62b5829 and 50e3141.

[lategoodbye]

The fact that we code the expectation of x parse errors is the whole problem. I don't want this in a script to generate XML files. Please keep it simple and easy to maintain. Also i don't want that we run generate-xml on the unsupported frames as a test. Except of that i'm happy to pull.

[fredrik-sk]

@lategoodbye , could you have another look?

[fredrik-sk]

@lategoodbye : New version, where I no longer check the count of errors, and instead accept that the two other directories does not exit with code zero.

[lategoodbye]

Thanks