Exporters: make jinja engine strict

[ilg-ul]

• configure the jinja Environment to raise exception when substitution
  variables are not defined
• trim spaces and lines, to avoid empty lines in generated files

[ilg-ul]

I may tack on a few commits after your changes in that PR to ensure that all of the exporters will work afterwards.

in case they don't, you can make this an optional feature, and activate it only in the exporters that work.

similarly for the white space options, perhaps it would be better to set them in each exporter and pass them to the base class.

[theotherjimmy]

Yeah, let me pull this down and test the exporters real quick.

[theotherjimmy]

Also, The diff is great. Nice and simple.

[theotherjimmy]

That IAR one is definitely a bug.

[theotherjimmy]

CMSIS one also clearly a bug.

[ilg-ul]

it looks like you asked for some trouble with this patch... :-(

in my opinion this setting should have been a jinja default, I fail to identify cases when ignoring missing definitions is acceptable.

[theotherjimmy]

it looks like you asked for some trouble with this patch... :-(

Yes I did. However, I think it's the good kind of trouble: it pointed out some bugs.

[theotherjimmy]

And that's all of them! They all should export again without issue.

/morph export-build

[theotherjimmy]

@ilg-ul We're almost ready to get this in. Just waiting on CI.

[0xc0170]

+1, thanks for enabling StrictUndefined

[mbed-bot]

Result: FAILURE

Your command has finished executing! Here's what you wrote!

/morph export-build

Output

mbed Build Number: 91

Exporter Build failed!

[theotherjimmy]

It looks like removing all blank lines is really bad for Makefiles.

[theotherjimmy]

@ilg-ul Would you mind if I made the boolean for removing empty lines and spaces optional and changed the makefiles to disable them?

[ilg-ul]

I would do the opposite, I would call gen_file() with additional optional arguments, and pass all to jinja.

[theotherjimmy]

I'm confused, I thought that's what I was suggesting. I worded the original suggestion poorly. I am was thinking of making gen_file take optional arguments when I asked.

[ilg-ul]

sorry, I was cryptic.

I mean calling gen_file() like:

self.gen_file('gnuarmeclipse/.project.tmpl', ctx, '.project', trim_blocks=True, lstrip_blocks=True)

the gen_file() definition should use variable args, and pass everything to jinja, not only those two, in case more flags will be necessary.

[theotherjimmy]

Yes, with True being the default for both, and the Makefile exporter explicitly calling with False.

So:

def gen(teplate_file, context, output, **kwargs):
    ...
    template.render(... **kwargs)

or thereabouts.

[ilg-ul]

hmmm... it looks I have difficulties to explain.

only exporters that need additional jinga options pass them, all existing ones need pass nothing, as they do now, so jinja defaults will apply.

[ilg-ul]

as implementation, add **kwargs to gen_file().

[theotherjimmy]

Also, Github is not the best way to do IRC... I edited by comment, probably while you were typing up a response to the old version 😆

[ilg-ul]

ok, go ahead with the changes.

[theotherjimmy]

/morph export-build

[theotherjimmy]

@ilg-ul I made the changes, could you review them? I want to make sure that we understood each other correctly.

[ilg-ul]

looks ok, please don't forget to add the options to the GNU ARM Eclipse exporter later

[theotherjimmy]

Yup! I have made a note.

[mbed-bot]

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph export-build

Output

mbed Build Number: 92

All exports and builds passed!

[ilg-ul]

Great!

What next now? Do we patiently wait for the planets to align in order for the PR to be merged? Or we better hum some incantations to hurry things up? ;-)

[theotherjimmy]

The incantation is: @sg- could you take a look?

[sg-]

LGTM from 38k ft.

[ilg-ul]

Thank you, @sg-!

FL 380 is pretty high. On the way to ISS? ;-)

[ilg-ul]

The next incantation should address the magic Merge button...

[theotherjimmy]

actually, I was hoping that @sg- would push the Merge button, but we could get @0xc0170 to do it instead.

[0xc0170]

actually, I was hoping that @sg- would push the Merge button, but we could get @0xc0170 to do it instead.

Will be reviewed on Monday (=integrated)

[ilg-ul]

any progress?

[ilg-ul]

thank you, Martin!