Skip parsing of transitional keyword

[gylstorffq]

This skips the transitional keyword to allow parsing of kernel config of Linux kernel => 6.18. It does not implement the transitional feature as this requires additional changes.

Also add the test files to .gitignore.

This fixes #25.

[tejlmand]

Thanks for this contribution.

However, i'm not immediately convinced that we should just silence / ignore the transitional keyword.

Skipping the transitional keyword means that the symbol holding this keyword suddenly becomes a valid symbol again, which potentially can enter the .config file.

Also it seems that this silencing will allow invalid config descriptions, such as having prompts on settings with the transitional keyword.

I do understand the wish for being able to parse Kconfig trees having transitional attribute, but when supporting / silencing this attribute then we should not allow invalid Kconfig files in that same process.

If not supporting the attribute but still allowing it to be present, then such implementation should still ensure the symbol does not become or behave like a regular Kconfig symbol, which unfortunately would be the case with this PR.

I hope to find time to look into proper support for transitional.

[gylstorffq]

Proper support would be better. This was short fix to get our own use case working again.

The skip looked correct in my initial testing expect the missing assignment of value, which uses the default symbol mentioned in issue #31 :

E               config transitional
E               CONFIG_MODULES=y
E             - CONFIG_NEW_BOOL=y
E             - CONFIG_NEW_TRISTATE=m
E             + # CONFIG_NEW_BOOL is not set
E             + # CONFIG_NEW_TRISTATE is not set
E             - CONFIG_NEW_STRING="test string"
E             ?                    -----------
E             + CONFIG_NEW_STRING=""
E             - CONFIG_NEW_HEX=0x1234
E             ?                ------
E             + CONFIG_NEW_HEX=
E             - CONFIG_NEW_INT=42
E             ?                --
E             + CONFIG_NEW_INT=
E               # CONFIG_NEW_BOOL_PRECEDENCE is not set
E               CONFIG_NEW_STRING_PRECEDENCE="user value"
E               CONFIG_NEW_TRISTATE_PRECEDENCE=y
E               CONFIG_NEW_HEX_PRECEDENCE=0xABCD
E               CONFIG_NEW_INT_PRECEDENCE=100
E               # CONFIG_NEW_DISABLED is not set
E               # CONFIG_DEPENDENCY_TEST is not set
E               # CONFIG_NEW_CONDITIONAL_DEFAULT is not set
E             - # CONFIG_REGULAR_OPTION is not set
E             + # CONFIG_REGULAR_OPTION is not set
E             ? 

Comment: Fixed typo

[tejlmand]

The skip looked correct in my initial testing expect the missing assignment of value

my main problem with the current PR as it stands is that it allows invalid Kconfig definition like this:

config FOO
	bool "Foo"
	transitional

according to the specification, this symbol definition is illegal, but with this PR this syntax is accepted because the transitional keyword is silently ignored.

I think it's worse to silently accept invalid Kconfig constructions, than it is to be lacking a feature.

Note, there might be other issues regarding the PR, but this is the main one I've found so far.