To avoid git complaining about duplicate worktrees, detach
the temporary trees. To avoid python modules translations
changing order (depending on how find traverses the tree),
sort the filenames before extraction.
- this would be the first 3.3 release, scripts still needed
cleanup to handle all the changes in 3.3
- support "-alpha1" kinds of designation
- support more versioning information in the headers
- use clang-format12 or 13 or unversioned executable, only
- check it's actually 12 or 13
- set language standard explicitly to C++17, "Cpp11" is now
an alias for "latest" which is weird
This change does lead to some thrashing when applying styles,
so I'm not going to do that across the board right now. Changes include:
- extra spaces in lambda-captures
- nicer alignment of lambda-bodies
The ordering of entries jumps around sometimes when reading from
Transifex (this might be Python unordered dictionaries, or based
on translation statistics -- I can't tell). Force an order by
sorting on language code and key-name so they all end up grouped
by language code, sorted Name Icon GenericName Comment.
Although this shuffles some more entries now, longer-term it
will reduce churn in the .desktop file.
This introduces a stub-implementation (fake) that mimics the
API offered by libcalamares (the library is actually exposed
to Python via Boost::Python, so it doesn't act like a C-extension).
Using that stub-implementation, we can check Python modules for
validity as part of the test-suite.
The stub-implementation is needed, because otherwise every
Python module already fails at `import libcalamares`.
- stub-implement the API that is actually used by the Python modules
- in globalstorage, be slightly smart about what keys are being
requested (so that e.g. all the modules that handle partitions
information get an empty list and can manipulate that, instead of
erroring out when they get a string)
