we need to propagate stacklevel so that the warning is emitted at the
usage line instead of deep in translations.py
i mean
we don't _need_ need to
it's just nicer
The regex enforces starting with an uppercase letter, which doesn't work for eCash and for xRhodium. An exception is added to handle this naming, as well as a comment to remind why the regex contains this special case.
This fixes `make defs_check` for eCash.
the error:
When built with PYOPT, Optiga will log its communication to the VCP. But
now that VCP config lives in coreapp and Optiga lives in kernel, there's
no way for Optiga to write to VCP, and the relevant calls simply don't
build.
We need a different method for logging inside the kernel.
Here we change all FOO=VALUE defines to be tuples ("FOO", "VALUE").
Also, VALUE is always the raw string you want to end up in the C file,
instead of attempting to shell-escape it while specifying.
By all rights scons _should_ be using shlex.quote() on the CPPDEFINES,
but it doesn't, so we hack it by specifying the define prefix as `-D'`
and suffix as `'`. That way the arguments in shell are '-escaped, and
we're (currently) not using ' in any argument value so this should work
fine.
At the same time, when passing the flags to cargo, we can shlex.quote
the whole thing and get the right strings passed all the way into
build.rs -- as long as no argument contains a comma, which is the split
character...
Supply the rust layout with dedicated paremeter type instead of plain
micropython::Obj. The types used are ConfirmBlobParams and
ShowInfoParams.
[no changelog]
- removes 2nd definition of ConfirmBlobParams, the choice of fields and
what supplied in ctor and what in `with_` methods should be thought
through again.
[no changelog]
Commit c300576d6c introduced
`confirm_blob_with_optional_pagination` which proved to be unpopular and
impractical. This commit brings back the old behaviour of having the
`ask_pagination` parameter on `confirm_blob`. It also reverts back to
using the old way of paginating `confirm_blob` on model R, which the
aforementioned commit ignored and re-implemented from scratch.
[no changelog]
request_anim_frame will register a timer for RequestPaint, which will
then cause a crash. This commit fixes the crash, but makes the marquee
component not work.
[no changelog]
Commit c300576d6c introduced the
`default_cancel` parameter to `show_warning` and `confirm_blob` rather
using the already existing `flow_warning_hi_prio` which was doing the
same thing. This commit reverts all the nonsense.
[no changelog]
