This also modifies the previous implementation by always propagating the
`include_skip_proof_annotations` flag, whereas previously some of the
internal properties would "forget" about it, for example the one on
DeclarativePart.
However it seems necessary to propagate the flag to correctly handle
`is_subject_to_proof` on local declarative parts nested in a subprogram
annotated with Skip_Proof.
This commit introduces two functions (one for GNATCOLL.Projects, one for
GPR2) to build an analysis context object so that the unit provider,
preprocessing setup, configuration pragmas and the default charset are
automatically inferred from the project file.
Some tests mean to trigger a parsing error in Ada sources. For them,
disable the printing of parsing errors in the inline playground to avoid
polluting baselines.
Add support for "--% X = ..." to avoid repeating subexpressions in
several commands, or to break a big expressions into several parts. In
Take this opportunity to rewrite intensive use of indexing to browse the
parsing tree (which made it hard to review tests).
For GitLab issue #925
On Windows, the default encoding (not UTF-8) may not allow non-ASCII
test outputs. Use Python's "ascii" builtin where needed to ensure pure
ASCII test outputs.
TN: U720-016
Move most of "ada/*" to the root directory (this makes sense, as this
repository has been dedicated to Libadalang for years), and rename
"ada/language" to "ada".
TN: T914-010
