On 15/02/17 22:44 +0100, François Dumont wrote:
Hi
Here is the end result. I eventually chose to detect usage of
versioned namespace while generating the .gdb file. I haven't use any
caching mecanism considering the limited number of test cases.
Tested under Linux x86_64 with and without versioned namespace.
* python/Makefile.am (use_versioned_namespace): New.
(gdb.py): Subst use_versioned_namespace.
* python/Makefile.in: Regenerate.
* python/hook.in: Adapt.
* python/libstdcxx/v6/printers.py (Printer.add_version): Add name
versioned namespace if _versioned_namespace is defined.
(Printer.add_one_template_type_printer): Likewise
(add_one_type_printer): Likewise.
(register_libstdcxx_printers): Add parameter to indicate if versioned
namespace is active.
* testsuite/lib/gdb-test.exp (get_use_versioned_namespace): New.
(gdb-test): Use latter.
* testsuite/libstdc++-prettyprinters/48362.cc: Prefer note-test to
regexp-test.
Ok to commit to trunk ?
No, I thought the point of the use_versioned_namespace parameter was
to optimise thigns later, if needed. Do we need to do it?
We certainly don't need to do it during stage 4, as it's not actually
fixing any bug.
Let's reconsider during stage 1.
