Re: [OE-core] [PATCH v2] lib/oe/package: Add strip keep-section support

[Mathieu Othacehe <othacehe@gnu.org>]

> Thanks. But you still need to look at existing minidebuginfo tests and
> see if you can enhance them. Otherwise we have a variable that isn't
> tested.

OK, I will check want I can do.

> Also, if minidebuginfo distro feature is enabled, do we actually need
> this level of control? Why not just always have the needed bits?

On some architectures, such as x86_64 and aarch64, the .eh_frame section
that is *not* stripped because used for C++ exception handling, can be
used for DWARF-based stack unwinding purposes.

On those architectures, one can rely on .eh_frame + minidebuginfo to get
DWARF based backtraces. So no need for the PACKAGE_KEEP_SECTIONS there.

However, on ARMv7, the .eh_frame section is empty and the ARM specific .exidx
section is not enough to get full backtraces, as discussed here:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117941

For that specific architecture, it then makes sense to instruct 'strip'
to leave the .debug_frame section. By setting for instance:

PACKAGE_KEEP_SECTIONS:armv7a = ".debug_frame"

So minidebuginfo and PACKAGE_KEEP_SECTIONS are two very different
things, serving different purposes. Even though the combination of both
is required on some architectures to get full backtraces, I don't think
that they should be tied together.

> You can split into at least four commits:
>
> - change of runstrip signature
> - change of strip_execs signature
> - adding new variable to package.py
> - adding new variable to staging class.

OK, thanks for the suggestion.

Mathieu