[Bug default/29175] handling of symbols outside the first corpus in a corpus group

["gprocida at google dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29175

gprocida at google dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|abidiff ignores symbols     |handling of symbols outside
                   |outside the first corpus of |the first corpus in a
                   |a corpus group              |corpus group

--- Comment #1 from gprocida at google dot com ---
There are a few things to consider in abidw and abidiff handling of corpus
groups.

Are symbols in later corpus groups picked up by abidiff? No (this is the
original bug report).

Can symbols be duplicated between different corpora? Yes. Someone is building
lots of modules in one tree (but devices only ship / use a subset of modules
appropriate to the specific hardware). So only certain subsets of the corpora
are meaningful.

Does abidw do the right thing in terms of scoping elf-symbol-id to within each
corpus rather than across the whole DWARF reader / corpus group? Possibly not.

What if symbol x is "int x" in one corpus but "int x(double)" in another?

What will happen if parts of one type (say different member functions) are
spread across multiple corpora? Please don't do this.

I think it would be most useful if we could invent some test corpus groups,
without actually building kernels and kernel modules. I'll talk to Matthias
about this.

It would then be much easier to test and validate abidw behaviour.

-- 
You are receiving this mail because:
You are on the CC list for the bug.