[Bug default/30034] [libabigail] Handle library splitting

["dodji at seketeli dot org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=30034

--- Comment #24 from dodji at seketeli dot org ---
"david.marchand at redhat dot com via Libabigail"
<libabigail@sourceware.org> a écrit:

[...]


> --- Comment #23 from David Marchand <david.marchand at redhat dot com> ---
> For the record, before I go off for some weeks.

[...]


> - You may want to fix typos (invalid -to suffix in the options) in the README
> file:

Fixed.

> - The mandatory aspect of the --add-binaries-toX options is disturbing, but it
> seems to work if I pass an empty list --add-binaries-to1=

Fixed.  The --add-binaries1= is no longer necessary if you have
specified a --add-binaries2=, for instance.

> - Requiring a '=' for those --add-binariesX options is surprising as other
> options won't accept a =.

Fixed.  The '=' is no longer necessary.

> - Now performance / runtime wise, before the DPDK split:
> * before the DPDK split, with 2.2 libabigail package from fc37:
> real    0m2,233s
> user    0m20,185s
> sys     0m2,354s
> * before the DPDK split, with libabigail binaries generated against this branch
> and no change in abidiff runtime:
> real    0m2,414s
> user    0m21,802s
> sys     0m3,502s
>
> And with the DPDK split:
> * using --follow-dependencies + --add-binaries-dirX:
> real    0m20,935s
> user    3m4,397s
> sys     1m45,828s
> * using --add-binaries1= --add-binaries2=librte_log.so + --add-binaries-dirX:
> real    0m2,703s
> user    0m24,006s
> sys     0m4,458s
>
> Which seems acceptable to me :-).

Good to hear.

I have updated the branch with the fixes.

Thank you for testing and providing this timely and valuable feedback.

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