[Bug binutils/28058] 300x performance regression in strip

[mail at nh2 dot me]

https://sourceware.org/bugzilla/show_bug.cgi?id=28058

Niklas Hambüchen <mail at nh2 dot me> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |matz at suse dot de

--- Comment #1 from Niklas Hambüchen <mail at nh2 dot me> ---
Ah, master actually has a fix from 1 month ago that I hadn't spotted:

    commit 956ea65cd707707c0f725930214cbc781367a831
    Author: Michael Matz <matz@suse.de>
    Date:   Mon Jun 7 15:52:31 2021 +0200

        bfd/elf: Don't read non-existing secondary relocs

        Without this we unconditionally try to slurp in secondary
        relocs for each input section, leading to quadratic behaviour
        even for strip(1).  On write-out we already used a flag to avoid
        this.

        So track existence of secondary relocs on read-in as well and
        only slurp in when needed.  This still doesn't implement a proper
        list of secondary reloc sections, and still would exhibit quadratic
        behaviour if most input sections have a secondary reloc section.
        But at least on normal input this avoids any slowdown from trying
        to handle secondary relocation sections.

It would be great if a point release could be made that includes it.

Thank you!

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