[Pkg-xen-devel] [PATCH 3/9] d/shuffle-binaries: Fix binary shuffling script for cross-building

[Elliott Mitchell]

On Mon, Nov 30, 2020 at 11:31:00PM +0100, Hans van Kranenburg wrote:
> On 7/17/20 8:37 AM, Elliott Mitchell wrote:
> 
> > so use `file` to detect scripts
> > and `strings | grep` for identifying linked libraries.  Even though
> > debhelper depends on file, make the requirement explicit.
> 
> Why would we stop to depend on debhelper? I don't mind to have another
> line with redundancy, but I'm just wondering what your thought behind
> this would be.

As Diederik de Haas noted, the concern is the dependency is indirect.  If
debhelper was to stop depending on file (unlikely at this time, but not
impossible in the future), then the Xen build mysteriously breaks without
any visible change.

> > Heavily simplify script inner loop.  While the core of that inner loop
> > was often executed only once, it is best to shrink inner loops.
> > 
> > With the reduced inner loop, less is being done inside pipes, so pipefail
> > is lower value (the early commands in pipes are unlikely to fail).  This
> > makes the script POSIX compliant so switch to /bin/sh.
> > 
> > Remove the useless extra argument passed to the script.
> 
> And that's 5? 6? 10? changes at once, and now I have to go dig into the
> diff and find out which change belongs to which line in the comments.
> 
> You already went through all of the thought work doing all of this. And
> that's much appreciated. Now make it explicit.
> 
> So, if you can split this out in a series of commits that just does one
> logical change with one simple explanation, and even if there's just a
> simple commit in the end that changes /bin/bash to /bin/sh with an
> explanation of "yay, now we got rid of all the stuff that prevented us
> from switching to /bin/sh for performance yay", no problem.

I felt it was a single change, "major rework of binary shuffle script".
I suppose some portions can be broken into separate commits.  Some will
be heavily overlapping and highly ordered though.  Issue is it will be a
bunch of tiny commits which cannot stand on their own, plus one big
atomic commit which cannot be split.  I suspect several will get squashed
back together on commit.

This will take a bit...


-- 
(\___(\___(\______          --=> 8-) EHM <=--          ______/)___/)___/)
 \BS (    |         ehem+sigmsg at m5p.com  PGP 87145445         |    )   /
  \_CS\   |  _____  -O #include <stddisclaimer.h> O-   _____  |   /  _/
8A19\___\_|_/58D2 7E3D DDF4 7BA6 <-PGP-> 41D1 B375 37D0 8714\_|_/___/5445