[PATCH] build: Make gzip archives reproducible

[Chris Johns]

On 14/8/2023 4:24 pm, Sebastian Huber wrote:
> On 12.08.23 18:29, Joel Sherrill wrote:
>> Piling on Chris' comments about where is the requirement for this. This is a
>> topic which must be addressed in the Software Engineering Guide and accepted
>> before it can be put into practices. A software development process of any
>> quality cannot be changed on a whim without discussion and modifying all the
>> artifacts necessary.
> 
> Sorry, it was not my intention to open another can of worms. This reproducible
> builds stuff appeared to be some small thing to do.

I think it would be good to understand what you intend and why? At the moment we
are wondering or guessing based on some patches and that is never good. Our main
focus is the long term liability change have for us. Anything you can add to
explain things helps.

I think reproducible builds is a great thing to attempt but in the past there
have been issues. For example at one point the register allocator in gcc for the
m68k used the libc sort and Windows and Linux differed when sorting "all the
elements are the same" so the code built on Windows did not match the code built
on Linux even though it was equivalent code, ie register ordering for push and
pops differed.

Chris