[petsc-users] reproducibility

[Jed Brown]

Matthew Knepley <knepley at gmail.com> writes:

> On Fri, May 28, 2021 at 12:05 PM Mark Adams <mfadams at lbl.gov> wrote:
>
>> On Fri, May 28, 2021 at 11:53 AM Lawrence Mitchell <wence at gmx.li> wrote:
>>
>>>
>>>
>>> > On 28 May 2021, at 16:51, Mark Adams <mfadams at lbl.gov> wrote:
>>> >
>>> > It sounds like I should get one branch settled, use it, and keep that
>>> branch in the repo, and to be safe not touch it, and that should work for
>>> at least a few months. I just want it to work if the reviewer tests it :)
>>> >
>>>
>>> Just bake a docker container and archive it on zenodo.org or figshare :)
>>>
>>
>> Are you serious?
>>
>
> Yes. That is what Zenodo is for.

Yeah, an alternative is to archive that commit of the PETSc repository if you don't want to pin the compiler toolchain, etc. (There are pros and cons of pinning the whole environment versus enabling tinkering. One option is to have the repository snapshot and a Dockerfile that specifies versions, which would allow you to test and them to spin up something similar. But usually that build involves an apt-get update so you'll get new maintenance releases of all the packages. Usually harmless, but something to keep in mind.)