Please make the manpages reproducible

[lamby]

Whilst working on the Reproducible Builds effort, we noticed that corosync could not be built reproducibly.

This is because, whilst it uses SOURCE_DATE_EPOCH, the output varies depending on the current timezone.

This was originally filed in Debian as #896441.

[knet-ci-bot]

Can one of the admins verify this patch?

[wferi]

Thanks, Chris, absolutely true, I omitted this from 14d251d by mistake. I vote for merging this fix.

Aside: this won't work on BSD, even the current code wouldn't: the -u and -d options are specific to GNU date. It isn't apparent in CI because Jenkins does not export SOURCE_DATE_EPOCH. An ad-hoc solution is given at https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal#Bash_.2F_POSIX_shell, a more elaborate one in kronosnet/kronosnet#143. I'm willing to put something together along those lines if the project lead agrees.

[lamby]

👍 :)

[fabbione]

Feri, corosync and knet do share a great deal of code and build facilities around (beside it´s the same people managing the 2 projects ;) ). There is a good 90% chance that a bug in one build system affects the other ;) I think it´s fine to port the knet patch to corosync,

[jfriesse]

@wferi: -u seems to be supported in Free/Net/OpenBSD (I don't think we really care about other implementations).

Because current code is using -d it's not compatible with BSD systems, so patch is not making things worse.

I have a question. Is there anything bad on An ad-hoc solution is given at https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal#Bash_.2F_POSIX_shell? It looks like quite simple compared to ax_prog_date.m4.

[lamby]

I have a question. Is there anything bad on An ad-hoc solution is given at https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal#Bash_.2F_POSIX_shell? It looks like quite simple compared to ax_prog_date.m4.

Can't parse this, sorry. Can you help? :)

[jfriesse]

@lamby Yep.

Is anything bad on solution described in the link https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal#Bash_.2F_POSIX_shell?

Or other way around. What advantages has using the ax_prog_date.m4 instead of solution described in the debian wiki link?

[wferi]

I don't think there's anything bad with it if you only use it once. If you want to use the date utility several times at different locations, then doing the checks once starts to make sense.

[jfriesse]

@wferi Ok, thank you for the reply. So I would say that because we have only one place where date is called we can stick with solution@debian wiki.

@lamby Would you mind to rework your patch in a way stated at https://wiki.debian.org/ReproducibleBuilds/TimestampsProposal#Bash_.2F_POSIX_shell ?

[lamby]

@jfriesse Done :)

[jfriesse]

@lamby: Nice patch so ACK and merged as 82b8199 (master) and 06ef44f (needle).

Thank you for the patch!

[lamby]

Thanks :)