Hello Niels, Thanks for your quick followup.
On Thu, Nov 29, 2018 at 07:31:00AM +0000, Niels Thykier wrote: > Hi Andreas, > > Thanks for filing the bug. Please feel free to wontfix/close this if you feel like it. [...] > This seems to still be a "whack-a-mole" strategy; we are just > centralizing it to debhelper (which does have some merit). There will for sure always be cases where we need to massage the package to make it reproducible on usrmerge vs non-merged. My thought was simply to cat a few cases before they even get introduced..... if possible (and worth it). [...] > Honestly, I am conflicted on this. On one hand, I do not want to > maintain such a list forever - which I would have if usrmerge will > remain optional forever. On the other hand, I can see how this would be > useful in deflating the current (and future) d-d mail threads about > usrmerge. > > I will be considering it for now. I've looked over the 32 (+ 1) cases and my unconfirmed view on them is that if debhelper passed the following variables to configure: PS alone fixes 2 GREP alone fixes 1,5 (positively effects 1 more) CAT alone fixes 1 (positively effects 1 more) TRUE alone fixes 1 GZIP alone fixes 1 (positively effects 1 more) SED alone fixes 1 (positively effects 3 more) ECHO alone fixes 1 MKTEMP alone fixes 1 (positively effects 1 more) above (MKTEMP SED) plus BASH fixes 1 more above (GREP SED) plus MKDIR_P fixes 1 more above (GZIP CAT) plus SH fixes 1 more Please note: All affected cases seems to either use autoconf or Makefile.PL build system. I have no idea about perl so only autoconf are part of above summary. Makefile.PL should be rare enough that it shouldn't affect things much. (The GREP -> 1,5 comes from that it fixes 2 but there's another comment in one of them that is still causes unreproducibility, but that is fixed by SED so.... could have said 2...). I'm not sure what conclusions to draw from this exactly. I'm also not sure what the reproducible build system staying on 32(+1) found for a while now means. Has it tested all packages in archive yet or not? If this is it, then I guess it's absolutely not worth bothering to do anything in debhelper to handle this.... Regards, Andreas Henriksson
