[Reproducible-builds] Repro Build report I do not understand

Dirk Eddelbuettel edd at debian.org
Sun Nov 8 21:17:14 UTC 2015


On 7 November 2015 at 00:11, Jérémy Bobbio wrote:
| Dirk Eddelbuettel:
| > | From the quick look I had, it seems symbols are sorted in a different
| > | order depending on the locale:
| > | https://reproducible.debian.net/dbd/unstable/amd64/littler_0.3.0-2.debbindiff.html#r-cran-littler_0.3.0-2_amd64.deb/data.tar.xz/data.tar/./usr/bin/r/objdump%20--disassemble%20--full-contents%20{}
| > | (It's my guess because both are sorted by the second build has lowercase
| > | 'a' grouped together with uppercase 'A'.)
| > | 
| > | The symbols match the ones in
| > | https://sources.debian.net/src/r-base/3.2.2-1/src/library/datasets/data/
| > | 
| > | Hope that helps,
| > 
| > That is very good too.  Should I ensure a locale during the build?  Any other
| > heavy hand?
| 
| You can try to set LC_ALL (or set LC_COLLATE and unset LC_ALL). But you
| might want to identify where the sorting happens first.

It is something else.  Both littler (r-cran-littler) and RInside
(r-cran-inside) embed R, and both are irreproducible.  I had overlooked how
RInside had failed this same issue and focused on the fact that I added
compilation-time timestamps (which I changed).

There is something else going on having to do with symbols from R. I did five
builds in quick succession, saving  the deb file under different names. They
all ended up with slightly different file sizes (!!) which is REALLY weird.

Not sure where to go from here.

Dirk (who knows the R build system pretty well)

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org



More information about the Reproducible-builds mailing list