Hi Tom,
On 25 February 2022 at 13:14, Tom Lee wrote: | Thanks for giving it a shot. Surprised to hear the patch didn't apply | cleanly, believe I worked directly from the source pulled via | | dget -a http://deb.debian.org/debian/pool/main/t/tiledb/tiledb_2.6.2-2.dsc Would it be possible for you to work off a git checkout? I still use the one started by Adam (which I took over) in the debian account: [email protected]:debian/tiledb.git | I even had a DD co-maintainer try the patch before I sent it your way so | not quite sure what I messed up. "Me neither". But sticking the patch into debian/patches/ lead to dpkg-* failing. The patch was also (very weirdly) non-unified. Why? | In terms of testing, tested by myself and the co-maintainer. I'm not sure | what to suggest here but I can try regenerating the patch, perhaps I should | try again using dpkg-source --commit. Yes please do work off git. Whether or not we use pbuilder (I do, "old" setup) or not should not matter. We could also proxy on "clean" environments like a Docker container or AWS instance. But I want to be sure this works. | Can't speak to the build failure, perhaps send through your own patch and | the full build log and I'll take a look. There is no "own" patch. I just tried to follow your (nice !!) logic of removing the enforced "exact" match (my colleagues at TileDB at strong proponents of pinnning as they are used to building in other environments <coughs, mutters conda, coughs some more> and your logic of recomputing the capnproto files (as you kindly explained in the other email). My pbuilder is current to unstable too. Dirk -- https://dirk.eddelbuettel.com | @eddelbuettel | [email protected]
