< hebasto>
just re-built 0.20.0 (point zero) with *clear* cache and got aae0b7e3a457a808335113f70448559d9d3b5414fa640e40f0ff7a31b390e2fa bitcoin-0.20.0-osx-unsigned.tar.gz
< hebasto>
it seems no gitian builder clears cache while building the previous 0.20.0 for macOS
< bitcoin-git>
[bitcoin] hebasto opened pull request #19547: doc: Update macOS cross compilation dependencies for Focal (master...200718-mac-dep) https://github.com/bitcoin/bitcoin/pull/19547
< achow101>
genisoimage is part of the build result because it's needed to make the code signed dmg
< achow101>
I wonder if it is a significant enough change to result in a different code-signed binary
< hebasto>
I wonder why genisoimage build is not deterministic
< wumpus>
achow101: I guess it would be possible to build genisoimage independently for the normal osx build and the signing, don't know how long it takes to build but can't be that bad?
< wumpus>
I agree it should ideally not matter, but it seeems this binary that is built for the host is the source of non-determinism every time, is it built with debug info perhaps? the chance of non-determinism in debug info seems to be slightly larger than in the stripped code
< achow101>
alternatively #19068 just removes it entirely, along with all of the other dmg building stuff
< wumpus>
tthe difference doesn't result in a different dmg, btw
< achow101>
interesting
< wumpus>
but that doesn't say much, if you compile genisoimage with gcc 4 and gcc 5 you'll likely also not get a different result - the logic wlil be the same
< achow101>
I think it makes sense to build that as needed, it'll just add to the build time of the signing step
< achow101>
or maybe we can make another inputs package that isn't part of the gitian signed result
< wumpus>
I know about #19068 but I don't think there is entirely agreement to stop with making dmgs in favor of zips