Stéphane Glondu: > [..] > > There are very few occurrences of dup3 and pipe2 in OCaml sources, and > looking at them, upstream indeed "tweaked some things to use dup3/pipe2 > instead". But the codepath is conditioned to the presence of HAS_DUP3 > and HAS_PIPE2. One has to make these variables disappear on kfreebsd. >
Thanks for digging into this a bit more, I had ran out of patience about that... anyway I have fixed it properly now and pushed it to experimental. Sadly we now have one new failure, on arm64: https://buildd.debian.org/status/fetch.php?pkg=ocaml&arch=arm64&ver=4.05.0-5&stamp=1500739972&raw=0 E: Build killed with signal TERM after 150 minutes of inactivity Even though this was built in experimental, it was built with the binutils from sid: [..] Get:21 http://mirror.netcologne.de/debian unstable/main arm64 binutils-dev arm64 2.28-6 [2182 kB] [..] Selecting previously unselected package binutils-dev. Preparing to unpack .../19-binutils-dev_2.28-6_arm64.deb ... Unpacking binutils-dev (2.28-6) ... [..] This did not happen with the previous builds, and I didn't change anything relating to arm64. However after reading https://caml.inria.fr/mantis/view.php?id=7585 I wonder if this is just the same issue? Apparently ocaml will generated aarch64 instructions that on some corner cases (which I don't understand) are incorrect and the new binutils simply rejects these? Perhaps with the old binutils, these same faulty corner cases will sometimes result in these test suite timeouts? X -- GPG: ed25519/56034877E1F87C35 GPG: rsa4096/1318EFAC5FBBDBCE https://github.com/infinity0/pubkeys.git
