On Sat, 30 Dec 2017 04:28:21 +0100 Axel Beckert <a...@debian.org> wrote: > Control: clone -1 -2 > Control: retitle -2 approx: FTBFS with ocaml 4.05.0: Some fatal warnings were > triggered > Control: found -2 5.8-1 > Control: fixed -2 5.9-1 > Control: close -2 > > Hi Ole and Eric, > > Ole Streicher wrote: > > @Adrian: Are you sure this is a regression introduced with 5.9-1? > > I fear so: > > ~ → links2 -dump > https://tests.reproducible-builds.org/debian/history/approx.html | fgrep > FTBFS | fgrep -v 5.9-1 > 2016-10-26 5.5-2 unstable amd64 FTBFS FTBFS 4m 9s > profitbricks-build5-amd64 profitbricks-build1-amd64 amd64_4/54728 > 2015-12-01 5.5-2 unstable armhf FTBFS FTBFS 5m 55s > armhf_14/164 > > So all approx FTBFS reported by the reproducible builds project were > from either 5.9-1 or 5.2-2 (more than a year ago) while 5.8-1 and > 5.7-1 never FTBFS, at least not in testing. > > There's a small chance that 5.8-1 would FTBFS with Ocaml from > Unstable. Since I couldn't reproduce the FTBFS with approx and ocaml > from unstable on amd64, I didn't try to build approx from testing with > ocaml from unstable on amd64. > > But I was able to reproduce the FTBFS of approx 5.9-1 on unstable > armhf six times in a row (with several variants: with and w/o > eatmydata, with different nice and concurrency levels). > > So I tried to build approx 5.8-1 with ocaml from unstable on armhf, > too. And it indeed FTBFS with ocaml from unstable. But not due to a > failing test suite but due to fatal compiler warnings: > > ocamlfind ocamlc -c -warn-error A -package nethttpd -package pcre -o > approx.cmo approx.ml > + ocamlfind ocamlc -c -warn-error A -package nethttpd -package pcre -o > approx.cmo approx.ml > File "approx.ml", line 283, characters 10-26: > Warning 3: deprecated: String.lowercase > Use String.lowercase_ascii instead. > File "approx.ml", line 1: > Error: Some fatal warnings were triggered (1 occurrences) > Command exited with code 2. > Makefile:15: recipe for target 'approx' failed > make[1]: *** [approx] Error 10 > > > 1. lower the severity to let it migrate, or > > Sounds like the wrong option to me. > > > 2. set the version-found to 5.8-1 > > Sounds wrong to me, too, given my findings. > > > to enable autoremoval of the package > > in testing (and then let the transition go). > > This nevertheless sounds like the correct approach. > > So IMHO the clean solution is a separate bug report covering the FTBFS > of 5.8-1 with ocaml from unstable, which is fixed with 5.9-1. I've
We still have test failures with 5.9-1 and ocaml 4.05.0, as can be seen in https://buildd.debian.org/status/package.php?p=approx E.g. from the i386 log: Error: 12:file_sha256sum_tests:1:(file_sha256sum "non-empty"). File "/<<PKGBUILDDIR>>/_build/oUnit-anon-x86-csail-01#01.log", line 273, characters 1-1: Error: 12:file_sha256sum_tests:1:(file_sha256sum "non-empty") (in the log). Raised at file "src/oUnitAssert.ml", line 45, characters 2-27 Called from file "src/oUnitRunner.ml", line 46, characters 13-26 expected: "0d43abb19c4f6fa228c0e577568a99cc6b3768d3ca0f0700e75377d0e08e8793" but got: "62457ada7c8ccf7dccee0d1e490dfe91d7eef6e29474643f9f7053b45d85c871" This is blocking the ocaml transition, so I may remove approx from testing to unblock the transition. Cheers, Emilio