+++ David Bremner [2014-08-07 12:47 -0300]: > Wookey <woo...@wookware.org> writes: > > > Package: notmuch > > Version: 0.18.1-1 > > Followup-For: Bug #756085 > > > > > > I applied the attached patch to build against current emacs and the > > build works fine on amd64, but fails one test (consistently) (on arm64). > > > > T380-atomicity: Testing atomicity > > PASS "notmuch new" is idempotent under arbitrary aborts > > FAIL detected 0>10 abort points > > test 0 -gt 10 > > > > Have you any idea what might be causing that? > > > > Yeah, it's probably some difficulty with gdb. As a hack, you could > remove the build-dep on gdb (on arm64) and the test will be disabled. > You can see the list of troublesome architecture in debian/control; some > of those may actually work now after a change to the test between 0.18 > and 0.18.1. If you want to debug further, you can try stepping through > test/atomicity.gdb.
OK setting up the env and running the gdb command we get: gdb -batch -x /home/buildd/packages/modified/notmuch-0.18.1/test/atomicity.gdb notmuch Breakpoint 1 at 0x4092d4 Failed to read a valid object file image from memory. [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/aarch64-linux-gnu/libthread_db.so.1". /home/buildd/packages/modified/notmuch-0.18.1/test/atomicity.gdb:54: Error in sourced command file: Unexpected error setting hardware debug registers And I get the same stepping through manually, so that does look like gdb just not working right. I have a gdb expert in my office so I'll get him to take a look at what's up. In the meatime turning that off for arm64 seems like the right thing to do, as there's nothing wrong with notmuch. > > Is there a way to get a more verbose test log? > > > > there is some more output in tmp.T380-atomicity after the test fails > > I plan to do an NMU with the patch below, so that at least this > > package is updated (unless you say no). Then we can worry about the > > arm64 test issue separately and there will be public build-logs I can > > point to (It will also stop the buildd trying to build this package > > over and over and over again (116 attempts so far!): > > http://buildd.debian-ports.org/status/package.php?p=notmuch&suite=sid > > Out of curiousity, isn't this a bug in the buildd? Well, Not really, although it may not be working quite right. It's designed to treat problems with installing build-deps as temporary. This actually makes sense - it's usually network errors, or an issue in one of the packages installed (md5sum mismatch or something else making install fail). Generally nothing wrong with the package that's trying to build (so long as its build-deps are satisfiable) So in this case the problem is that the build-deps say emacs23 | emacs24 but emacs23 doesn't exist for this arch (and probably isn;t going to as it's going away anyway and isn't simple to fix) but the buildds only take the first option. I'm not quite sure why the wanna-build analysis thinks that everything it need _is_ available (I guess it's incorrectly allowing for the the emacs24 alternative?) (and that does seem like a bug). > Why doesn't the same > problem occur on all the other BD-Uninstallable architectures? If the package is BD-Uninstallable then the system doesn't try to build it. > > Would you prefer a patch where we just drop the emacs23 part or does > > that remain useful for backport builds? > > I was waiting for the emacs24 backport before moving forward with this, > I think your patch looks sensible; there are still developers hand > building debian packages with emacs23. If possible I'd prefer a git > patch against the release branch of > > git://git.notmuchmail.org/git/notmuch Hmm. git. OK. Let's see if I can make this work. <fx: pokes git gingerly> OK. here's hoping this is what you meant. attached Wookey -- Principal hats: Linaro, Emdebian, Wookware, Balloonboard, ARM http://wookware.org/
diff --git a/debian/changelog b/debian/changelog index f873b1b..943f620 100644 --- a/debian/changelog +++ b/debian/changelog @@ -1,3 +1,10 @@ +notmuch (0.18.1-2) unstable; urgency=medium + + * Update build-deps to use current emacs24 on buildd (Closes:756085) + * Disable gdm atomicity test for arm64 as it doesn't work + + -- David Bremner <brem...@debian.org> Thu, 07 Aug 2014 09:34:40 +0000 + notmuch (0.18.1-1) unstable; urgency=medium * New upstream bug fix release diff --git a/debian/control b/debian/control index e2f2708..262e032 100644 --- a/debian/control +++ b/debian/control @@ -17,9 +17,9 @@ Build-Depends: python3-all (>= 3.1.2-7~), python-sphinx (>= 1.0), ruby, ruby-dev (>>1:1.9.3~), - emacs23-nox | emacs23 (>=23~) | emacs23-lucid (>=23~) | - emacs24-nox | emacs24 (>=24~) | emacs24-lucid (>=24~), - gdb [!s390x !ia64 !armhf !armel], + emacs24-nox | emacs24 (>=24~) | emacs24-lucid (>=24~) | + emacs23-nox | emacs23 (>=23~) | emacs23-lucid (>=23~), + gdb [!s390x !ia64 !armhf !armel !arm64], dtach (>= 0.8), bash-completion (>=1.9.0~) Standards-Version: 3.9.4
