Re: [fis-gtm] builds with pbuilder
On 02/07/2013 12:21 PM, Yaroslav Halchenko wrote:
On Thu, 07 Feb 2013, Amul Shah wrote:
Andreas/Yaroslav,
Thanks to Luis, I setup a pbuilder environment and built GT.M. There
are a few lintian warnings and some package naming oddities that I
am unsure about. How do I grab the build log, aside from using tee?
spoiled me uses git-buildpackage (could be called with --pbuilder
option) and that generates a nice .log for me
I say package naming oddity, because the generated deb is named
fis-gtm-6.0-001_6.0-001-1_amd64.deb, where I expected to see a deb
named fis-gtm_6.0-001-1_amd64.deb.
per our discussion at kitware some time ago -- we agreed to have
versioned binary package (i.e. version in the name) to signal that per
se you can't just upgrade fis-gtm to a new major.minor version to still
access previous DB -- it needs to be migrated. And that is why it is
better to be able to co-install 2 (or more) versions at the same
time. I do not remember though having -001 revision in there, and
would have expected fis-gtm-6.0_6.0-001-1_amd64.deb, but I could be
wrong.
[amul:2] That -001 is the sub minor version. Our next release is going to be V6.0-002. I think we need that in there to allow
V6.0-001 and V6.0-002 to coexist. Correct? We don't need one of those version numbers, that's for sure. Where are they coming
from? I know one is in debian/control.
I can't find the lintian warnings in pbuidler's output, but I did see them in debuild's output. This is what I see in debuild:
W: fis-gtm source: changelog-should-mention-nmu
W: fis-gtm source: source-nmu-has-incorrect-version-number 6.0-001-1
are you listed in Maintainers/Uploaders as well (with identical name in
the last changelog entry signature)?
[amul:2] I'm listed as an uploader (or at least I think I am). I don't have a GPG key yet (well I did, but I forgot the
password). My name is in both files, but the email address case is different.
> grep -i fisglobal debian/*
debian/changelog: -- Amul Shah <amul.shah@fisglobal.com> Fri, 25 Jan 2013 23:13:11 -0500
debian/control: Amul Shah <Amul.Shah@fisglobal.com>
W: fis-gtm-6.0-001: setuid-binary usr/lib/fis-gtm/V6.0-001_x86_64/gtmsecshr 4755 root/root
W: fis-gtm-6.0-001: non-standard-dir-perm usr/lib/fis-gtm/V6.0-001_x86_64/gtmsecshrdir/ 0500 != 0755
W: fis-gtm-6.0-001: setuid-binary usr/lib/fis-gtm/V6.0-001_x86_64/gtmsecshrdir/gtmsecshr 4500 root/root
W: fis-gtm-6.0-001: executable-is-not-world-readable usr/lib/fis-gtm/V6.0-001_x86_64/gtmsecshrdir/gtmsecshr 4500
I think we had discussion on those "security" measures -- would need to
look in emails to rehears what was our conclusion ;)
[amul:2] Sure. I'll dredge them up.
I'm not sure what nmu is.
http://wiki.debian.org/NonMaintainerUpload
[amul:2] Ah, I guess I'm not an uploader.
The flagged permissions for gtmsecshr are
what we require and check for. Do I need to suppress those warnings?
probably
[amul:2] Ok, I'll take care of them.
These are the warnings that I see in pbuilder's output:
dpkg-shlibdeps: warning: debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/plugin/libgtmcrypt.so
contains an unresolvable reference to symbol gtm_free: it's probably
a plugin
dpkg-shlibdeps: warning: 6 other similar warnings have been skipped (use -v to see them all)
dpkg-shlibdeps: warning: package could avoid a useless dependency if
and it is a plugin, so might rely on the main process to have the
namespace loaded for it.,.. so should be safe to ignore (not sure if
there is a way to suppress)
[amul:2] Ignoring works for me. :)
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/ftok
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/gtcm_server
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/lke
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/mumps
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/mupip debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/gtcm_gnp_server
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/gtcm_pkdisp
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/dse
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/libgtmshr.so
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/gtcm_shmclean debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/gtmsecshrdir/gtmsecshr
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/gtcm_play
were not linked against libncurses.so.5 (they use none of the
library's symbols)
dpkg-shlibdeps: warning: package could avoid a useless dependency if debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/plugin/libgtmcrypt.so
was not linked against libgpg-error.so.0 (it uses none of the
library's symbols)
gtm_free is provided by
debian/fis-gtm-6.0-001/usr/lib/fis-gtm/V6.0-001_x86_64/libgtmshr.so.
if those statements are correct -- you might like to adjust your CMake*
files ... but that is not critical really
[amul:2] That's the plan. I won't change the current release unless it is needed.
Do I need to suppress this warning? I will look into whether or not
we can avoid the dependency for libncurses and ligpg-error.
probably it is not that you need to avoid dependency -- it is just that
you are linking against them where needed and not. You might like (not
sure if tollerable ;) ) use
-DCMAKE_SHARED_LINKER_FLAGS="-Wl,--as-needed"
-DCMAKE_EXE_LINKER_FLAGS="-Wl,--as-needed"
?
[amul:2] Good suggestions. I'll give those a try.
What are the next steps?
let's decide on versioning and above NMU false-positives. And I guess
Andreas' blessing ;)
[amul:2] Agreed!
thanks, Amul
_____________
The information contained in this message is proprietary and/or confidential. If you are not the intended recipient, please: (i) delete the message and all copies; (ii) do not disclose, distribute or use the message in any manner; and (iii) notify the sender immediately. In addition, please be aware that any message addressed to our domain is subject to archiving and review by persons other than the intended recipient. Thank you.
Reply to: