[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: RFS: hivex



On Sun, 2010-05-09 at 12:35 +0800, Paul Wise wrote:
> 
> Looks like some symbols dissappeared from libhivex0, which now causes
> the package to FTBFS because dpkg-gensymbols fails. I'm assuming they
> were the private symbols that are now hidden. I'd suggest removing
> them from the symbols file in that case.

Yes, done. I think I was asleep at the wheel because I totally neglected
to do any build tests for some reason - sorry for wasting your time on
that!

> Apart from that, the package looks fairly good. I'm unable to commit
> to sponsoring it though, hopefully you'll find someone else with the
> ongoing time to do so.

Thanks, I'll work on that.


> Remaining minor stuff:
> 
> lintian:
> 
> W: libwin-hivex-perl: maintainer-not-full-name TJ
> I: libwin-hivex-perl: unused-override libhivex-perl binary:
> maintainer-not-full-name

Aha! I forgot to change the specified library name when making the Perl
package-name adjustments.

> I: libwin-hivex-perl: spelling-error-in-manpage
> usr/share/man/man3/Win::Hivex::Regedit.3pm.gz reencode re-encode

Looks like this one escaped upstream when they patched. I'll chase that
up and add a manual patch for now.

> E: libwin-hivex-perl: binary-or-shlib-defines-rpath
> ./usr/lib/perl5/auto/Win/Hivex/Hivex.so
> /tmp/buildd/hivex-1.2.2/perl/../lib/.libs

Hmmm, another rpath issue. Don't recall seeing that in previous reports.
I'll ask upstream about it. Hope I don't have to re-implement the
chrpath workaround again.

As a matter of interest, would you mind sharing the lintian commands you
use since it is still something of a mystery to me. I use one command
against the source package but I'm guessing now you're running it
against each binary package too? In fact, it'd help me a great deal to
understand the entire process you go through in doing the review - if I
can do that before uploading it'd probably make for cleaner, better,
uploads and less iterations.

> dpkg-gencontrol: warning: package hivex: unused substitution variable
> ${perl:Depends}

I recall removing this from debian/control and it resulting in a bunch
of other messages so I put it back in. I'll remove it again and look
more closely at the build log this time.

> You may want to look at debhelper 7 instead of cdbs:
> 
> https://penta.debconf.org/dc9_schedule/events/418.en.html

I feel more comfortable with CDBS since I like simple-patchsys, although
now upstream has incorporated my patches it is no longer used. Reading
those slides didn't really help and I don't have time to listen to the
presentation but it mostly made my eyes glaze over :p

> You may want to look at DEP-5 for the copyright file:

> http://dep.debian.net/deps/dep5/

Crikey, another format to master :p

I've started on it but it looks to be slightly involved for hivex since
the current copyright is different for "Makefiles and programs" and
"libraries and language bindings" - figuring out the patterns for the
File specifications is going to be 'fun'.

> 
> debian/libhivex0.install can be reduced to one line: usr/lib/libhivex.so.*

Ahh yes, that list of libraries was a leftover from getting the split
between the -dev and library packages correct.

> 
> Is it nessecary to ship the static library? and the .la file?

Probably not, but which would be preferred in the -dev package? Looking
at other -dev packages there doesn't seem to be a clear preference.

> README probably isn't nessecary to ship in libhivex0, possibly the
> same for libhivex-ocaml and perl packages.

I think I was enjoying the easy life of not having to figure out/write
the rules to prevent the installation. I'm finding after a certain point
the need to find yet more documentation on how to fine-tune the package
becomes tiresome.

> It would be nice if the package descriptions were less template-like
> and describe the individual packages instead of the whole hivex
> project. Consider the audience for each package description when
> writing them. For example libhivex0 will almost always be
> auto-installed, so it can have a minimal description.

I see your point - changed.

I'll upload the changes in the next day or so. Thanks for the great
feedback.


Reply to: