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

Re: pdebuild-cross fixes in wook-shed branch

On Mon, 27 Sep 2010 15:00:15 +0100
Wookey <wookey@wookware.org> wrote:

> +++ Neil Williams [2010-09-25 13:34 +0100]:
> > On Tue, 14 Sep 2010 02:28:12 +0100
> > localised package with the changes in the branch. Migrate the current
> > PO files to check.
> I admit that I have not the fainest understanding of how all this
> works. Anything I touched was the blindest of blundering about and is
> no doubt completely wrong. I just fiddled till it built and the right
> files were in the right packages. 
> It would be best if I didn't touch any of that, although I'm very
> happy to have it all explained to me sometime.

The top level po directory is basically a Makefile copied out of an
autotools package. The work happens in LINGUAS, Makevars and
POTFILES.in and the rest is up to intltool-*.

I'll write some documentation in the po4a project at some point....

> > 1: use feature "switch"
> > 
> > I'd much rather use simple if() {} elsif () {} else {} layout. It's
> > easier to debug within perl. It's no slower. Unlike C, perl doesn't
> > take longer to compare strings than it does to compare identifiers.
> erm OK. You want me to change that? As a perl person who knows what
> they want you'll probably get what you want better if you do it
> yourself. I thought switch was actually part of perl in more recent
> versions so there is no real reason to avoid it?

... except that we'd be better using a config file approach as outlined
in the other message.
> > 3: force-yes - NO. Use the actual no-auth support in apt, not the
> > --force-yes sledgehammer. (See the multistrap code for code to turn
> > off SecureApt.) This is a complete deal breaker - I will not upload code
> > using --force-yes merely to disable SecureApt.
> I had to do it that way because apt >=0.8 didn't work with no-auth. It
> still doesn't, although the failure has changed from the original:
> https://bugs.launchpad.net/ubuntu/+source/apt/+bug/625042
> to
> https://bugs.launchpad.net/ubuntu/+source/apt/+bug/646499

(That such a disruptive change was made this close to a release
infuriates me. It's plain dumb.)
> Until we work out (or someone tells us) how to do authenticated
> downloads into a chroot we have to use --force-yes or neither
> multistrap nor pdebuild-cross actually work at all. I realise this is
> pretty shoddy but it seemed a lot better than nothing.

Shoddy doesn't even start to cover it. Is aptitude any better behaved?
> I've got rid of that hook in latest checkin as it just installs
> apt-cross and pile of pabout 15 perl packages in _eery single_ build,

Good. If xapt is usable as the default, drop apt-cross entirely.

> > 7: xapt could do with bash_completion support (pdebuild-cross too,
> > actually).
> True. I have no idea how to write such a thing.

There are examples in this package and the other emdebian-* ones -
basically convert the --help output into shell variables. I'll take a
> > 8: temporary files in SVN:
> > doc/pdebuild-cross/man/man1/pdebuild-cross-create.1
> I have failed to grok what is made from which in all that i18n stuff
> so I was conservative about binning things that _might_ have been
> important. As I say please either explain it all or just correct it. I
> am thoroughly confused.

OK, the manpage generation is po4a and po4a-build. It's a little bit
opaque when used with source packages that build multiple binaries.
What it tries to do is generate translated XML and then parse that XML
into translated manpages. I corrected the svn, I can try and explain
the rest - the main thing is that the "important" stuff is only the
English XML and the po/* contents. The rest is in the Makefile.

> > There is - use the normal apt support for disabling SecureApt, not
> > --force*. -o Apt::Get::AllowUnauthenticated=true
> I would if it worked.

That is really, really poor behaviour by apt. I'm very disappointed
that this has been broken so badly so close to a release. 


Neil Williams

Attachment: pgpNMAn72VMFB.pgp
Description: PGP signature

Reply to: