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

Re: Automatic Debug Packages



Stefano Zacchiroli wrote:
> On Wed, Jul 29, 2009 at 02:18:49PM +0200, Emilio Pozuelo Monfort wrote:
>> I've written down the details in the wiki [2], and I'll appreciate
>> it if you could give some feeback. I don't want to trash this
>> completely though, so no drastic changes preferred :)
> 
> I wonder how C-specific is your proposal. In particular, I wonder if
> other compilers / interpreters which might provide debugging symbols
> can benefit from it. My specific case is OCaml, where we can build
> libraries with debugging information even though we usually don't do
> that. While I guess the archive part of the infrastructure (.ddeb
> handlings) is pretty independent from .ddeb content, I'm not sure
> about the helpers, but we can of course provide our own extra
> helpers. Can you comment on that?

There should be no problem in which content you ship on a ddeb (using common
sense, of course), so they could be used for other languages that don't use
build ids. However the "one ddeb per source package" requirement may expose
problems to some packages if two binary packages built from the same source
package ship the same files built differently. E.g. we could imagine pygtk being
built twice with different compiler flags, creating python-gtk2 and
python-gtk2-optimized binary packages. Both would ship
/usr/lib/pyshared/python2.5/gtk-2.0/pango.so, and their python-dbg modules would
be /usr/lib/pyshared/python2.5/gtk-2.0/pango_d.so. It wouldn't be possible to
ship both _d.so in the .ddeb

This should be the exception, but is still a limitation, so maybe we should
consider one ddeb per binary package (the reasons for one ddeb per source
package were mainly related to build-id, but I could implement it with one ddeb
per binary package too). Note that the changes file may become be somewhat
bloated if we allow one ddeb per binary package ;)

Do you think that limitation is enough to change from one ddeb per source
package? Should we maybe allow any number of ddebs, and only build one ddeb in
the common case? Or allow both one ddeb per source package and one per binary
package?

Cheers,
Emilio

Attachment: signature.asc
Description: OpenPGP digital signature


Reply to: