Re: Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-04 Thread Andreas Tille

On Fri, 3 Apr 2009, Raphael Geissert wrote:


One additional thought: We store packaging Vcs and upstream homepage
in debian/control.  What about storing upstream Vcs there as well and
teach uscan to look there as well.  Just a rough thought.


And do what with it? how will it know what it should look for? that's why
all that information is in and belongs to the watch files.


Why not teaching uscan reading uscan watch *and* control file.  IMHO
the upstream VCS might be useful for other tools than uscan as well
and duplicating information is bad.

Kind regards

  Andreas.

--
http://fam-tille.de


--
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-04 Thread Raphael Geissert
Andreas Tille wrote:

 On Fri, 3 Apr 2009, Raphael Geissert wrote:
 
 One additional thought: We store packaging Vcs and upstream homepage
 in debian/control.  What about storing upstream Vcs there as well and
 teach uscan to look there as well.  Just a rough thought.

 And do what with it? how will it know what it should look for? that's why
 all that information is in and belongs to the watch files.
 
 Why not teaching uscan reading uscan watch *and* control file.

What's the benefit? what would uscan do with the control file? how will it
know the pattern it should use?

 IMHO 
 the upstream VCS might be useful for other tools than uscan as well
 and duplicating information is bad.

Uscan needs extra information to do something useful with a URI which, if
embedded in the control file, would confuse other tools or would require
other tools to massage the URI before using it (in that case, why not
require the other tools to obtain that information from the watch file?).
And as for duplication: sure, but currently the copyright file usually
mentioned where the source code was obtained from (usually upstream's home
page) and the control file usually contains the home page as well; that's
duplication (triplication in case it is the same base address the watch
file uses).


Cheers,
Raphael Geissert



-- 
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-03 Thread Daniel Leidert
Ben Finney wrote:
 Raphael Geissert atomo64+deb...@gmail.com writes:
 
  I planned to add support for svn in version 4 watch files (it would
  be a matter of svn info svn://domain.tld/path/to/repo and some data
  massaging).
 
  But well, now that everyone is talking about it why not just tell
  what is missing so that it can be addressed in version 4 watch
  files?

I would like to propose a complete rethinking of the watch files.
We have several packages, which must be prepared from a VCS or
must be repckaged. ATM several people changed to use something like

$url $pattern debian /bin/sh get-orig-source-script

One can't use

$url $pattern debian /usr/bin/make debian/rules get-orig-source

because make will error out because of the unknown switch
--upstream-version.

The next propblem is, that such a script doesn't have the same purpose
as uupdate has, as the first is used to create a new tarball and the
second to apply an existing diff to the new tarball.

So I was thinking about, how this could be solved. One idea, that came
to my mind is: Why not source debian/watch by uscan? One could then
write a watch file like this:

- first declare $url $pattern and maybe debian/uupdate like in the old
  watch file versions (probably this requires a new syntax, if
  debian/watch is sourced by the uscan script)

- then declare a function

sub prepare_orig_tarball () {
  ... the arguments should be upstream-version, debian-version,
  downloaded tarball name, orig-tarball name and ??? ...
  ... then add the commands to prepare/repack the tarball ...
}

uscan should then try, if this function exists, if not, simply download
the tarball/file specified by the pattern.

The function could simply contain a call like

/usr/bin/make debian/rules get-orig-source \
VERSION=$uvers PACKAGE=$orig_tarball

or call the already written get-orig-source script with the appropriate
arguments.

Opinions?

I'm pretty busy atm, but I would like to help fixing/improving uscan.

Regards, Daniel
-- 
Pt! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: 
http://www.gmx.net/de/go/multimessenger01


-- 
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-03 Thread Andreas Tille

On Fri, 3 Apr 2009, Daniel Leidert wrote:


I would like to propose a complete rethinking of the watch files.
We have several packages, which must be prepared from a VCS or
must be repckaged. ATM several people changed to use something like

$url $pattern debian /bin/sh get-orig-source-script

One can't use

$url $pattern debian /usr/bin/make debian/rules get-orig-source

because make will error out because of the unknown switch
--upstream-version.


I used to work around this in my packages by providing a

debian/get-orig-source

script which is just called in the get-orig-source target of
debian/rules.  This might be a way to circumvent this specific
problem (nothing against rethinking in general for sure).


The next propblem is, that such a script doesn't have the same purpose
as uupdate has, as the first is used to create a new tarball and the
second to apply an existing diff to the new tarball.


Exactly - so my suggested workaround just solves the problem of
the old way of thinking.


...
Opinions?


Interesting.

One additional thought: We store packaging Vcs and upstream homepage
in debian/control.  What about storing upstream Vcs there as well and
teach uscan to look there as well.  Just a rough thought.

Kind regards

  Andreas.

--
http://fam-tille.de


--
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-03 Thread Raphael Geissert
Andreas Tille wrote:
[...]
 
 One additional thought: We store packaging Vcs and upstream homepage
 in debian/control.  What about storing upstream Vcs there as well and
 teach uscan to look there as well.  Just a rough thought.
 

And do what with it? how will it know what it should look for? that's why
all that information is in and belongs to the watch files.

Cheers,
Raphael Geissert


-- 
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-03 Thread Raphael Geissert
Ben Finney wrote:
[...]
 
 For my purposes, I would like to be able to specify “fetch from the
 VCS at $URL, getting the specific working tree referenced by
 identifier $ID” and then the rest of what uscan does for generating
 an original source archive from the working tree.
 
 That's made more complex, of course, by:
 
 * the fact that the URL doesn't necessarily indicate what VCS tool is
   needed to interact with it (just about all of them allow an HTTP URL
   for public read-only access, so that's what will commonly be used
   here)

That can easily solved, svn+http://...

 
 * the plethora of different concepts for mapping identifiers to
   specific working trees in different VCSen (revision-id and branches
   and tags, oh my!)
 

uscan would retrieve the information (say the output of svn info
svn://domain.tld/repo/) and match the rest of the pattern against the
output. Whataver is specified is what would uscan would svn export.

 Ideally this would be predicated on the Final and Definitive Interface
 to All VCSen Now and Future™, but I think that might be delayed by at
 least a month or so.
 

s/month/decade?

Cheers,
Raphael Geissert



-- 
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-03 Thread Raphael Geissert
[No CC please]

Daniel Leidert wrote:
[...]
 
 I would like to propose a complete rethinking of the watch files.
 We have several packages, which must be prepared from a VCS or
 must be repckaged. ATM several people changed to use something like
 
 $url $pattern debian /bin/sh get-orig-source-script
 
 One can't use
 
 $url $pattern debian /usr/bin/make debian/rules get-orig-source
 
 because make will error out because of the unknown switch
 --upstream-version.
 

Sure, that's something that could be modified in version 4 watch files. I'm
thinking about a format string a-la printf.

 The next propblem is, that such a script doesn't have the same purpose
 as uupdate has, as the first is used to create a new tarball and the
 second to apply an existing diff to the new tarball.
 
 So I was thinking about, how this could be solved. One idea, that came
 to my mind is: Why not source debian/watch by uscan? One could then
 write a watch file like this:
 

I'm against sourcing anything, this is far beyond the pourpose of the watch
files and can be done by a specified helper script.

Cheers,
Raphael Geissert



-- 
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Improvements to ‘debian/watch’ for fetching from VCS (was: This topic died off; any resolution?)

2009-04-02 Thread Ben Finney
Raphael Geissert atomo64+deb...@gmail.com writes:

 I planned to add support for svn in version 4 watch files (it would
 be a matter of svn info svn://domain.tld/path/to/repo and some data
 massaging).
 
 But well, now that everyone is talking about it why not just tell
 what is missing so that it can be addressed in version 4 watch
 files?

For my purposes, I would like to be able to specify “fetch from the
VCS at $URL, getting the specific working tree referenced by
identifier $ID” and then the rest of what uscan does for generating
an original source archive from the working tree.

That's made more complex, of course, by:

* the fact that the URL doesn't necessarily indicate what VCS tool is
  needed to interact with it (just about all of them allow an HTTP URL
  for public read-only access, so that's what will commonly be used
  here)

* the plethora of different concepts for mapping identifiers to
  specific working trees in different VCSen (revision-id and branches
  and tags, oh my!)

Ideally this would be predicated on the Final and Definitive Interface
to All VCSen Now and Future™, but I think that might be delayed by at
least a month or so.

-- 
 \   “I don't know anything about music. In my line you don't have |
  `\ to.” —Elvis Aaron Presley (1935-1977) |
_o__)  |
Ben Finney


-- 
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org