This patch removes the old implementation of generate-packages-sources
that used apt-ftparchive instead of the metadata in the Postgres
database.

We don't want cron.unchecked to do anything while we run process-policy.
Especially not running another process-policy process or moving files
out of queue/accepted/ while we are installing files there.

Signed-off-by: NMark Hymers <mhy@debian.org>

This avoids triggering the taint check in spawn().

newstage is gone. Instead we need to run process-policy and grab
accepted files from $queuedir/accepted.

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Renamed functions and methods:

  Find              -> find
  FindB             -> find_b
  newConfiguration  -> Configuration
  ParseCommandLine  -> parse_commandline
  ParseSrcDepends   -> parse_src_depends
  ReadConfigFileISC -> read_config_file_isc
  SubTree           -> subtree
  ValueList         -> value_list
  VersionCompare    -> version_compare

Removed functions:

  debExtractControl: replaced with utils.deb_extract_control

Other changes:

  x.has_key(y) -> y in x
  x.Find(y)    -> x[y]

The last change was only done in places where `y' is sure to exists (`x.Find'
would return None otherwise, `x[y]' would raise an exception).

helps avoiding accidental *changes commandline arguments
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

We don't want process-upload and generate-packages-sources2 or similar
combinations running at the same time. So just create a lock around such
sections.

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

merge bugfixes from Torsten and Mark:
- built uncompressed and bz2 packages/sources too for buildd
- set a codename in the release file
- strip epoch from version
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NJoerg Jaspert <joerg@debian.org>

entirely rewritten. much smaller and nicer and better and whatnot.
except preparing advisories, this should be all they need. lets see.
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

* replaces functions write_filelist(s) in msfl
Signed-off-by: NTorsten Werner <twerner@debian.org>

Signed-off-by: NChris Lamb <lamby@debian.org>

Signed-off-by: NChris Lamb <lamby@debian.org>

Signed-off-by: NChris Lamb <lamby@debian.org>

add a little function in daklib/utils.py to return the unix username, make nsi and transitions use this
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NMark Hymers <mhy@debian.org>

no longer check the daily lock when we start looking at a package, but
when we try to do modifying actions.
No reason to not allow people to do Check/Reject/Note, those are actions
we can do whenever. Its just modifying overrides which *have* to be
avoided at certain times.
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

its database.get_suite_architectures everywhere
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

new function, daklib.database.get_suite_architectures.
Use that instead of Cnf.ValueList("Suites::%s::Architectures") everywhere
to find out which suite supports which architectures
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

oh well, use a wrapper.
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

enable the package-file-map
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

doc strings, dont we love em?
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Rather than scatter regexes all over the place (for things
which should probably be functions hiding the implementation
detail anyways such as "is this a valid package version"),
move them into daklib.regexes.

I'm not convinced this is an immediate win, but I think it's
one step on a long road...

Oh, and it's untested
Signed-off-by: NMark Hymers <mhy@debian.org>

remove a few trailing ;
Signed-off-by: NJoerg Jaspert <joerg@debian.org>

Signed-off-by: NMark Hymers <mhy@debian.org>

The buildd team believes that the fact that the packages are in the security
pool after this point should be good enough.
Signed-off-by: NMark Hymers <mhy@debian.org>