GitLab

And use it in couple of places. Some whitespace fixes too.

We might want to add an --update-index option to misc applets. For
example:

  apk add --update-index -u package
  apk version --update-index

We upgrade when we have an oldpkg.

implements 'apk add --virutal metaname dep1 dep2...' where metaname will
be an empy meta package with dep1 and dep2 as dependencies.

This is useful to prevent abuild to add each makedepend to world which
causes some headache when it comes to unintalling them after sucessful build.

The syntax is: apk index -d /path/to/APK_INDEX.gz pkg...

It does not seem like its possible to remove packages in the db so we
trick apk_db_index_write() by setting the repo to on-zero.

It's still not perfect since it does not recalculate the dependencies.

This allows you to query the state db even if you do not have write
permissions in cache dir (which is needed if you have remote repositories)

This should speed up things and save some memory if you have big and slow
remote repositories.

apk_db_add_repository does its own error reporting and does not set errno
reliable. This patch also tell which repository it failed to load the index
for rather than just say it failed to open "APK_INDEX.gz" which does not tell
much.

Cache non-local index files always locally.
Introduce 'update' applet to force refresh of cached index files.
Fixes #19.

Fixes #24.

Calculate changesets directly by stabilizating the package graph instead of
recalculating the whole graph and then diffing (similar approach as seen
in 'smart' package manager). The algorithm is not complete: defferred
search space forking is missing. So you don't always get a solution on
complex graphs.

Benefits:
- usually the search state tree is smaller (less memory used)
- speed relational to changeset size, not database size (usually faster)
- touch only packages related to users request (can work on partitially
  broken state; upgrades only necessary packages, fixes #7)

Also implemented:
- command prompt to confirm operation if packages are deleted or downgraded
- requesting deletion of package suggests removal of all packages depending
  on the package being removed (you'll get list of packages that also get
  removed if you want package X removed)
- option --simulate to see what would have been done (mainly for testing)
- an untested implementation of versioned dependencies and conflicts

A lot has changed, so expect new bugs too.

This will be used later by the commit change calculator (for improved
changeset calculation, ref #7). Will be also used by "apk info" to show
reverse dependencies or "required by" information.

So we don't get artificial limits on the amount of dependencies
(fixes #8).

So you will not get .apk-new files of the new configuration files.

we let bb use /bin and /sbin, all packages overriding busybox must
install the bins in /usr/bin and /usr/sbin

apk_hash_delete() actually deletes the file entry, so unlinking
from other lists need to happen before that. Also free the diri
instead of leaking it.

We dont care if a directory have references or not, we care if directory
is relly there.

Only the creator of the direcotry will set the permissions.

This fixes the situation where you install a package owning a dir then
you remove this dir manually and try install a second package also owning
the dir.

For example:

  apk add e2fsprogs-dev
  rm -rf /usr/lib/pkgconfig
  apk add zlib-dev

Otherwise creating rootfs from scratch with --initdb cannot be locked
due to the lock file path missing.

Protect files in protected dirs even if the file is not previously
registered in the database. We do so by always extracting to a
.apk-new file name, compare the checksums afterwards and rename if
file was identical.

Add flags field to db open call. Also make error reporting quite a bit
more detailed.

Otherwise installed db load fails due to internal reasons. This would
cause a lot of other funny stuff happen.

Ignore /etc/apk/repositories, so additional repositories that depend
on other repositories need to have explicit --repository reference on
command line when generating the index (to avoid warnings).

This was broke on --initdb addition to add applet since apk_db_create
changes the working directory.

this allows non-root users to query the database.

But we dont bother try create all subdirs (i.e mkdir -p ...)