Fixes 37fbafcd by adding more input to the hash than just second
grained time stamp - collisions would happen when running apk
scripted.
For virtual package the hash works only as unique identifier, so
try to add elements that should make it unique in most cases.
Fixes#10648
Add also some testing to make sure help, long help and handling
of invalid arguments works as expected.
Based on pull request #19 originally by Laurent Arnoud (@spk).
The return -1 seems to have been left over from earlier code, and
could have been treated as -EPERM. This helps to fix the other command
line handling that potentially require changing.
Notably this fixes occasional issues when doing upgrade with multiple
versions of same packages. Without this the upgrade flag is not always
propagated properly down the dependency chain.
Originally the virtual packages could have dependencies added to it.
However, commit b06e3b99 broke this behaviour to fix error reporting.
The root cause however was that the virtual depedency package was not
properly versioned.
This fixes to use current date/time as the package version, and
constructs the "faked" package hash from it. This effectively makes
"add -t virtpkg deps.." replace the dependencies which should be the
desired behaviour for "abuild deps".
'world' dependency to the generated virtual package is also now
versioned to make sure it get's upgraded.
BIT(1) corresponds with decimal 2, which is the first available repository.
Before this fix, `apk list -O` would list every package installed from the
first available repository, which is the 'system' repository on most Adélie
Linux computers.
After this fix, `apk list -O` correctly lists only the packages which are
no longer available.
merge_options() will write one more entry to the options table
which is the end-of-table indicator. Allocate memory for it too.
valgrind did not pick it up due to being in stack; changing alloca
to malloc would make valgrind notice the issue too.
Reported-by: Mobile Stream <info@mobile-stream.com>
A crafted .apk file could to trick apk writing unverified data to
an unexpected file during temporary file creation due to bugs in handling
long link target name and the way a regular file is extracted.
Several hardening steps are implemented to avoid this:
- the temporary file is now always first unlinked (apk thus reserved
all filenames .apk.* to be it's working files)
- the temporary file is after that created with O_EXCL to avoid races
- the temporary file is no longer directly the archive entry name
and thus directly controlled by potentially untrusted data
- long file names and link target names are now rejected
- hard link targets are now more rigorously checked
- various additional checks added for the extraction process to
error out early in case of malformed (or old legacy) file
Reported-by: Max Justicz <max@justi.cz>
Most applets return whatever apk_solver_commit() returns. It is the
number of errors found (or negative for hard error). Sanitize the
error value to not give false success exit code in the unlikely case
of errors % 256 == 0.
Reported-by: Max Justicz <max@justi.cz>
It's common for a pre-install script to do something like
addgroup -S group 2>/dev/null
When apk installs files after this, it sets the owner/group based on id cache
but currently the id cache is stale and doesn't contain the new group at that
point: instead the file will be installed with gid that the build host
happened to have for that group -- on target this might mean a non-existing
group or a completely different group.
We can't know if the script really did modify id cache contents so make sure
to reset the id cache on every script execution.
Virtual packages have the origin pointer set to NULL. Trying to print it
using the BLOB_PRINTF macros causes a segmentation fault.
Inspired by the `print_origin_name` function from `src/search.c` this
commit attempts to fix it by checking whether `pkg->origin` is NULL
before attempting to print it. If it is NULL the pkg name is printed
instead.
Since printing the pkg name requires a different format string this
commit splits the printf call for printing the package line into
multiple ones. The output format shouldn't have changed at all though.
Run apk_solver_solve once with all args as dependencies instead of
running apk_solver_solve for each arg. This is neccesary so the
install_if calculation is done correctly.
During netboot on systems without RTC, time() will be near zero,
and the index fill not exist. Thus the plain test of st.st_mtime
against system time failed. Verify that fstatat() succeeds.
The list applet provides a convenient way of inspecting both the available
and installed package databases by listing their contents. In some ways,
it is similar to `apk search` but is considered to be a superset of
`apk search` functionality.
A few `apk list` criterion are not yet ready though, such as `apk list --depends`
which searches by runtime dependency (replacing `apk info --rdepends`).
This modifies apk cache for indexes to be automatically refreshed
periodically without explicit 'update' or '--update-cache' usage.
The default is to do if-modified-since request if the local copy
is older than 4 hours. This age can be changed with --cache-max-age.
Using --update-cache will change this age to 60 seconds to make
sure the cached copy is relatively new. The small age is in order
to try to avoid downloading indexes second time when apk-tools is
upgraded and apk re-execs after self-upgrade.
Accordingly using explicitly 'apk update' will now enforce
--force-refresh and request the very latest index by requesting
any potential http proxy to do refresh too.
This unloads --force as several of the things are really not wanted
together. E.g. --force-refresh is a lot different from --force-broken-world
and doing --force to get the other might introduce unwanted behaviour.
--force is still kept for backwards compatibility and it enables
most things --force was used for.
This flag skips running hook scripts
This flag *must* be used during initramfs tmpfs initial install.
The reason that this new flag is needed is that the hooks will currently
always fail as musl and /bin/sh is missing at this stage on diskless.
This change just changes to keep deleted directory items in
the hash with ref count zero and modified flag set. Those entries
are reused when needed. The side effect is that fire_triggers()
will now see those removed direcotries and reports them. Other
enumerators of the directories hash are protected to skip removed
directories when appropriate.
In discovery phase, there was logic to not process packages
multiple times. However, that logic failed to account the package's
depth and install_if state for the name being processed. This
caused install_if processing failure in certain topologies of the
dependency graph. Adds also a test case that should catch this
issue reliably.
By introducing a new package metadata field, `provider_priority`
(index letter `k`), we can specify default packages to satisfy a
virtual.
If a user wishes to select an alternative provider for the virtual,
a changeset swapping the default provider for the selected provider
will be generated by the dependency resolver.
this fixes package selection when a 'real' package exists, but would
need to be provided by another package with 'virtual provides'.
In current package database this can happen with postgresql which is
also provided by postgresql-bdr. Normally postgresql would be satisfied
by postgresql, but if any package depends on postgresql-bdr and there's
no versioned dependency on postgresql this will help apk figure out
that postgresql-bdr should be used.
APKs have been created with GNU tar so far, which uses the
GNU extensions for long names. In order to increase portability
support the standard header's 'prefix' portion in case
the GNU extensions are not present.
Depending how the directory entries are ordered, the cached dir
instance might not have been updated correctly. This has not been
a problem as the entries have been ordered, but is now triggered
on ppc.
The value from tar header is unsigned int; keep it casted to
unsigned int and size_t instead of (signed) int, otherwise
the comparisons fail to do their job properly. Additionally check
entry.size against SSIZE_MAX so the rounding up later on is
guaranteed to not overflow.
Fixes CVE-2017-9669 and CVE-2017-9671.
Reported-by: Ariel Zelivansky from Twistlock
- rework the progress bar a little bit, basically removing the [ and ] to give a more
modern aesthetic.
- if utf-8 locale is enabled, use unicode codepoint 0x2588 instead of # to give the
progress bar a nicer look.
- if APK_PROGRESS_CHAR environment variable is defined, use the character defined there
for the progress bar.
This is useful for requirements such as: python3=~3.6, which would match python3-3.6.[0-9].
This implementation should in theory be backwards compatible with the implementation in Adelie.
Fixes regression from commit 5ba27c90 which caused stdio
buffering issues now that output is split to stderr/stdout.
See also commit 51737872 for some of the history.
Fixes#7107
On some (probably buggy) terminals, printing up to the rightmost column
may end up with the terminal issueing a newline (probably due to putting
the cursor at the next char -> new line).
Some other progress bars avoid it by not reaching the rightmost column.
Shorten the bar width from term-width - 7 to -8 (the 7 are "xxx% []").
fixes#5616
The original intention was not use unnecessary space on tmpfs
e.g. if the cache directory is a mount point, but accidentally
left unmounted. But there are valid cases when packages are
intentionally wanted to be cached on tmpfs. If caching is not
desired, the user can just remove the cache directory.
This allows for instance integration of etckeeper
[TT: Reorganized code a bit, and modified to use single
directory commit_hooks.d with argument for script of stage.]
Currently apk only knows about powerpc and ppc64. I am adding support for ppc64le.
ppc64le is the based on the ABI v2, which defines the endianess as little,
while ppc64 is based on the first 64-bits ABI.
It is unreasonable to assume that all package writers would except
to reset umask themselves. It's done currently in most packages,
but we had first issue of this kind recently, so better just reset
umask.
Make it possible to individually override openssl, zlib and libfetch
cflags and linker flags. This makes it possible to build apk-tools
without having pkg-config installed.
it's only used to speed up things, and having it non-existant
is not a fatal error - all included things in index will be
passed in command line anyway
which will delete any .apk package on output directory that were
not downloaded by fetch
this allows apk fetch to incrementally build repositories for
binary images
- self-conflicts when the exact same version of a name is provided
twice is now properly detected and diagnozed
- don't print redundant satisfies diagnostic
Implement --no-cache. The index is read directly from network and not
cached. This is useful for docker, where you install a set of packages
and directly after purge the cache. (see
1fc9e59d16/builder/scripts/apk-install)
fixes#4905
preserve [am]time for all packages and indexes. this fixes the caching
error that 'apk update' is after new index is generated, but before
the used mirror is synchronized. this caused local apkindex timestamp
to be newer than file in mirror, when in fact it was outdated index.
this also fixes fetched files to have build timestamp so that files
going to .iso or custom images have proper timestamps (rsync with
appropriate --modify-window now works)
the problem is that var/lock is on root installs symlink to /run/lock
(on tmpfs) and does not exist if doing chroot() to that root. fixes
apk to work when chrooted to existing rootfs install.