This fixes a bug introduced by my previous changes which changes the
behavior of IgnorePkg/IgnoreGroup to allow the user to remove unresolvable
packages from the transaction. The bug is that the target-list was no
longer being consulted first to resolve dependencies, which means that if
two packages in the sync database satisfied a dependency, and the user
explicitly requested one of those two packages in the sync, the other
package was still being pulled in.
A new test was added, sync993.py, to verify the desired behavior.
Signed-off-by: Bryan Ischo <bji-keyword-pacman.3644cb@www.ischo.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
From now on _alpm_db_find_fileconflicts() works with upgrade and remove
target lists (like checkdeps), which makes it transaction independent
(we still need a trans param because of the progressbar). This is a small
step towards the universal transaction. So we call this function directly
from sync.c before commiting the remove transaction. This is much safer,
but we can get false fileconflict error alarms in some tricky cases
("symlinks puzzle" etc).
The patch on find_fileconflict looks complex, but it is mainly an
"indent-patch", the new code-part can be found after the
/* check remove list ... */ comment, and I modified something around the
"file has changed hand" case (see comment modifications in the code).
Unfortunately sync.c became more ugly, because we have to create 2 parallel
internal transactions: to avoid duplicated work, upgrade transaction is
used to load package data (filelists). This problem will disappear, when
we finally get rid of internal transactions.
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Signed-off-by: Dan McGee <dan@archlinux.org>
Don't prompt the user for unignore of IgnorePkg/IgnoreGroup packages,
except for packages explicitly listed for sync by the user. This
eliminates many unnecessary prompts when IgnorePkg/IgnoreGroup is
used.
Signed-off-by: Bryan Ischo <bryan@ischo.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Enabled a new prompt to ask the user if they'd like to remove
unresolvable packages from the transaction rather than failing it.
Many pactest tests that used to fail now return success codes, because
pacman now issues a prompt allowing the user to cancel rather than
failing many transactions, and the pactest scripts always choose to
cancel with no error rather than failing. The only net effect is that
the return status of pacman is now 0 in cases where it used to be
nonzero.
Signed-off-by: Bryan Ischo <bryan@ischo.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
The HoldPkg feature is even more important when the packages to be held are
pulled automatically by pacman, in a -Rc and -Rs operation. Before, it only
applied when the packages were explicitly requested by the user to be
removed. This patch extends holdpkg to -Rc and -Rs by doing the HoldPkg
check just before trans_commit.
Additionally, the whole HoldPkg stuff was moved to the front-end.
I changed the default behavior to "don't remove", so I modified remove030.py
pactest as well.
See also: FS#9173.
Original-work-by: Xavier Chantry <shiningxc@gmail.com>
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Signed-off-by: Dan McGee <dan@archlinux.org>
When we do our sed edits, we really don't need every command printed out to
the terminal. Now with "make -s", the output is quite palatable.
Signed-off-by: Dan McGee <dan@archlinux.org>
Do the checks in the tests that need it, and get rid of some of the
cluttered output when it is not available (one line per test run).
Signed-off-by: Dan McGee <dan@archlinux.org>
This was a stupid and unimportant regression caused by commit
4476598e4e .
When libdownload is not available, a xfercommand is needed for this pactest
to run properly.
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Added a Makefile.am for the pactest/tests dir. This is a blatant ripoff
of scripts/Makefile.am, which replaces predefined expressions in
NAME.py.in pactests with configure variables.
This can be used to write pactests which consider compile time options.
Signed-off-by: Henning Garus <henning.garus@gmail.com>
[Dan: autotools are tough, make a few adjustments for correctness]
Signed-off-by: Dan McGee <dan@archlinux.org>
Instead of pausing 1.5 seconds on tests that check file mtimes, change the
mtimes to something in the far past so we can immediately tell if a file was
modified and/or touched. This saves a decent amount of time on the upgrade
tests which often check mtimes.
355 was a completely arbitrary time value, don't ask me why I picked it.
Signed-off-by: Dan McGee <dan@archlinux.org>
This will allow the return code of pactest to be useful, for such things as
use in a git-bisect test script.
Signed-off-by: Dan McGee <dan@archlinux.org>
This will enable us to mark tests we know currently fail to differentiate
them from those that we know should pass. Regressions should be easier to
spot this way.
Signed-off-by: Dan McGee <dan@archlinux.org>
This vercmp issue has been a sticking point but this should resolve many of
the issues that have come up. Only a few minor code changes were necessary
to get the behavior we desired, and this version appears to beat any other
vercmp rendition on a few more cases added in this commit.
This commit passes all 58 vercmp tests currently out there. Other 'fixes'
still fail on a few tests, namely these ones:
test: ver1: 1.5.a ver2: 1.5 ret: -1 expected: 1
==> FAILURE
test: ver1: 1.5 ver2: 1.5.a ret: 1 expected: -1
==> FAILURE
test: ver1: 1.5-1 ver2: 1.5.b ret: 1 expected: -1
==> FAILURE
test: ver1: 1.5.b ver2: 1.5-1 ret: -1 expected: 1
==> FAILURE
4 of 58 tests failed
Signed-off-by: Dan McGee <dan@archlinux.org>
Now '-S provision' handling is done in the back-end.
In case of multiple providers, the first one is selected (behavior change:
deleted provision002.py). The old processing order was: literal, group,
provision; the new one: literal, provision, group. This is more rational,
but "pacman -S group" will be slower now. "pacman -S repo/provision" also
works. Provision was generalized to dependencies, so you can resolve deps by
hand: "pacman -S 'bash>2.0'" or "pacman -S 'core/bash>2.0'" etc. This can be
useful in makepkg dependency resolving. The changes were documented in
pacman manual.
alpm_find_pkg_satisfiers and _alpm_find_dep_satisfiers functions were
removed, since they are no longer needed.
I added some verbosity to "select provider instead of literal" and
"fallback to group".
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Signed-off-by: Dan McGee <dan@archlinux.org>
Also remove some meaningless pactests (broken requiredby, requiredby*.py
tests). requiredby001.py was renamed to upgrade076.py.
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Signed-off-by: Dan McGee <dan@archlinux.org>
I divided resolvedeps into 2 functions. The new _alpm_resolvedep function
will resolve one dependency, for example the 'foo>=1.0-1' dependency. It
can be useful in sync_addtarget refactoring.
The resolvedeps parameters were changed, to be coherent with recursedeps:
* the target-list is an alpm_list* instead of alpm_list**. This is OK,
because alpm_list_add == alpm_list_add_last
* syncpkg param was removed. list contains the to-be-installed packages,
resolvedeps will add the required dependencies into this list
* trans param was removed, it was used in QUESTION() only, which can be used
on the main (handle->trans) transaction only (because the front-end cannot
access our pseudo-transactions at all!).
The patch fixes some wrong dynamic pmdepmissing_t usage.
I did a behavior change (and sync1003.py was modified accordingly), which
needs some explanation: The old resolvedeps didn't elect packages from
'remove' list. I've dropped this because I don't want that 2nd excluding
list param. In fact, in real life, we ~never need this rule. Resolvedeps is
called before checkconflicts, so only -Su's %REPLACES% packages are sitting
in 'remove' list. This means, that we have the replacement packages in our
target list. Usually "foo replaces bar" means, that bar isn't in our repos
any more, so resolvedeps *cannot* elect it; but usually it won't try it at
all, because foo is in the target list, and it is expected to satisfy
'bar>=1.0-1'-like dependencies too. Since checkdeps and checkconflicts is
done after resolvedeps, this cannot cause any harm.
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Signed-off-by: Dan McGee <dan@archlinux.org>
Commit 8428367285 introduced the regression,
and a previous commit introduced the vercmptest.sh test script to track down
these issues. This commit solves the problem by removing the previous
attempt at locating the pkgrel portions and replacing it with something that
performs the correct logic.
While tracking down everything I needed to, I also found a mistake in one of
the pactests which is fixed here as well as increased the functionality and
verbosity of the vercmptest script to both print out each test it is running
as well as automatically run the mirror of each test case.
Signed-off-by: Dan McGee <dan@archlinux.org>
Commit 8428367285 updated the versioncmp code
in libalpm. Unfortunately for us, it also introduced the regression that
becomes apparant with the following upgrade:
warning: sonata: local (1.5-2) is newer than extra (1.5.1-2)
Add a vercmptest.sh test script that is run during the make check phase
which now points out three regressions in the version comparison function
that will need fixing. All current tests in this script pass with the old
versioncmp code.
Signed-off-by: Dan McGee <dan@archlinux.org>
This adds the sync equivalents of upgrade010.py and upgrade070.py.
Both additional test pass.
Signed-off-by: Allan McRae <mcrae_allan@hotmail.com>
Acked-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
This adds a test for when removing multilpe packages recursively from a
chain of dependent packages. This situation can occur when removing
installed dependencies with makepkg if a "makedepend" recursively depends on
a "depend" or if redundant dependancies are included.
Signed-off-by: Allan McRae <mcrae_allan at hotmail.com>
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Only use fakeroot and fakechroot when they are found AND required.
fakechroot only had the first condition, and fakeroot only the second.
When they are required (user != root) but not found, display a warning.
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
This patch offers a way to fix FS#9228.
By putting "SyncFirst = pacman" in pacman.conf, the version check will
happen before the transaction really starts, and before any replacements is
made.
Otherwise, no version check is done.
The sync301 pactest was updated to use this SyncFirst option.
Example session with SyncFirst = pacman, and a newer pacman version
available :
$ pacman -Su (or pacman -S <any targets>)
:: the following packages should be upgraded first :
pacman
:: Do you want to cancel the current operation
:: and upgrade these packages now? [Y/n]
resolving dependencies...
looking for inter-conflicts...
Targets: pacman-x.y.z-t
Total Download Size: x.xx MB
Total Installed Size: x.xx MB
Proceed with installation? [Y/n] n
As Nagy previously noted, doing this check on any -S operations might look
intrusive, but it can be required.
For example, the case where you want to install a package with versioned
provisions, using a pacman version which didn't support that feature yet
(and there is already a newer pacman in sync db supporting it).
Signed-off-by: Chantry Xavier <shiningxc@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
We were a bit juryrigged using one call to mkstemp() before rather than
extracting the new files side-by-side and doing our comparisons there. We
were also facing some permissions issues. Instead, make our life easier by
extracting all temp files to a '.paccheck' extension, doing our md5
comparisons, and then taking the correct actions.
Still to be done here- a cleanup of the use of PATH_MAX which should not be
necessary if we use dynamic allocation on the heap.
Signed-off-by: Dan McGee <dan@archlinux.org>
This should allow some future tests to set modes and ensure they are set
after installation. It is also in anticipation of a test for checking
permissions on pacnew files.
Signed-off-by: Dan McGee <dan@archlinux.org>
Previously, tar was called manually with os.system. This caused one fork per
package/db creation, which is costly, especially on cygwin. Besides, it also
caused some problems with directory with whitespaces (that could also be
fixed with quotes, but well..)
Using tarfile module is cleaner and more efficient, and still easy enough.
Benchmark (time make check) :
- windows / cygwin
prepatch:
real 6m36.360s
user 2m28.914s
sys 2m35.866s
postpatch:
real 5m25.428s
user 1m26.029s
sys 2m0.006s
- linux
prepatch:
real 1m22.629s
user 0m31.498s
sys 0m18.899s
postpatch:
real 1m11.465s
user 0m26.382s
sys 0m12.986s
Signed-off-by: Chantry Xavier <shiningxc@gmail.com>
This update causes query003 to fail before commit
4b8ada818e. Hopefully the issue doesn't crop
up again but we might as well test for it.
Signed-off-by: Dan McGee <dan@archlinux.org>
This patch should avoid duplicated target names in the backend.
1. sync_loadtarget will return with PM_ERR_TRANS_DUP_TARGET when trying to
add a duplicated target
2. sysupgrade never pulls duplicated targets
3. resolvedeps won't pull duplicated targets anymore
A pulled list was introduced in sync_prepare to improve the
pmsyncpkg_t<->pmpkg_t list conversion by making it more direct.
Also replace sync1005 and sync1006 by the sync1008 pactest, which is
similar but more interesting (the provisions are dependencies instead of
explicit targets).
sync1005 didn't work as expected anyway. It was expecting that pacman
failed, and pacman indeed failed, but not for the good reason. It didn't
fail during the preparation step because of conflicting targets, but during
the commit step, because of a md5 error...
And sync1006 didn't pass and was not really worth fixing. We have already
enough failing pactests more important than these two.
sync1008 pass with this patch.
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Signed-off-by: Chantry Xavier <shiningxc@gmail.com>
This gets rid of a lot of the unnecessary verbosity in the --valgrind output
when running all the tests. It should also make diff-ing output between test
runs a lot easier.
Signed-off-by: Dan McGee <dan@archlinux.org>
These case insensitive comparisons didn't work in some locales, like tr_TR
where upper(i) != I. So a second case sensitive comparison had to be made
for each directive.
Only keeping case sensitive comparisons make the code cleaner and treat all
locales equally.
Ref: http://www.archlinux.org/pipermail/pacman-dev/2008-March/011445.html
Also fix pactests to use the correct case.
Signed-off-by: Chantry Xavier <shiningxc@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
* -Rss removes all dependencies (including explicitly installed ones).
* updated documentation
* two pactest files added to test the difference between -Rs and -Rss
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Its implementation was quite broken:
* add_loadtarget() might have silently filtered out some targets when
replacing an older version.
* This was used in sync.c to determine whether a target is implicit or not,
which is incorrect behavior. Before this patch we silently removed user
confirmed replacements; now we always warn on a replacement.
* remove001.py behavior was quite odd in adding same target 5 times to the
target list, we can change this behavior to be a failure.
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
[Xav: changed remove001 pactest accordingly]
Signed-off-by: Chantry Xavier <shiningxc@gmail.com>
[Dan: rewrote commit message]
Signed-off-by: Dan McGee <dan@archlinux.org>
This hasn't been updated in forever, and a simple python program could
regenerate it anyway. Or even grep.
Signed-off-by: Dan McGee <dan@archlinux.org>
This is in anticipation of removing the -A/--add option from the pacman
frontend. I've went through each of the pacman pactests that used the -A
operation and decided whether they were worth keeping, whether there was
already an upgrade test doing the same thing, or whether it should be moved
over.
The GIT rename log should make several of the moves obvious, but for those
that were deleted:
add001: handled by upgrade004
add002: worthless
add004: worthless
add010: handled by upgrade011
add050: handled by upgrade010
Signed-off-by: Dan McGee <dan@archlinux.org>
This is the first step of fixing FS#9547. This should not break any existing
code that may rely on this function behaving the way it did, and should be
good for inclusion in a maint release.
In addition, update pactest so it fills the FILENAME field in the DB entries
it creates so we can move forward with a real fix to this issue.
Signed-off-by: Dan McGee <dan@archlinux.org>
If the end of the pB list was reached before the end of pA, we failed to
read any remaining files from the pA list. Add an additional loop to ensure
all entries of pA are added to the return list regardless of whether we have
reached the end of pB.
This new loop also eliminates the now-unnecessary check for a null pB, as we
need to ensure we are excluding directories in the resulting output anyway.
Signed-off-by: Dan McGee <dan@archlinux.org>
This patch fixes upgrade040.py and upgrade041.py (041 now fails!):
* the old pactests didn't check the existence of the relocated file
* upgrade041.py was broken due to a typo (missing comma)
New upgrade046.py pactest was added (derived from the fixed upgrade041.py).
This fails because the file relocation check is _hacked_ to conflict.c, and
_alpm_db_find_fileconflicts is not called in case of --force.
Signed-off-by: Nagy Gabor <ngaba@bibl.u-szeged.hu>
Signed-off-by: Chantry Xavier <shiningxc@gmail.com>
Bryan Ischo