| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
aka having @recommends@ lines instead of @suggests@ ones
|
|
|
|
|
| |
in order to reflect reality
but keep a suggests alias for compatibility for the time being
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
rpm-4.12.0 introduces a new weak deps implementation (the 3rd one to be
used in mdv/mga!)
The tag we used to use (RPMTAG_SUGGESTSNAME) is now OLDSUGGESTSNAME
=> we need to handle both old & new tags
Hopefully rpm-4.12.x enables to check both old & new tags when using
xxxNVRS variants
We thus have a choice between RECOMMEND & SUGGEST tags, that both
fallback on older SUGGESTS tags if no new ones are found.
The difference is that they filter old suggests depending on
RPMSENSE_STRONG hint.
As we used suggests in the recommends role, this is the perfect time to
switch from suggests to recommends
(with a patch on rpm in order to fallback RECOMMEND on weak (old)
SUGGESTS instead of on strong (old) ENHANCES as we didn't used the later
in mga)
Our rpm is patched so that RECOMMENDNAME fallbacks too.
In Suse, only RECOMMENDNEVRS is the only RECOMMEND variant that handle
the fallback on old tags.
However, using RPMTAG_RECOMMENDNEVRS makes URPM->suggest_nosense to
return versionated suggests, which while not supported by urpmi, do are
used by a couple packages (such as git or gnome-user-share)
This breaks urpmi as suggests on "foo >= x.y" is not interpreted by
urpmi which thinks that "foo" is a new suggests and reinstalls a
previously removed suggests
Hence we patched rpm.
|
| |
|
|
|
|
|
|
| |
suggests to use compare_pkg instead
(needs a new rpmdrake)
|
|
|
|
| |
We don't have those since 2004 (10 years+)...
|
| |
|
| |
|
|
|
|
|
|
| |
this broke installing 32 bit kernel for drakx-installer-binaries
(spotted by Pascal Terjan)
|
|
|
|
|
|
|
| |
'Argument "19697|19..." isn't numeric in array element at ...URPM/Resolve.pm'
introduced by pixel in commit 1d945d2a483598dc982cf73ae932d83d4e35df9
on Sep 7 2007 ("add some debug messages)
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
(for rpmdrake --env)
|
| |
|
|
|
|
|
| |
'Pointer "foo" returned by "_free(foo)" is never used'
(improve coverity scan's SNR)
|
|
|
|
| |
thus fixing urpmi testsuite...
|
| |
|
| |
|
| |
|
|
|
|
|
| |
enabling eg to switching back to packages from core/release after
having installed some from core/updates_testing, then disabling it
|
| |
|
|
|
|
|
|
| |
commit r6545 on 2012-12-2 broke delta rpm support:
"(Pkg_description, Pkg_sourcerpm) split UTF-8 strings from raw strings
functions, then simplify"
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
revert part of commit r5052 (June 28 2012, "light speedups")
|
| |
|
|
|
|
| |
(that cause rpm to abort on assert while creating a new pool)
|
| |
|
|
|
|
|
| |
RPMFILE_UNPATCHED is no more used and RPMFILE_EXCLUDE only exists during
spec parse (see rpm commit 8d277bbb)
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
this mean prefering package from first medium when same NEVR exists in
several media, thus prefering 64bit media over 32 bit ones (mga#7524)
complement commit r2909 on Feb 7 2012:
" (find_required_package) really prefer packages from first media, thus
fixing downloading noarch packages from 32 bit repository (mga#1603)"
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
found accidently
could be triggered by the following (calling enough times either
requires() or provides() while traversing rpm db) which resulting in
an "Invalid read of size 8":
use URPM;
my $db = URPM::DB::open()->traverse_tag_find('whatprovides', 'kernel',
sub {
my ($p) = @_;
warn $p->requires, $p->requires, $p->requires, $p->requires, $p->requires,
$p->requires, $p->requires, $p->requires, $p->requires, $p->requires, $p->requires,
$p->requires, $p->requires, $p->requires, $p->requires, $p->requires, $p->requires;
return 1;
});
bug introduced in commit r4947 on June 14 2012:
"(_run_cb_while_traversing) introduce this helper"
|
| |
|
| |
|
| |
|
|
|
|
| |
(it should have been named suggests_nosense in the first place, bad pixel...)
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
calling ->provides or ->requires in traverse_tag_find() results in an
invalid read if provides/requires list is big (or if called several
times)
basically reverting commit r6476 on 2012-11-16
("(unsatisfied_requires) (slightly) optimize")
|
| |
|
| |
|
| |
|