| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|
|
|
|
| |
o enhancement: only parse updated synthesis
|
|
|
|
|
|
|
| |
o drop option "-c" which used to clean /var/cache/urpmi/headers
since we don't use that dir anymore :)
|
|
|
|
|
|
|
|
|
| |
(this will allow checking xml media_info is up-to-date even if we don't
update it at the same time as synthesis is updated)
and further cleanup urpm::media code
(esp. modify {force} behaviour, may need some adjustments)
|
|
|
|
|
| |
o drop hdlist support (will be replaced with xml media_info)
|
|
|
|
|
|
|
|
| |
o do remove __db* on priority-upgrade
(fix regression introduced in 4.10.15)
(may fix #36190)
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
o fix handling "post-clean: 0" in urpmi.cfg
(#36082, regression introduced 2007-11-28)
- library:
o urpm::media::read_config() doesn't read urpmi.cfg global options anymore,
use urpm->get_global_options or urpm->new_parse_cmdline
|
| |
|
| |
|
|
|
|
|
|
| |
o adapt to perl-URPM 2.11 API to parse pubkey files
(nb: $urpm->{keys} is not used anymore)
|
|
|
|
|
| |
o handle --suggests (to override urpmi.cfg global option "no-suggests") (#36007)
|
|
|
|
|
|
|
|
| |
o exit code 1 when a medium can't be added
and simplifies a mess in urpmi.addmedia: this mess is not needed anymore since
update_media() will exit on error
|
|
|
|
|
|
|
|
| |
o exit code 1 when a (selected) medium can't be updated (#35952)
o leave early with error if no medium were successfull updated
nb: also fix returning "unmodified" instead of "1" when a medium is not updated
|
|
|
|
|
|
| |
o enhance parsing of urls with login:password for logins with "@"
so that password doesn't end up in urpmi.cfg
|
| |
|
| |
|
|
|
|
|
|
|
| |
o fix regression introduced in 4.10.15:
"urpmi --force valid invalid" should warn about "invalid" but still
install "valid"
|
| |
|
|
|
|
|
|
|
|
| |
o small transactions should have at least 8 packages
(ie --split-length is now 8 by default)
o do not do a big transaction if installing less than 20 packages
(ie --split-level is now 1 by default)
|
|
|
|
|
| |
o bugfix previous release: "urpmq --fuzzy foo" should display all "*foo*" matches
|
| |
|
|
|
|
|
| |
o new option --requires-recursive (alias of option -d)
|
| |
|
|
|
|
|
|
|
| |
o exit with code 1 after displaying usage (instead of exit code 0)
(urpmi exit code "1" is correctly documented in urpmi's manpage)
|
|
|
|
|
|
|
|
|
| |
o --auto-update is quite unsafe, but at least now it should be cleaner
(ensure $urpm doesn't have media twice)
bug triggered by --auto-update --update. still --auto-update --update should
not update non-update media...
|
|
|
|
|
| |
o fix --limit-rate (regression introduced in 4.9.12)
|
| |
|
|
|
|
|
|
|
|
| |
o fix "urpmi --install-src" (regression introduced in 4.10.9)
nb: packages_by_name can't be used here since $urpm->{provides} doesn't have src.rpm
so a full search is needed
|
|
|
|
|
|
| |
o with rsync, use --copy-links (to have the same behaviour as http/ftp, and
so allow symlinks on the server)
|
| |
|
|
|
|
|
| |
o urpm::media::add_distrib_media: add option "only_updates" for rpmdrake
|
| |
|
|
|
|
|
| |
o remove prefix/var/lib/rpm/__db* after installing pkgs rooted
|
| |
|
|
|
|
|
|
|
|
|
| |
o ensure a second pass is done even if media has not been modified
(useful for the %trigger trick done in urpmi to rebuild synthesis with
suggests)
without it, file provides are not computed properly
|
|
|
|
|
| |
o fix -a and -o
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
previously only index by raw package name (w/o version and w/o arch)
which resulted in packages being overwritten (ie on a biarch, we would
only saw one of both ia32 & x86_64 packages)
|
| |
|
| |
|
|
|
|
|
|
|
| |
o fix --synthesis
(by handling synthesis argument the same way in urpmq, urpmi and urpmf)
|
|
|
|
|
|
|
| |
o default to selecting all the prefered packages according to installed
locales (need perl-URPM 1.81 to work)
|
|
|
|
|
|
|
|
|
| |
o do not prompt questions if not waiting for user answer
(for urpmi --force or urpme --force)
create message_input_() which deprecated message_input(), $o_default_input is
dumb IMO
|
| |
|