summaryrefslogtreecommitdiffstats
path: root/urpm
Commit message (Collapse)AuthorAgeFilesLines
* adapt to aria2c-1.16+ output (mga#16500)Shlomi Fish2016-07-111-3/+24
| | | | | | | | | | | | | | | | aria2c > 1.14 changed the output of the progress download string it's no longer what follows: "[#1 SIZE:176.0KiB/2.5MiB(6%) CN:3 SPD:256.22KiBs ETA:09s]" but here is the new one from aria2c 1.16 and beyond: "[#2c8dae 496KiB/830KiB(59%) CN:1 DL:84KiB ETA:3s]" So the old parser no longer works... The string parser has been rewritten by Shlomi Fish with use of multiline mode and comments for better readability, should aria2c progress download string change again Dug & fixed by Giuseppe Ghibò <ghibomgx@gmail.com> and Shlomi Fish <shlomif@shlomifish.org>
* Prefer http mirrors when using a mirrorlistAnssi Hannula2016-07-111-1/+3
| | | | | | | | | | Sort http mirrors to the beginning of the mirror list. This avoids an rsync mirror being selected as the mirror sometimes, which causes the preferred aria2 downloader from not being used. This also has the effect that in practice aria2 will only be given http mirrors. If this causes issues, the sorting code can be changed to reverse sort with the presence of "rsync://" instead.
* fix using a proxy without an auth user (mga#11265)Thierry Vignaud2015-08-271-1/+1
|
* always display the right package name on errorsunknown2015-02-261-13/+2
| | | | | fixes mga#15032 and also simplify... needs URPM >= 5.04
* default to "unknown"Thierry Vignaud2015-02-061-1/+1
|
* Don't give a wrong package name when it is unknown in a callback (mga#15032)Pascal Terjan2015-02-051-2/+12
|
* s/suggests/recommends/Thierry Vignaud2014-09-071-2/+2
|
* s/suggested/recommended/Thierry Vignaud2014-09-072-3/+3
|
* adapt to URPM-5.00Thierry Vignaud2014-09-072-2/+2
|
* typo fixThierry Vignaud2014-09-071-1/+1
|
* rename '--suggests' option as '--recommends'Thierry Vignaud2014-09-071-1/+5
|
* add some missing methods, allows to do urpmf --qf '%sourcerpm:%recommends'Thierry Vignaud2014-09-071-0/+1
|
* use ->recommends_nosense() instead of suggests"()Thierry Vignaud2014-09-072-4/+4
| | | | | this is the new method name (even if compatibility still works). needs a new URPM.
* rename 'allow-suggests' as 'allow-recommends'Thierry Vignaud2014-09-071-3/+6
| | | | but keep an alias (with a warning) for compatibility
* rename 'no-suggests' option as 'no-recommends'Thierry Vignaud2014-09-072-5/+13
| | | | keep --no-suggests but with a warning
* rename 'no_suggests' parameter as 'no_recommends'Thierry Vignaud2014-09-051-1/+1
|
* fix retrieving rpm version with RCsThierry Vignaud2014-08-271-1/+1
|
* fix some warnings with perl 5.20Pascal Terjan2014-06-161-2/+2
|
* add some missing methods, allows to do urpmf --qf '%sourcerpm:%requires'Pascal Terjan2014-05-311-0/+5
|
* Use /proc/self/mountinfo for ro filesystem checks mga#11319Colin Guthrie2013-11-281-2/+7
|
* add support for --deploops optionThierry Vignaud2013-11-053-1/+3
|
* download: fix parsing download progress with rsync >= 3.10Olivier Blin2013-10-151-1/+1
| | | | | | | | | | | | | rsync 3.10 now defaults to human readable output, with sizes being printed with a separator between group of 3 digits. This broke our download progress parser. We can fix this by using the --no-human-readable command line option, to get the old rsync progress output format. This option already existed in older rsync versions (tried in Mageia 2), so this change should be safe to include in backports of urpmi to older distributions.
* adapt to perl-5.18's random hash keys orderingThierry Vignaud2013-10-041-1/+1
| | | | | | | | | | | | | | | | | | | | (translate_why_removed_one) ensure we got the right reason (testsuite) adapt to new URPM predictible behaviour it was previously relying on the following fixed hash order: $ perl -MData::Dumper -e 'warn Dumper [ keys {'rr1', 'rr1', 'rr2', 'rr2'} ]; \ warn Dumper [ keys {'bb1', 'bb1', 'bb2', 'bb2-1'} ]' $VAR1 = [ 'rr2', 'rr1' ]; $VAR1 = [ 'bb1', 'bb2' ];
* prevent removing orphan kernels if we failed to detect running kernel versionThierry Vignaud2013-10-041-1/+5
| | | | (eg: running testsuite in a chroot)
* revert fix for mga#10254 in previous release (mga#10335)Thierry Vignaud2013-05-301-1/+1
| | | | | | | the bug was actually in media.cfg basically reverting commit r8361 on 2013-05-24 (was: "fix wrongly enabling backports/testing for tainted/nonfree (mga#10254)")
* prevent mga#10254 bug to ever happen againThierry Vignaud2013-05-241-0/+1
|
* fix wrongly enabling backports/testing for tainted/nonfree (mga#10254)Thierry Vignaud2013-05-241-1/+1
| | | | (when such packages are already installed)
* library: ensure priority updates are installed in a single transaction.Colin Guthrie2013-05-191-0/+1
| | | | | | | | | | | | | | This fix is needed in cases where rpm's libdb changes. If the priority updates happen to be installed in more than one transaction, the %post script in rpm which deletes the old index files (/var/lib/rpm/__db.*) may happen at the end of the first of those transactions. This causes the indexes to be regenerated by the *old* db version and causes the second transaction to fail. By ensuring the priority updates are handled in a single transaction, the indexes will be recreated after the urpmi restart and will thus use the correct, new libdb to create the indexes. Patch and fix from Thierry Vignaud
* really auto enable tainted media if needed when adding distrib mediaThierry Vignaud2013-05-171-1/+1
|
* auto enable nonfree/taitned media if needed when adding distrib mediaThierry Vignaud2013-05-171-3/+17
|
* (is_media_to_add_by_default) split it out of add_distrib_media()Thierry Vignaud2013-05-171-8/+15
| | | | (needed for next commit)
* enable to see if nonfree/tained packages are installed (mga#8368, #8379 & #9758)Thierry Vignaud2013-05-171-0/+21
|
* fix leaving read locks on rpmdb (mga#9248)Thierry Vignaud2013-04-201-0/+1
| | | | | regression introduced in commit r7280 on Feb 5 2013: "(install_logger) unify with callback_uninst(), thus fixing counting for erasures"
* fix package count when a package has a %pretrans scriptletThierry Vignaud2013-04-011-4/+2
| | | | | | | | | | | | | | | | (spot by Colin on installing a minimal system with urpmi) callback_open got called one too much (it got called twice for first package which is actually filesystem), thus we bump the index one too much. the call chain is: rpmtsRun() -> runTransScripts() -> rpmteProcess() which says: /* Dont bother opening for elements without pre/posttrans scripts */ But since we now have package ('filesystem') with %pretrans due to /usr migration, it goes further and calls rpmteOpen()->rpmteFDHeader() which notifys us about a package opening in order to get its file descriptor.
* display exceptions that happened while downloading (mga#9529)Thierry Vignaud2013-03-261-0/+1
|
* document 'post_download' callback tooThierry Vignaud2013-03-211-0/+8
|
* note that urpm::util is a subset of MDK::CommonThierry Vignaud2013-03-211-0/+2
|
* kill unused return valuesThierry Vignaud2013-03-211-2/+0
|
* previous commit enables to further simplifyThierry Vignaud2013-03-211-1/+1
|
* simplify, let mount finds the loop deviceThierry Vignaud2013-03-212-20/+1
|
* treat formatList like others...Thierry Vignaud2013-03-212-2/+3
|
* simplifyThierry Vignaud2013-03-211-1/+1
|
* difference2 && untaint are not neeededThierry Vignaud2013-03-211-1/+1
|
* only a couple func really are neededThierry Vignaud2013-03-2119-19/+19
|
* drop 'callback_report_uninst'Thierry Vignaud2013-03-162-6/+2
| | | | rationale: 'callback_uninst' now enables to get erasure progress
* fix lookup deps of updates in other media when using --updateThierry Vignaud2013-03-161-4/+4
| | | | | | | reverting commit r1931 from 2011-09-14: "(configure,_auto_update_media,parse_media,update_media) fix (g)urpmi --auto-select --update defaulting to all media when there're no update media (#1024)"
* really enable callers to pass their own callback for erasure progressThierry Vignaud2013-03-162-1/+4
|
* simplifyThierry Vignaud2013-03-161-4/+1
|
* document trans_log parametersThierry Vignaud2013-03-161-1/+1
|
* hint @URPMThierry Vignaud2013-03-162-1/+5
|