- build_transaction_set: new sort algorithm which allow returning sets of circular dependent packages, taking into account obsoleted packages (fixes #31969). It may still fail in presence of conflicts - allow running transaction with justdb option - fix split_length > 1 (eg: "urpmi --split-length 2 a b c" will only install 2 pkgs) - spec2srcheader: workaround parseSpec returning a header where ->arch is set to %{_target_cpu} whereas we really want a header similar to .src.rpm (see #32824) Version 1.75 - 12 August 2007, by Pascal "Pixel" Rigaux - fix dropping tags from rpm header. it hasn't work since MDK8.1 and rpm 4.0. it may break urpmi!! but potentially allows a much smaller hdlist.cz :) Version 1.74 - 12 August 2007, by Pascal "Pixel" Rigaux - sort choices per media, then per version Version 1.73 - 11 August 2007, by Pascal "Pixel" Rigaux - allow running transaction with replagekgs option Version 1.72 - 10 August 2007, by Pascal "Pixel" Rigaux - modify parse_hdlist so that partial hdlist reading can be used (needed when some stuff is already done in the callback) Version 1.71 - 9 August 2007, by Pascal "Pixel" Rigaux - compilation fixes on rpm < 4.4.8 Version 1.69 - 9 August 2007, by Pascal "Pixel" Rigaux - "suggests" are no more handled as "requires" - resolve_requested support "suggests": a newly suggested package is installed as if required (can be disabled with option no_suggests) Version 1.68 - 3 August 2007, by Pascal "Pixel" Rigaux - add $trans->Element_version and $trans->Element_release Version 1.67 - 22 June 2007, by Olivier "Nanar" Thauvin - add osscore, archscore and platformscore function to URPM - is_platform_compat function to Pkg object - fix call to rpm function in spec2header() - fix some compilation warnings Version 1.66 - 2 July 2007, by Pascal "Pixel" Rigaux - fix --auto-select skipping some packages because of other packages providing a more recent version, but no obsolete between those packages. the fix is to revert commit from Aug 2002: "fixed propable old package (according provides) requested by request_packages_to_upgrade." Version 1.65 - 22 June 2007, by Olivier Thauvin - really fix arch_score evaluation Version 1.64 - 12 June 2007, by Pascal "Pixel" Rigaux - hack on $pkg->is_arch_compat to make it return true for noarch packages when using rpm 4.4.8 (#31314) Version 1.63 - 9 May 2007, by Pascal "Pixel" Rigaux - add $trans->Element_fullname Version 1.62 - 3 May 2007, by Pascal "Pixel" Rigaux - pass the virtual package name as a parameter to {callback_choices} in ->resolve_requested - add $trans->NElements and $trans->Element_name to be able to display name of uninstalled package in callback_uninst - fix b--obsoletes-->a and c--conflicts-->a prompting for upgrading a (need a fix in urpmi which rely on the $state->{rejected} to upgrade (-U) b instead of installing (-i) it) Version 1.60 - 8 March 2007, by Pascal "Pixel" Rigaux - more debugging hooks - create $urpm->packages_providing($name) and use it - create $urpm->packages_by_name($name) Version 1.59 - 1 March 2007, by Olivier Thauvin - rpm 4.4.8 adaptions - load rpm config files at module load, improve the mechanism Version 1.58 - 14 February 2007, by Pascal "Pixel" Rigaux - don't check signature and digest in ->traverse and ->traverse_tag (=> x15 speedup, ie 2.5 speedup on urpmi --auto-select and rpmdrake) Version 1.57 - 9 February 2007, by Pascal "Pixel" Rigaux - allow upgrading from noarch to x86_64 even if strict-arch Version 1.56 - 19 January 2007, by Pascal "Pixel" Rigaux - tell perl that strings from rpm headers are utf8 - add URPM::bind_rpm_textdomain_codeset() to set encoding of messages returned by rpmlib, and tell perl that those strings are utf8 - really use strict-arch by default on x86_64 Version 1.55 - 10 January 2007, by Pascal "Pixel" Rigaux - bug fix release - fix "not selecting foo-1 since the more recent foo-1 is installed" causing urpmi to try to remove the package it wants to install (#28076) Version 1.54 - 9 January 2007, by Pascal "Pixel" Rigaux - if we have a choice between foo-1 and bar-1 and foo-2 is installed, prefering bar-1 instead of foo-1 (otherwise we can hit: "the more recent foo-2 is installed, but does not provide xxx whereas foo-1 does", cf bug #27991) - bar is needed, foo-1 does provide bar, installed foo-2 does not provide bar: do not let the algorithm use foo-2 as if it also provides bar - allow understanding what ->resolve_requested is doing through a callback ($urpm->{debug_URPM}) - cleanup some code in ->resolve_requested - make the documentation for ->is_arch_compat more clear