summaryrefslogtreecommitdiff
path: root/usr.sbin/pkg_add
AgeCommit message (Collapse)Author
2006-08-05actually perform error-checking on every read/write.Marc Espie
2006-08-01If pkg_delete is trying to remove more than one package at the same time andBernd Ahlers
is finding some dependency problems, explain which package depend on what. (with -v, -i or '-F verbosedeps') idea & ok espie@
2006-07-31cosmetic: comments, and better method name.Marc Espie
2006-07-31add missing licence info.Marc Espie
use strict/warning where applicable, and fix relevant warnings.
2006-05-26make pkg_add exit gracefully if it cannot read from a file handle, whichSteven Mestdagh
happens when the ssh connection could not be established. ok espie@
2006-05-04more clues for people who update their system.Marc Espie
2006-04-22Pass over the right object to OpenBSD::Delete::validate_plist().Bernd Ahlers
This fixes an error while checking for collision with old libs. okay espie@
2006-04-08Add missing 'use OpenBSD::Error'.Bernd Ahlers
spotted by marco@ 'go ahead' espie@
2006-03-21fix endless loop in case of error in archive header.Marc Espie
2006-03-19Partial revert of previous changes, nasty regexps not greedy the way they ↵Marc Espie
should. Broke md5.
2006-03-17un-special case old constructs: make them inherit from the same class,Marc Espie
and remove any special code from Factory. As it's simpler remove local variables.
2006-03-13clean-up some name handling:Marc Espie
- package locations always contain a pkgname (except if we don't have it) - complete it on the repository side to grab the right file. - no need to pass pkgname/arch on $location->grabPlist(), since it's already in the location.
2006-03-13repair pkg_add -Marc Espie
2006-03-08propagate stuff to be able to explicitly clean-up repositories.Marc Espie
(we can't use a normal destructor, since this will be called for each fork()ed subprocess).
2006-03-08fix race condition in SCP for real.Marc Espie
We can't control a grand-child death through gzip, so stop spawning grand-children: create two children connected by hand through a pipe, and when we close gzip, explicitly wait for the second child to die as well. This avoids race conditions between sigpipe and sigusr1. (and as usual, the resulting code is easier to follow once you get through the pipe/fork). Thanks theo for the comment. ;-)
2006-03-07do not write partial files out if the PKG_CACHE filesystem overflows.Marc Espie
2006-03-07Fix pkg_copy().Bernd Ahlers
2006-03-07properly zap temporary file.Marc Espie
2006-03-07rendez-vous with subchild through signal USR1.Marc Espie
2006-03-07make pkg_copy be overidable by subclass.Marc Espie
2006-03-07only the grand child can have issuesMarc Espie
2006-03-07special case: repository holds several packages (say, old versions), andMarc Espie
you choose the most recent one, which also happens to be the currently installed version -> then explicitly say no update is needed and proceed.
2006-03-07let scp repositories work. Avoid run-away processes by just having one singleMarc Espie
ssh connection for each repository.
2006-03-07pass the grandchild pid through the pipe, in order to be able to monitor it.Marc Espie
2006-03-06zap object pid when done with it.Marc Espie
2006-03-06actually zap the codeMarc Espie
2006-03-06typoMarc Espie
2006-03-06take some infrequently used code completely out of the way.Marc Espie
2006-03-05avoid running mkfontdir/fc-cache if they don't exist.Marc Espie
2006-03-05SynchMarc Espie
2006-03-05- kill whitespace at eolJason McIntyre
- kill double .Pp
2006-03-04cut down the Locator code into maintainable chunks.Marc Espie
2006-03-04accurate description of script execution.Marc Espie
2006-03-04Spell out normal modes of operation for pkg_add more.Marc Espie
Explain about ambiguities, and expand a bit on interactive mode.
2006-03-04reorganize code a little bit: Update.pm becomes Replace.pm (since it matchesMarc Espie
the -r option), and the find_update code moves to Update.pm, so that it doesn't get parsed if -u is not used. Also make the has_new_sig and uses_old_libs local methods of PackingList, as it is their natural API. Reorg a few more routines so that they can be found by pkg_add and by Update.pm. May need to split off more stuff to avoid parsing everything...
2006-03-04zap more success messages. Keep pid of distant location pipe, just in case.Marc Espie
2006-02-23Let 'pkg_info -P' work with package distribution files and URLs to aBernd Ahlers
package. ok espie@
2006-02-22Fix typo.Bernd Ahlers
2006-02-21detect temporary connection errors, and retry with exponential timeout.Marc Espie
2006-02-21ask question corresponding to -F updatedepends too (commonalize code).Marc Espie
Use key 'update', 'updatedepend' to avoid answering the same question again and again.
2006-02-21add optional 3rd choice [y/N/a] for always, so that this question typeMarc Espie
will always match.
2006-02-16validate() can also take sample files into account.Marc Espie
2006-02-10reset $state->{dir} to the correct value after deleting the replaced package.Marc Espie
Fixes a problem noticed by steven.
2006-02-10Use the ProgressMeter engine to display the update messages. This makesBernd Ahlers
'pkg_add -u' less chatty. (turn on verbose output to restore the old behaviour) Advise the user to re-run 'pkg_add -u' in interactive mode if there are ambiguities. Discussed with steven@ and espie@. Testing and some ideas by steven@, thanks. ok espie@
2006-02-10If the length of the message in OpenBSD::ProgressMeter::message() is shorterBernd Ahlers
than the length of the playfield, fill up the difference with whitespace. Prevents leftover characters from previous message()s. ok espie@
2006-02-09Turn validate_plist() into a real visitor that calls validate() for eachMarc Espie
object in the packing-list. Move some checks earlier, so that update of packages fails before deleting the old package. Add more checks, like verifying that install/deinstall scripts are indeed located on exec'able partitions.
2006-02-09use new interface (*close*)Marc Espie
optimization: store dir into state earlier, so that do_script and other stuff can have access to it.
2006-02-09Add some nice interfaces instead of the hint() argument to close().Marc Espie
This allows the cache to finish properly even in some weird cases like fat packages (calling finish_and_close).
2006-02-07synchMarc Espie
2006-02-07minor independent tweaks:Marc Espie
- make Newuser and Newgroup subclasses of Newauth, so that code can be shared. - add property type() for these, that expands to user/group - add property to SpecialFile() that says whether this is executed on add/delete, so that we can check for noexec fs and error out early on these.