Stefan Schubert [Thu, 7 Feb 2008 09:20:01 +0000 (09:20 +0000)]
cleanup
Stephan Kulow [Thu, 7 Feb 2008 09:16:44 +0000 (09:16 +0000)]
for the RPM-challenged
Stephan Kulow [Thu, 7 Feb 2008 09:13:23 +0000 (09:13 +0000)]
sync with autobuild
Stephan Kulow [Thu, 7 Feb 2008 09:11:36 +0000 (09:11 +0000)]
let's increase against factory
Klaus Kaempf [Thu, 7 Feb 2008 08:53:19 +0000 (08:53 +0000)]
introduce Solvable.kind to remove the need for strcmp()/strchr() when
accessing the name in libzypp and other applications.
Its also in preparation of per-kind solving (not implemented yet).
The name prefix (e.g. "pattern:") is still kept but only needed for
capability matching. kind is implemented as an offset (unsigned int)
matching the length of the name prefix. Applications can simply use
(solvable.name + solvable.kind) to strip the prefix from the name.
To prevent name clashes, "product:" was renamed to "prod:". Being
seldomly used, this seems to be acceptable.
Missing understandable documentation on the format of .solv files, the
kind value is currently NOT stored but computed on read.
tools/repo_susetags.c is already prepared to write the kind, solv
version6 should support storing kind values in .solv files.
Michael Matz [Wed, 6 Feb 2008 20:13:36 +0000 (20:13 +0000)]
Fix merge_attrs, so that not (e.g.) two medianr attributes show up in
susetags converted solvables using =Shr .
Michael Matz [Wed, 6 Feb 2008 19:55:36 +0000 (19:55 +0000)]
Make mergesolv work again.
Michael Matz [Wed, 6 Feb 2008 19:37:45 +0000 (19:37 +0000)]
Make repo_lookup_num work.
Stefan Schubert [Wed, 6 Feb 2008 10:05:44 +0000 (10:05 +0000)]
Prefer installed solvables first if there are two available with the
same name and evr. Bug#351941
Michael Matz [Wed, 6 Feb 2008 02:30:32 +0000 (02:30 +0000)]
Doesn't segfault anymore.
Duncan Mac-Vicar P [Wed, 6 Feb 2008 02:08:15 +0000 (02:08 +0000)]
expose repo_lookup_str and add lookup_num (nonworking)
Duncan Mac-Vicar P [Wed, 6 Feb 2008 02:08:09 +0000 (02:08 +0000)]
unbreak
Duncan Mac-Vicar P [Wed, 6 Feb 2008 02:08:04 +0000 (02:08 +0000)]
avoid segfault. FIXME please
Michael Matz [Wed, 6 Feb 2008 01:46:44 +0000 (01:46 +0000)]
I understand the solver a bit now , enough to hack around the problem causing
#354404 . See the elaborate comment. No additional test suite errors
(i.e. only those from the disabling of locale/language support).
Michael Andres [Tue, 5 Feb 2008 20:33:39 +0000 (20:33 +0000)]
fix segv when str2id(pool,NULL)
Michael Matz [Tue, 5 Feb 2008 17:34:17 +0000 (17:34 +0000)]
Iterating over the literals of a rule requires setting up the literal
list pointer from rule->d.
Michael Matz [Tue, 5 Feb 2008 14:06:26 +0000 (14:06 +0000)]
Bah, we have to handle the empty string too. Whenever there were empty
strings (STRID_EMPTY), e.g. in s->vendor, the SOLV file was fairly
messed up in subtle ways. This might fix #354404 (at least with this I
can't reproduce it). Probably most SOLV files need to be regenerated.
Stefan Schubert [Mon, 4 Feb 2008 13:41:07 +0000 (13:41 +0000)]
added pool to arguments
Duncan Mac-Vicar P [Fri, 1 Feb 2008 09:11:41 +0000 (09:11 +0000)]
trying to get a translated attribute working
Duncan Mac-Vicar P [Fri, 1 Feb 2008 09:11:35 +0000 (09:11 +0000)]
- unify join and makeevr
Duncan Mac-Vicar P [Thu, 31 Jan 2008 17:54:12 +0000 (17:54 +0000)]
xml store capabilities are parsed now.
next: attributes
Duncan Mac-Vicar P [Thu, 31 Jan 2008 17:54:11 +0000 (17:54 +0000)]
try to move split and join out
support for more attributes of xmlstore
I think I will need to use another strategy for
the capability tag, the stupid state machine cant handle it
Duncan Mac-Vicar P [Thu, 31 Jan 2008 17:54:09 +0000 (17:54 +0000)]
initial support for different kinds and pattern/patch store
tag names.
next: add support for store type of dependencies, which are slightly different
to the yum ones
Michael Schroeder [Wed, 30 Jan 2008 14:56:38 +0000 (14:56 +0000)]
- use namespace instead of solvable for languages
(breaks some test cases, solv files have to be regenerated)
Michael Schroeder [Wed, 30 Jan 2008 12:05:35 +0000 (12:05 +0000)]
- add optimization comment
Michael Schroeder [Wed, 30 Jan 2008 11:17:31 +0000 (11:17 +0000)]
allow empty strings
Michael Schroeder [Wed, 30 Jan 2008 11:17:07 +0000 (11:17 +0000)]
- allow empty string space
Michael Matz [Wed, 30 Jan 2008 02:53:07 +0000 (02:53 +0000)]
Use repodata_merge_attrs to implement =Shr again. Currently suboptimal,
as the blob data isn't shared in the file (the old attribute store
simply copied only ofs/len, the new one creates a new blob for each
reference to the same data snippet).
Michael Matz [Wed, 30 Jan 2008 02:51:16 +0000 (02:51 +0000)]
Split repodata_insert_keyid from repodata_set to create
repodata_merge_attrs.
Michael Matz [Wed, 30 Jan 2008 02:49:45 +0000 (02:49 +0000)]
A subtle but fatal off-by-one.
Michael Matz [Wed, 30 Jan 2008 01:13:11 +0000 (01:13 +0000)]
Storing the schema in front of the incore data releaves us from fiddling
with the entryschema members.
Michael Matz [Wed, 30 Jan 2008 00:14:15 +0000 (00:14 +0000)]
Document repodata_internalize a bit, after understanding what it does :)
Michael Matz [Tue, 29 Jan 2008 23:28:03 +0000 (23:28 +0000)]
I'm fairly sure this is right. Probably untested code path.
Duncan Mac-Vicar P [Tue, 29 Jan 2008 22:34:58 +0000 (22:34 +0000)]
compile
Michael Matz [Tue, 29 Jan 2008 20:55:09 +0000 (20:55 +0000)]
Increment needid[] of the right member.
Michael Andres [Tue, 29 Jan 2008 19:34:46 +0000 (19:34 +0000)]
remove superfluous Pool argument in policy_illegal_ functions
Michael Matz [Tue, 29 Jan 2008 18:51:51 +0000 (18:51 +0000)]
no comment
Michael Matz [Tue, 29 Jan 2008 17:24:52 +0000 (17:24 +0000)]
Small nit.
Michael Schroeder [Tue, 29 Jan 2008 16:57:10 +0000 (16:57 +0000)]
- split assignment to shut up stupid g++ compiler
Michael Schroeder [Tue, 29 Jan 2008 16:14:03 +0000 (16:14 +0000)]
- add key filtering to repo_write
- change repo_write so that it combines all available data
- integrate attr_store into repodata
- write storage parameter with every key, bump solv revision to 5
- don't create system rules for atoms
- change repo_susetags to use repodata interface
Stefan Schubert [Mon, 28 Jan 2008 16:16:52 +0000 (16:16 +0000)]
Added callbacks for policy handling.
Michael Schroeder [Fri, 25 Jan 2008 17:50:39 +0000 (17:50 +0000)]
- add dirpool support
Klaus Kaempf [Fri, 18 Jan 2008 16:21:21 +0000 (16:21 +0000)]
introduce 'Covenants' as a generic representation of 'Locks'
Covenants are bound to the Solver and should be taken into account
when solving.
Still incomplete !
Michael Matz [Thu, 17 Jan 2008 16:28:52 +0000 (16:28 +0000)]
Very subtle sign/zero-extension bug occuring on 64bit.
Michael Schroeder [Tue, 15 Jan 2008 15:37:29 +0000 (15:37 +0000)]
- add dir reading code to solv reader
- change DU skeleton to use new layout
- add new type TYPE_DIR
Michael Matz [Mon, 14 Jan 2008 16:33:13 +0000 (16:33 +0000)]
Let's give fastlz a license, shall we?
Michael Matz [Mon, 14 Jan 2008 16:25:33 +0000 (16:25 +0000)]
It's nicer to represent the special numbers as TYPE_VOID with the actual
value sitting in key->size.
Stephan Kulow [Mon, 14 Jan 2008 16:17:58 +0000 (16:17 +0000)]
package submitted
Stefan Schubert [Mon, 14 Jan 2008 10:52:08 +0000 (10:52 +0000)]
reduced looging
Michael Schroeder [Mon, 14 Jan 2008 10:08:11 +0000 (10:08 +0000)]
- add evrcmp_str for ma
Michael Schroeder [Fri, 11 Jan 2008 20:42:25 +0000 (20:42 +0000)]
- fix support for splitprovides
- add some unfinished code for DU queries
Michael Schroeder [Wed, 9 Jan 2008 19:08:06 +0000 (19:08 +0000)]
- clarify
Stefan Schubert [Wed, 9 Jan 2008 15:59:27 +0000 (15:59 +0000)]
docu added
Michael Schroeder [Wed, 9 Jan 2008 15:32:39 +0000 (15:32 +0000)]
- beautify a bit
Klaus Kaempf [Wed, 9 Jan 2008 13:53:13 +0000 (13:53 +0000)]
Problem.to_s and Decision.to_s
Michael Schroeder [Wed, 9 Jan 2008 11:55:50 +0000 (11:55 +0000)]
- rename xmalloc/... functions to sat_malloc, as we're a
library and mustn't conflict with other libraries
Michael Schroeder [Tue, 8 Jan 2008 18:23:35 +0000 (18:23 +0000)]
- assertify
- add direct decision counter to make reset_solver faster and easier
to understand
Klaus Kaempf [Tue, 8 Jan 2008 16:43:32 +0000 (16:43 +0000)]
fprintf line number immediately before calling abort()
... as an incentive to convert it to a proper solution ;-)
Stephan Kulow [Tue, 8 Jan 2008 15:28:01 +0000 (15:28 +0000)]
sync with autobuild
Stephan Kulow [Tue, 8 Jan 2008 15:02:53 +0000 (15:02 +0000)]
prepare another autobuild round
Michael Schroeder [Mon, 7 Jan 2008 18:51:06 +0000 (18:51 +0000)]
- also push level 1 rules on leanrt stack
- this also fixes a nasty bug where a learnt rule was
not disabled by enabledisablelearntrules()
Klaus Kaempf [Mon, 7 Jan 2008 16:37:01 +0000 (16:37 +0000)]
update to current solver format (data taken from 10.3-i386)
Klaus Kaempf [Mon, 7 Jan 2008 16:36:07 +0000 (16:36 +0000)]
set pool architecture
Klaus Kaempf [Fri, 4 Jan 2008 14:53:05 +0000 (14:53 +0000)]
adapt to swig bindings (incomplete, non-working)
Klaus Kaempf [Fri, 4 Jan 2008 12:25:53 +0000 (12:25 +0000)]
further steps towards Perl bindings, still incomplete
- merge stuff from Kiwi
- work around bugs in CMake FindPerlLibs
Stefan Schubert [Thu, 3 Jan 2008 16:25:17 +0000 (16:25 +0000)]
comment added
Stefan Schubert [Thu, 3 Jan 2008 15:10:22 +0000 (15:10 +0000)]
comments added
Stefan Schubert [Thu, 3 Jan 2008 08:30:00 +0000 (08:30 +0000)]
comments
Klaus Kaempf [Sun, 30 Dec 2007 20:22:21 +0000 (20:22 +0000)]
negate p before use as offset
Klaus Kaempf [Sun, 30 Dec 2007 16:58:01 +0000 (16:58 +0000)]
add Decision (solver result element) and testcase
Klaus Kaempf [Fri, 28 Dec 2007 20:37:05 +0000 (20:37 +0000)]
fix examples/ruby/basic.rb
add create_repo, create_transaction and create_solver to Pool
rename 'erase' to 'remove', its more natural
rename Pool.count_repo to Pool.count_repos
Michael Matz [Fri, 28 Dec 2007 00:05:04 +0000 (00:05 +0000)]
Install content2solv.
Michael Matz [Fri, 28 Dec 2007 00:04:38 +0000 (00:04 +0000)]
Memory efficient representation of the dirtree (important for
unabridged packages.DU). Store and reread the dirtree (in a hacky way,
should use the info block). Make dumpattr show the directory names.
On disk the same info as in my 10.3 packages.DU (1.4MB, 330 KB gzipped)
only takes 146 KB.
Michael Matz [Thu, 27 Dec 2007 21:49:26 +0000 (21:49 +0000)]
No idea why I had this as intlist.
Michael Matz [Thu, 27 Dec 2007 21:44:04 +0000 (21:44 +0000)]
Parse also +Dir:, i.e. the packages.DU file. For now the info is stored
simply as an ATTR_INTLIST and the dirtree isn't saved at all.
Michael Matz [Thu, 27 Dec 2007 21:39:00 +0000 (21:39 +0000)]
Enable ATTR_INTLIST to contain zeros. Use the same encoding as IDARRAY.
Michael Matz [Thu, 27 Dec 2007 20:25:17 +0000 (20:25 +0000)]
Some infrastructure for directory trees.
Michael Matz [Thu, 27 Dec 2007 17:32:46 +0000 (17:32 +0000)]
I'll have a need for inserting substrings into a stringpool.
Duncan Mac-Vicar P [Sat, 22 Dec 2007 17:16:20 +0000 (17:16 +0000)]
fix example broken by module renames
Klaus Kaempf [Sat, 22 Dec 2007 08:17:02 +0000 (08:17 +0000)]
prepare to pass rootdir for reading rpmdb from other directory
(unfinished)
Michael Schroeder [Fri, 21 Dec 2007 18:16:33 +0000 (18:16 +0000)]
- add evrmatch
- first shot at adding file provides
Michael Schroeder [Fri, 21 Dec 2007 14:42:40 +0000 (14:42 +0000)]
- fix bug in repo_addid_dep
- make repo_addid_dep handle different markers, so that we can add
the file provides with SOLVABLE_FILEMARKER
Michael Matz [Fri, 21 Dec 2007 11:58:38 +0000 (11:58 +0000)]
Improve the page eviction a bit.
Add possibility to merge attributes between entities, use it to handle
=Shr susetags headers.
Michael Matz [Fri, 21 Dec 2007 10:13:29 +0000 (10:13 +0000)]
Let's not compare strings in an if-cascade or lookup the same names all
the time.
Michael Schroeder [Thu, 20 Dec 2007 21:30:41 +0000 (21:30 +0000)]
- some bug fixes + stop handling, nothing fancy
Michael Matz [Tue, 18 Dec 2007 21:09:58 +0000 (21:09 +0000)]
One-valued types for media number. They can be encoded in the schema,
no need for storing 1, 2 or 3 all the time.
Michael Schroeder [Tue, 18 Dec 2007 18:39:55 +0000 (18:39 +0000)]
- some more things to think about
Michael Schroeder [Tue, 18 Dec 2007 15:14:36 +0000 (15:14 +0000)]
- do not exit on bad solv files
Michael Schroeder [Mon, 17 Dec 2007 17:53:01 +0000 (17:53 +0000)]
- removed verticals for now to make work easier
- allow NULL idmap
Michael Schroeder [Fri, 14 Dec 2007 20:11:33 +0000 (20:11 +0000)]
- forgot to store start and end of id block
Michael Schroeder [Fri, 14 Dec 2007 20:08:42 +0000 (20:08 +0000)]
- some pieces of code for the unified lookup/search interface
Michael Schroeder [Tue, 11 Dec 2007 13:14:00 +0000 (13:14 +0000)]
- beautify
Michael Matz [Tue, 11 Dec 2007 03:18:57 +0000 (03:18 +0000)]
Handle most of the susetags tags. Major exception is 'Shr', where I'm
still undecided if I want to copy the info verbatim (easier to use), or
if I want to store the reference expecting the users to follow the chain
on his own. packages.DU and filelists not yet handled.
Michael Schroeder [Mon, 10 Dec 2007 20:16:17 +0000 (20:16 +0000)]
- use schema in repodata to make it extendible
Michael Andres [Mon, 10 Dec 2007 15:09:01 +0000 (15:09 +0000)]
include stdio
Michael Schroeder [Mon, 10 Dec 2007 14:58:44 +0000 (14:58 +0000)]
- add evrcmp mode parameter
Michael Matz [Sat, 8 Dec 2007 04:58:04 +0000 (04:58 +0000)]
Pfeww. I'm tired, but now you can add refers from repo SOLV files to
attribute store files. Good night.
(Like so:
% addstore bla.solv ding.attr dong.attr > bla2.solv
% dumpsolv bla2.solv
repo refers to 2 attribute stores:
ding.attr has 4 keys
authors
group
keywords
license
dong.attr has 4 keys
authors
keywords
license
...
)
Michael Matz [Sat, 8 Dec 2007 03:08:32 +0000 (03:08 +0000)]
I'm going to use the file info block real soon now and noticed that
doing it with the schema way brings only hassle for no good. The info
is per file, hence probably the same scheme won't be reused by anything
else. Hence going back to nearly the old way. I do have use for a
counted type, in order to skip safely over unknown entries. Old SOLV
files can be read when they don't have any file info. None should
have.
Michael Matz [Sat, 8 Dec 2007 01:53:58 +0000 (01:53 +0000)]
Also activate paging for embedded attr stores.
Michael Matz [Sat, 8 Dec 2007 01:24:52 +0000 (01:24 +0000)]
Pull together attribute store and repo. dumpsolv actually now really
can also show the content of test.attr, yeah.