summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
Stephan Kulow [Wed, 31 Oct 2007 08:45:27 +0000 (08:45 +0000)]
fixing misunderstanding
Michael Matz [Wed, 31 Oct 2007 04:11:20 +0000 (04:11 +0000)]
The blobs are now stored and loaded. If the input file for the attr
store is seekable we load the blobs on demand, and as we do this in
pages we don't even need much more memory (currently I'm using 4 pages
at 32KB each, so the constant memory overhead for all authors and
descriptions and other blobs is only 128Kb per attr store). And it
isn't even slow :-)
Michael Matz [Tue, 30 Oct 2007 21:45:56 +0000 (21:45 +0000)]
Parse also a second packages file (e.g. packages.en) catted to stdin,
so that we can annotate already existing packages, e.g. with summary and
description. Store authors and description as blobs.
Michael Matz [Tue, 30 Oct 2007 21:42:51 +0000 (21:42 +0000)]
We can create blobs now. But can't yet store them into the file :)
Michael Schroeder [Tue, 30 Oct 2007 18:53:36 +0000 (18:53 +0000)]
- stuff to discuss
Michael Schroeder [Tue, 30 Oct 2007 17:55:35 +0000 (17:55 +0000)]
- add vendor change policy (but fixes no test case, as there is no
good test case)
Michael Schroeder [Tue, 30 Oct 2007 11:27:12 +0000 (11:27 +0000)]
- add vendor id to solvables
Michael Matz [Tue, 30 Oct 2007 01:19:17 +0000 (01:19 +0000)]
An iterator over attributes, as GCC scalarizes it the code is equivalent
to hand-written code that doesn't use a structure:
attr_iterator ai;
FOR_ATTRS (store, entry, &ai)
{
do_something (ai.name, ai.type);
switch (ai.type)
{
case ATTR_STRING: ai.as_string; break;
case ATTR_INT: ai.as_int; break;
case ATTR_LOCALIDS: iterate (ai.as_numlist); break;
}
}
Michael Matz [Mon, 29 Oct 2007 22:22:59 +0000 (22:22 +0000)]
Write and read the read-only representation directly.
Michael Schroeder [Mon, 29 Oct 2007 19:02:15 +0000 (19:02 +0000)]
- fix last failing testcase classified as bug
Michael Schroeder [Mon, 29 Oct 2007 17:25:45 +0000 (17:25 +0000)]
Use Offset instead of unsigned int.
Michael Schroeder [Mon, 29 Oct 2007 16:45:27 +0000 (16:45 +0000)]
- rename 'system' repo to 'installed'
Michael Schroeder [Mon, 29 Oct 2007 15:00:30 +0000 (15:00 +0000)]
- rename queue* to queue_*, inline a bit more
- rename map* to map_*
Klaus Kaempf [Mon, 29 Oct 2007 12:05:53 +0000 (12:05 +0000)]
keep 'bool'ness in the comment
Michael Schroeder [Mon, 29 Oct 2007 11:20:32 +0000 (11:20 +0000)]
- cleanup strange code
Michael Matz [Mon, 29 Oct 2007 02:39:38 +0000 (02:39 +0000)]
A DWARF like in-memory representation (with abbreviations), which uses
only a quarter of memory for the Attrs itself (i.e. without the string
spaces).
Michael Matz [Sun, 28 Oct 2007 23:09:45 +0000 (23:09 +0000)]
Start of an attribute store. It can hold arbitrary attributes for a
number of entities. When those are build as mirroring the solvables in
a Repo, we can attach non-solver information to them. Not yet really
integrated, uses too much in-core memory, and arbitrary blobs are not
yet supported (for descriptions and maybe large author lists).
Stephan Kulow [Sun, 28 Oct 2007 19:06:04 +0000 (19:06 +0000)]
make it debugable - use -DCMAKE_BUILD_TYPE=debug in the cmake call
Stephan Kulow [Sun, 28 Oct 2007 19:05:52 +0000 (19:05 +0000)]
continue compiling with -Werror
Stephan Kulow [Sun, 28 Oct 2007 19:05:49 +0000 (19:05 +0000)]
don't play with pointers
Michael Schroeder [Fri, 26 Oct 2007 23:01:13 +0000 (23:01 +0000)]
- no more bool, too many conflicts
- moved declaration of pool_freerepo to repo.h
Michael Schroeder [Fri, 26 Oct 2007 20:42:46 +0000 (20:42 +0000)]
- move dep2str buffer into pool so that the memory gets released
- directly install translation solvables
- some minor code cleanups
Duncan Mac-Vicar P [Fri, 26 Oct 2007 20:25:13 +0000 (20:25 +0000)]
Fix the ruby example, and remove a obsolete one
Duncan Mac-Vicar P [Fri, 26 Oct 2007 20:25:08 +0000 (20:25 +0000)]
to follow SUSE tools conventions
Source is now Repo
Stephan Kulow [Fri, 26 Oct 2007 19:08:40 +0000 (19:08 +0000)]
fix memory leaks
Duncan Mac-Vicar P [Fri, 26 Oct 2007 16:50:05 +0000 (16:50 +0000)]
- install the libraries
Duncan Mac-Vicar P [Fri, 26 Oct 2007 14:35:30 +0000 (14:35 +0000)]
this is not in old cmake (10.2) add it so it works for everyone
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:26 +0000 (13:57 +0000)]
bye autofoo
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:24 +0000 (13:57 +0000)]
- iterators for solver decisions
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:22 +0000 (13:57 +0000)]
- fix packaging
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:21 +0000 (13:57 +0000)]
examples
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:18 +0000 (13:57 +0000)]
revert debug
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:16 +0000 (13:57 +0000)]
name the enum so I can use it in a typemap
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:13 +0000 (13:57 +0000)]
each_solvable in Source almost working
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:11 +0000 (13:57 +0000)]
Find libraries, and wrap the ruby native module into a ruby file
Duncan Mac-Vicar P [Fri, 26 Oct 2007 13:57:08 +0000 (13:57 +0000)]
- add cmake support
- add ruby bindings
Michael Schroeder [Fri, 26 Oct 2007 11:00:22 +0000 (11:00 +0000)]
treat 'nothing provides' jobrules in a sane way
Michael Schroeder [Thu, 25 Oct 2007 17:18:26 +0000 (17:18 +0000)]
- do not add rules that are always fulfilled
Michael Schroeder [Thu, 25 Oct 2007 16:42:16 +0000 (16:42 +0000)]
- add support for namespace callbacks
- implement rpmlib test in relation expansion
- implement modalias() globbing
Michael Schroeder [Thu, 25 Oct 2007 14:23:28 +0000 (14:23 +0000)]
- make suggests/enhances first class, needs more rules but it simplifies
the code and the result is more reliable
Michael Schroeder [Wed, 24 Oct 2007 22:01:26 +0000 (22:01 +0000)]
- do a better job at presenting a solution
- implement source priorities, first half
Michael Schroeder [Wed, 24 Oct 2007 10:46:09 +0000 (10:46 +0000)]
- goodbye deps, we no longer need you
Michael Schroeder [Wed, 24 Oct 2007 09:59:42 +0000 (09:59 +0000)]
- zero first element in idarray just in case
- make runtest compare with solution2 and solution3
- do not add same language solvables multiple times, fixes
one test case
Michael Matz [Tue, 23 Oct 2007 17:57:17 +0000 (17:57 +0000)]
No need for testing null-ness anymore.
Stephan Kulow [Tue, 23 Oct 2007 17:13:50 +0000 (17:13 +0000)]
fix compile
Michael Schroeder [Tue, 23 Oct 2007 15:24:36 +0000 (15:24 +0000)]
- switch dependencies from Id pointers to Offsets
Michael Schroeder [Tue, 23 Oct 2007 12:40:00 +0000 (12:40 +0000)]
- map locale(xx) to language:xx to be compatible to libzypp
- treat system:modalias() as modalias()
Stephan Kulow [Tue, 23 Oct 2007 08:07:55 +0000 (08:07 +0000)]
fix compile
Michael Schroeder [Mon, 22 Oct 2007 18:50:33 +0000 (18:50 +0000)]
- fix idarray overflow checking
- do not create weak rule for packages scheduled for removal
(fixes one testcase)
- fix some explanations in README.FAILS
Michael Schroeder [Mon, 22 Oct 2007 18:01:35 +0000 (18:01 +0000)]
- first step in getting rid of dependency pointers:
store back pointer to source in Solvable.
Klaus Kaempf [Mon, 22 Oct 2007 15:25:04 +0000 (15:25 +0000)]
make it compile on 64bit
Michael Schroeder [Mon, 22 Oct 2007 14:39:00 +0000 (14:39 +0000)]
- add experimental weak rules support, solver tries a bit harder to
fulfill them
Michael Schroeder [Mon, 22 Oct 2007 10:19:55 +0000 (10:19 +0000)]
- beautify a bit
Stephan Kulow [Sun, 21 Oct 2007 07:35:23 +0000 (07:35 +0000)]
fixed mean bug in an optimization of addrule:
Differ between binary rules and non-binary rules
with one literal (e.g. two kernels)
(fixed 5 test cases)
Michael Matz [Sat, 20 Oct 2007 16:43:57 +0000 (16:43 +0000)]
Now that I can read content files, let's use that.
Michael Matz [Sat, 20 Oct 2007 16:32:08 +0000 (16:32 +0000)]
content2solv: a parser for the content file of Yast2 repos. Generates
a SOLV with only one solvable, the product.
Michael Matz [Sat, 20 Oct 2007 16:30:36 +0000 (16:30 +0000)]
I need C++ here too.
Stephan Kulow [Fri, 19 Oct 2007 21:01:30 +0000 (21:01 +0000)]
don't crash too often
Stephan Kulow [Fri, 19 Oct 2007 20:39:17 +0000 (20:39 +0000)]
trying to mimic the libzypp handling of kernel deps - first step
Michael Schroeder [Fri, 19 Oct 2007 18:12:25 +0000 (18:12 +0000)]
- clean up the code a bit
Michael Schroeder [Fri, 19 Oct 2007 16:34:41 +0000 (16:34 +0000)]
- do away with complex reenablerule, instead reset propagate_index to 0
and rely on propagate to move the watches
Michael Schroeder [Fri, 19 Oct 2007 14:05:14 +0000 (14:05 +0000)]
- fix some bugs in refine_suggestion()
Michael Schroeder [Wed, 17 Oct 2007 16:30:58 +0000 (16:30 +0000)]
move solver reset to top of refine_suggestion
Michael Schroeder [Wed, 17 Oct 2007 15:59:55 +0000 (15:59 +0000)]
- added pool_installable() inline function
- renamed addrelproviders() to pool_addrelproviders()
- also use dep_possible() in enhances
- print statistics about solvable reduction
- check is suggested packages are really installable
Michael Schroeder [Wed, 17 Oct 2007 13:19:20 +0000 (13:19 +0000)]
- inline source_name function
- fix dependency resolving loop to restart on any action
- fix supplement rule creation, it did not work for complex
(i.e. AND) supplements
Michael Schroeder [Wed, 17 Oct 2007 09:50:31 +0000 (09:50 +0000)]
unify whatprovides data to save memory.
Michael Schroeder [Mon, 15 Oct 2007 16:54:08 +0000 (16:54 +0000)]
find suggested packages
Michael Schroeder [Mon, 15 Oct 2007 15:06:01 +0000 (15:06 +0000)]
add arch filtering for obsoletes data
Michael Schroeder [Mon, 15 Oct 2007 15:00:45 +0000 (15:00 +0000)]
implemented GNADENLOS in a better way, now option "noupdateprovide".
Michael Schroeder [Mon, 15 Oct 2007 13:25:30 +0000 (13:25 +0000)]
also prune to suggests/enhances, fixes tpctl testcases
Michael Schroeder [Mon, 15 Oct 2007 11:06:02 +0000 (11:06 +0000)]
make prune_to_recommended look at recommends, too.
Michael Matz [Mon, 15 Oct 2007 02:58:32 +0000 (02:58 +0000)]
Simple shell script accepting a directory with local repository data,
detecting which type the repo is, and emitting exactly one SOLV file
for the whole thing.
Michael Matz [Mon, 15 Oct 2007 02:57:27 +0000 (02:57 +0000)]
To be able to accept a set of patches simply cat'ed behind each other
we need to accept a collector tag. I chose '<patches>', so you can
parse multiple patches by:
( echo '<patches>'; cat patch-*.xml; echo '</patches>' )
| grep -v '\?xml'
| patchxml2solv
Michael Matz [Mon, 15 Oct 2007 02:55:00 +0000 (02:55 +0000)]
Blaeh. LIBS doesn't provide DEPENDENCIES, hence we don't to use that,
but LDADD is overwritten (instead auf augmented) by specific X_LDADDS,
so we need to add LDADD everywhere explicitely.
Michael Matz [Mon, 15 Oct 2007 01:59:12 +0000 (01:59 +0000)]
I have use for a tool to merge .solv files.
Stephan Kulow [Sun, 14 Oct 2007 12:13:03 +0000 (12:13 +0000)]
parse <product> out of helix sources
Michael Matz [Fri, 12 Oct 2007 15:44:30 +0000 (15:44 +0000)]
Make the satsolver header usable with C++ (only pool.h and source_solv.h
for now, they drag in most of the usefull stuff anyway).
Stephan Kulow [Fri, 12 Oct 2007 15:14:34 +0000 (15:14 +0000)]
- fixing obsoletes not checked in every case (fixing one test case -
mls)
- adding a GNADENLOS mode not set by default (fixing 6 more, but
degrades performance by 28%)
Michael Matz [Fri, 12 Oct 2007 15:08:28 +0000 (15:08 +0000)]
The .spec file also referred to the shared lib. Fix that.
Michael Matz [Fri, 12 Oct 2007 15:07:26 +0000 (15:07 +0000)]
Enable maintainer-mode (so that Makefiles are automatically rebuilt if
Makefile.am changes).
Michael Matz [Fri, 12 Oct 2007 14:58:08 +0000 (14:58 +0000)]
For now only build a static (but PIC) libsatsolver.a. No sense in doing
a shared lib with an unstable API.
Michael Schroeder [Fri, 12 Oct 2007 14:50:32 +0000 (14:50 +0000)]
fix typo
Michael Schroeder [Fri, 12 Oct 2007 14:36:47 +0000 (14:36 +0000)]
fix makefiles a bit
Michael Schroeder [Fri, 12 Oct 2007 14:08:29 +0000 (14:08 +0000)]
delete some spaces
Stephan Kulow [Fri, 12 Oct 2007 11:45:43 +0000 (11:45 +0000)]
fixing some bugs:
- remove the original package on upgrade, not the new
one
- always set i686 when no arch is set (as required)
- handle "install" as capability not as name
- handle different kind of solvables in install
- install locale solvables into an extra channel
Michael Schroeder [Tue, 9 Oct 2007 15:47:41 +0000 (15:47 +0000)]
- do not select src packages in yps
- use some buffers to speed things up a bit
Michael Schroeder [Tue, 9 Oct 2007 15:33:44 +0000 (15:33 +0000)]
- show better problem reports
Michael Schroeder [Tue, 9 Oct 2007 10:51:18 +0000 (10:51 +0000)]
- fix arch handling
- fix makefile
- add modalias and installed namespaces
- fix red carpet upgrade detection
Michael Schroeder [Mon, 8 Oct 2007 09:17:33 +0000 (09:17 +0000)]
fix segfault
Michael Schroeder [Fri, 5 Oct 2007 23:19:53 +0000 (23:19 +0000)]
- don't pass name spaces for now
- don't supplement bad architectures
Michael Schroeder [Fri, 5 Oct 2007 22:53:16 +0000 (22:53 +0000)]
- support for legacy rewriting
- support for and/or/with deps
- support freshens
Michael Schroeder [Fri, 5 Oct 2007 17:24:41 +0000 (17:24 +0000)]
do not add self-provides to src rpms
Michael Schroeder [Fri, 5 Oct 2007 14:07:58 +0000 (14:07 +0000)]
fix typo, add some more architectures
Klaus Kaempf [Tue, 2 Oct 2007 14:54:29 +0000 (14:54 +0000)]
prep for package
Klaus Kaempf [Tue, 2 Oct 2007 12:40:07 +0000 (12:40 +0000)]
current state of 'sat-solver'