| Age | Commit message (Collapse) | Author | Files | Lines | |
|---|---|---|---|---|---|
| 2010-01-30 | Add a first draft of a class to represent *operations* on tiers. | Daniel Burrows | 3 | -2/+141 | |
| It's a bit awkward that it uses tiers to represent its components; on the other hand, its components are *almost* exactly tiers. Something to ponder. | |||||
| 2010-01-29 | Allow any level, even one that would be out-of-bounds, to be set via ↵ | Daniel Burrows | 1 | -2/+17 | |
| set_user_level(). Any intervening levels that are added as a result of this operation are initialized to tier_limits::minimum_tier. | |||||
| 2010-01-28 | Add tier.cc to version control. | Daniel Burrows | 1 | -0/+60 | |
| 2010-01-28 | Add a convenience method on tiers to retrieve the number of user levels. | Daniel Burrows | 1 | -0/+3 | |
| 2010-01-27 | Stop using flyweights to store tiers, since they tend to cause crashes on ↵ | Daniel Burrows | 1 | -33/+31 | |
| shutdown. :-( | |||||
| 2010-01-26 | Write a public method on tiers to retrieve individual user levels. | Daniel Burrows | 1 | -0/+3 | |
| 2010-01-26 | Since the tier::set_*() methods are non-mutating, they can and should be const. | Daniel Burrows | 1 | -2/+2 | |
| 2010-01-26 | Rename set_user_tier to set_user_level, in keeping with the earlier ↵ | Daniel Burrows | 1 | -1/+1 | |
| terminology change. | |||||
| 2010-01-26 | Add the new .cc files to the build. | Daniel Burrows | 1 | -1/+3 | |
| 2010-01-23 | Declare the operator<< override on tiers. | Daniel Burrows | 1 | -0/+2 | |
| 2010-01-23 | Put the hash_value() overload in the global namespace. | Daniel Burrows | 1 | -5/+2 | |
| 2010-01-23 | Implement comparison operators on tiers. | Daniel Burrows | 1 | -0/+15 | |
| 2010-01-23 | Write an assignment operator for tiers. | Daniel Burrows | 1 | -0/+7 | |
| 2010-01-23 | Give tiers public constructors instead of static create() routines. | Daniel Burrows | 1 | -10/+8 | |
| 2010-01-23 | Write a default constructor for tiers. | Daniel Burrows | 1 | -0/+7 | |
| 2010-01-23 | Rewire the rest of the dependency solver to account for the new tier type. | Daniel Burrows | 8 | -198/+56 | |
| Mostly this just removes typedefs from the universe-specific tier types and adds #includes for tier.h. The tier limits object was rewritten somewhat more, to account for the new terminology ("levels") and to make the integers explciitly represent structural levels, not tiers. | |||||
| 2010-01-21 | Implemment operator== and operator!= on tier-impls, and make the definitions ↵ | Daniel Burrows | 1 | -12/+16 | |
| in tier use the flyweight versions of these operators for speed. | |||||
| 2010-01-21 | Declare the create() methods to be static. | Daniel Burrows | 1 | -4/+5 | |
| 2010-01-21 | Fix get_hash_value() on tiers to use the hashed value of the referenced impl ↵ | Daniel Burrows | 1 | -1/+1 | |
| object. | |||||
| 2010-01-21 | Instead of computing a tier's hash whenever it's needed, cache it and just ↵ | Daniel Burrows | 1 | -12/+28 | |
| use the cached value. | |||||
| 2010-01-21 | The set_* routines should return tiers, not flyweight<tier>s. | Daniel Burrows | 1 | -2/+2 | |
| 2010-01-21 | Add a missing "return" statement. | Daniel Burrows | 1 | -1/+1 | |
| 2010-01-21 | Fix the return type of get_impl() (was "impl", should be "tier_impl"). | Daniel Burrows | 1 | -1/+1 | |
| 2010-01-21 | Change how the "return a modified tier" routines are implemented. | Daniel Burrows | 1 | -13/+14 | |
| The structural level can be changed with the normal impl constructor, so the special constructor now only deals with user levels (fixing an off-by-one error in the process). | |||||
| 2010-01-21 | Don't template the tier-from-structural-level constructor. | Daniel Burrows | 1 | -1/+0 | |
| 2010-01-21 | Reduce terminological confusion by describing the components of a tier as ↵ | Daniel Burrows | 1 | -80/+102 | |
| "levels". Not sure this is the best term, but it's a lot better than describing the components of a tier as "tiers". | |||||
| 2010-01-20 | Add an initial definition of a more flexible, generic tier object. | Daniel Burrows | 2 | -1/+285 | |
| The idea here is to support tiers that are dynamically defined by settings, so the user can (for instance) override the default ordering. | |||||
| 2009-11-17 | Have the resolver ignore unresolved dependencies that were unresolved before ↵ | Daniel Burrows | 4 | -16/+80 | |
| the current run of aptitude started. (Closes: #556042) This is done by filtering them out of the initial set of dependencies that the root node is initialized with. NOTE: I still need to convince myself that we can't somehow end up reactivating a dependency that should have been skipped. However, we might really need to legitimately activate it...something to think over, anyway. (I particularly suspect that the weird behavior of Conflicts might be a problem) | |||||
| 2009-09-09 | Don't crash when dumping information about a solver without a validity ↵ | Daniel Burrows | 1 | -1/+5 | |
| condition. | |||||
| 2009-07-28 | Always use make_shared to construct shared pointers. | Daniel Burrows | 1 | -1/+2 | |
| 2009-07-15 | Don't waste time building information about a solver set if nothing would be ↵ | Daniel Burrows | 1 | -10/+33 | |
| done with it. If the tier of the solver set is not high enough for us to extract any useful information, there's no point in calculating the reasons for that tier. | |||||
| 2009-07-11 | Fix compile problems introduced by the backout. | Daniel Burrows | 1 | -12/+63 | |
| 2009-07-10 | Merge with backout. | Daniel Burrows | 3 | -37/+294 | |
| 2009-07-10 | Eliminating the use of hash tables to save memory: backed out changeset ↵ | Daniel Burrows | 4 | -169/+196 | |
| 7c43f786ebba | |||||
| 2009-07-10 | Merge backout. | Daniel Burrows | 3 | -142/+105 | |
| 2009-07-10 | Eliminating the use of hash tables to save memory: backed out changeset ↵ | Daniel Burrows | 3 | -142/+105 | |
| c8aec0d44f84 | |||||
| 2009-07-10 | Merge backout. | Daniel Burrows | 1 | -11/+82 | |
| 2009-07-10 | Eliminating the use of hash tables to save memory: backed out changeset ↵ | Daniel Burrows | 1 | -60/+45 | |
| 8ff97afb1605 | |||||
| 2009-07-10 | Eliminating the use of hash tables to save memory: backed out changeset ↵ | Daniel Burrows | 2 | -62/+47 | |
| 13ac768ed385 | |||||
| 2009-07-07 | Switch the unresolved dep set representation to a boost::unordered_map. | Daniel Burrows | 2 | -47/+62 | |
| 2009-07-07 | When increasing the tier of a solver that's newly deferred, only increase ↵ | Daniel Burrows | 1 | -7/+9 | |
| exact matches (including matches to the dependency). Necessary because deferrals due to violating an approval constraint only apply when a version is chosen to solve the same dependency. This is just one other way that approval constraints are wonky. I'm thinking that it might be good on multiple levels to throw them out: they're a bit counterintuitive, they tend to produce unstable results (the set of legal successors can vary depending on which dependency is chosen as the next one to resolve, for instance), and because they mean that deferrals happen on a per-dependency basis, they need a lot of careful code and a lot of extra memory allocation to get right. I'm not sure what would replace them, though. | |||||
| 2009-07-07 | Fix recompute_step_tier() to correctly compute the tier. | Daniel Burrows | 1 | -4/+73 | |
| It wasn't taking into account the actions themselves, or promotions that only included them. This fixes the big glaring problem with the new code, namely that it was spinning off indefinitely (because depending on the order in which some normally idempotent things were applied, they would trip this bug and miss the fact that a line of reasoning was supposed to be chopped off). There's still one odd problem, where I get different results each time I run it, probably because of some other order dependency. | |||||
| 2009-07-07 | Fix the sanity-check stuff to account for recent changes to the resolver code. | Daniel Burrows | 1 | -45/+60 | |
| 2009-07-07 | Sort hashed sets of choices when displaying them, so log files are more ↵ | Daniel Burrows | 3 | -105/+142 | |
| comparable across runs. Not every set of choices is sorted yet; there might be a few more of these to write. | |||||
| 2009-07-07 | Log the semantic content of deferral/validity expressions rather than object ↵ | Daniel Burrows | 2 | -18/+166 | |
| pointers. This should make log files a lot more reproducible. | |||||
| 2009-07-07 | Make the constructor of var_e protected so that it can be subclassed. | Daniel Burrows | 1 | -0/+1 | |
| 2009-07-07 | Add an extra sanity-check that all solvers have the right associated ↵ | Daniel Burrows | 1 | -0/+3 | |
| dependencies. | |||||
| 2009-07-06 | Hopefully fix how the number of deferred steps is counted. | Daniel Burrows | 1 | -0/+4 | |
| 2009-07-06 | Put every step, even deferred ones, into "pending". | Daniel Burrows | 1 | -2/+1 | |
| Necessary for the sake of correctness: otherwise, set_step_tier doesn't know whether a step is "closed" and hence whether it should be put back on the pending list. An alternative would be to have an explicit flag on a step that gets set when it's "closed"; then we could throw deferred steps out of the pending tier entirely, which would be nice and would save a little time. | |||||
| 2009-07-06 | When recomputing a step tier from scratch, really recompute it from scratch. | Daniel Burrows | 1 | -1/+7 | |
| Two bugs here: it wasn't being set back to the minimum tier, and it was being set directly rather than via set_step_tier(). | |||||
 |
index : aptitude | |
| [no description] | Dyson |
| summaryrefslogtreecommitdiff |