Summary: 229 instances, 211 unique

Text	Count
* TODO: remove non-editable CFG option	1
// TODO: factor this code a little	1
// TODO: There is a potential register allocation issue when we instrument	1
// TODO: Don't mattern match at all, but make this a deep semantic	1
// TODO: Consider throwing a custom exception, possibly created by code	1
* TODO: detect changes and recompute when necessary.	1
// TODO: Implement	1
// TODO: this is horrible. After we remove methods, we shouldn't	1
// TODO: Make this an error once all existing configurations have been	1
// TODO: Is this really safe for all virtual/interface invokes? This	1
// FIXME: This isn't deterministic. We're keeping whichever class we loaded	1
// TODO: merge this and MethodSplicer.	1
// TODO: It's not worth the goto to merge return-only blocks. What size is	1
// TODO: replace this with an annotation.	1
// TODO maybe later? If we expand to non-clinit methods, `analyze()`	1
// TODO: check for unterminated markers	1
// TODO: The test below will fail for Win32 paths. Fix later or ignore.	1
// TODO: T31388603 -- Remove unused results for true virtuals.	1
// TODO: Is it possible to recover a valid state of iterators into the CFG	1
// TODO: Handle floats and doubles via Float.floatToIntBits and	1
// TODO: Consider making those methods static if they can be renamed,	1
*    of visibility but could be made static (FIXME: This flag currently seems	1
// TODO: Support load-param instruction in blocks other than the entry	1
// TODO: We should probably have an API to offer iterators into the blocks map	1
* TODO: Add a way to define a real store.	1
// @TODO - Instead of keeping track of just one group, in the future we can	1
// @TODO - Instead of only keeping one group and calculate best savings	1
* TODO (anwangster) Is method-method opt data needed?	1
// TODO: Currently not able to merge classes that only extend	1
// TODO: If we split by interaction, we could check here specifically.	1
* TODO:	1
// TODO t19478845	1
// TODO: detect the type of constant for each alias group	1
// TODO: tail call optimization (if callsite_split is a return & inline_after)	1
// TODO: Try some other variations.	2
// TODO: Instead of custom fixpoint computation using WTO, consider using the	1
// TODO: "Linking" error: Change this to an assertion failure	1
// TODO: Enable the following flag. It's off for now to ensures that the	1
* TODO: Some split blocks might not actually get dedup, as deduping checks	1
// TODO: allow static initializers that have other unrelated code	1
// TODO: inheritance?	1
// TODO: Collapse equivalent source blocks?	1
// TODO: The following loops are probably the most expensive thing in this	1
// TODO: Support ranges in ConstantPropagation instead of this one-off pass.	1
// TODO: Merge candidates that are equivalent except that one returns	1
# TODO: support 64-bit	1
// TODO: 064 hash table is different!	1
// TODO: Consider splitting all in-edges of the canonical block and	1
// TODO: Weaking across enums is technically correct, but exposes a	2
// TODO: Can merge named classes.	1
// TODO: make naming of methods smart	1
// TODO: Improve TypeInference so that we never have to deal with	1
// TODO: Make the set of excluded rules configurable.	1
"TODO: Document me!",	1
// TODO: Per-interaction stats.	1
* TODO: cover other branch type for constant values.	1
* TODO: have more robust support for above 1 and 2	2
// TODO: If the successors of the successors do not contain	1
// TODO: Share `m_config` with `optimize_enums::replace_enum_with_int`	1
* TODO: Deprecate this method and use the new version instead.	1
// TODO: Instead of just dropping all blocks in this case, do finer-grained	2
// TODO: Maybe do this?	1
//    TODO: also do a similar step for type strings pool, and any empty type	1
// TODO: deep-inspect this, but doesn't seem to occur in practice	1
// TODO: Investigate error P444184021 when merging simple classes without	1
// TODO: size estimate	2
// TODO: Investigate why. Until then, disable that pass to test CSE.	1
// FIXME: Copying the outgoing throw edges isn't enough.	1
// TODO: Give up if there's exception handling.	1
// TODO: add std::string API for contains	1
// TODO: make it private and find a better way to allow code creation	1
// TODO: Does the above need to filter? Do we need to transitively close	1
// TODO: There must be a logarithmic approach to this. dmethods are sorted!	1
// TODO: Forwarding may leave behind trivial conditional branches that can	1
// TODO: Figure out a most general way of doing this.	1
// TODO: Support capturing pre-state values	1
// TODO: Deal with duplication.	1
// TODO There are elements of this is used in InstructionSequenceOutliner and	1
// TODO: This normalization optimization doesn't really belong to local-dce,	1
// TODO: Extend this (with Proguard parsing) to support the following:	1
// TODO: Remove the XML-related keep reasons	1
// TODO: Should we undo this when we are done writing the map?	1
// TODO: Find true root cause, and make this exception more targetted.	1
# TODO printing IRCode seems to crash gdb	1
// TODO: Revisit.	1
// TODO: Do better, use less. A subset and its superset can share a temp.	1
// TODO: Bridgee won't necessarily be direct once we expand this	1
* TODO: The complete solution requires some kind of call graph analysis from	1
// TODO: I saw that most often, when multiple object type demands	1
// TODO: make order based on field position if a proper annotation is	1
* TODO (anwangster) Is class-class opt data needed?	1
// TODO: We also likely need a move if there are more than 16 args	1
// TODO: Encapsulate (primary|all) dex flags under one config.	1
// TODO: Can we be even more lenient without actually deduping and	1
// TODO: Try harder.	1
// TODO: Handle not just conditional trees, but DAGs or even arbitrary	1
// TODO: (in the future) add spill here given dex constraints	1
// TODO (cnli): put back 'always_assert(m_parent->editable());'	1
// TODO: we could be redexing object some day... :)	1
* - Other minor TODO ideas in the code	1
* TODO: The method is deprecated since it doesn't delete the edges associated	1
// FIXME: I don't think this handles wides correctly	1
// TODO: Explore other (variations of this) heuristics.	1
// TODO: Is this really necessary? If so, fill in correctly.	1
// TODO (T91001948): Integrate protobuf dependency in supported platforms for	3
// TODO: Positions should be connected to instructions rather than preceding	1
// 4. Will not have any fields (TODO We could extend to support sfields)	1
// TODO: Support properly.	1
// FIXME assert if already concrete	1
// TODO: probably want the real name here.	1
// TODO: Check the orderings.	1
// TODO: Relax this. Some portions of those virtual scopes could still	1
// TODO: make sure names reported handles collisions correctly.	1
// TODO: Sort the outputs by edge type, case key, and throw index	1
* TODO: Run constant-propagation in caller, and then do callsite-specific	1
// TODO: this is a hack to fix a problem with enclosing methods only.	1
// TODO: If a whole cluster does not have IODI, we should emit base	1
// TODO: revisit for methods to be careful around Object	1
// TODO: What about no exit blocks possibly due to infinite loops? Such case	1
// TODO: Handle compiled classes. Need to read method bitmap	1
// TODO: Also check that reduced_cfg's are equivalent	1
// TODO: do we need to handle endian-ness? I think all platforms we	1
// TODO: fix the optimizations that don't track registers size correctly.	1
// TODO: For better human readability, change this to CSV of native,java?	1
* TODO: We can assume no side effect for more cases, like if it only accesses	1
// TODO: Do some careful widening.	1
// TODO: do we need to re-check linear_alloc_limit?	1
// TODO: revisit this for multiple callee call graph.	1
// TODO: Support side-effect-free instruction sequences involving	1
// TODO: walk through annotations	2
* TODO: someday, this should probably be subsumed by a more general purpose	1
// TODO: account for "close" defs and uses. See [Briggs92], section 8.7	1
* TODO: we should be able to create a new FileMap that is a subset of	1
// TODO: Analyze other common opcodes, and try to support them.	1
// TODO: More complicated version of the above.	1
// TODO: Remove this limitation imposed by symbolication infrastructure.	1
// TODO we should coalesce here too, but we'll need to avoid removing	1
static constexpr const char* default_doc() { return "TODO: Document this"; }	1
* TODO: phase out edits to the IRCode and move them all to the editable CFG	1
// TODO: filter out length zero/one states?	1
// TODO: Instead of this custom meet function, the ConstantValue should get a	2
// TODO: SUB to RSUB	2
// TODO: Consider using the SummaryGenerator to analyze AOSP classes to	1
m_external_side_effect_summaries_file, "TODO: Document me!",	1
// TODO In case of invokes, we want to show the deobfuscated name for clarity,	1
// TODO: Remove this limitation imposed by symbolication infrastructure.	1
// TODO: There are still blocking issues in instrumentation test that are	1
* TODO: It now requires inliner to inline the merge virtual methods to	1
// TODO: Are we doing something about abstract methods without any overrides	1
// TODO: We could order them instead of looping ...	1
// TODO (cnli): use editable CFG and update insert logic.	1
// TODO: Try some other variations.	1
// TODO: Make this logic cleaner when these features get enabled for non-root	1
ss << "<data>"; // TODO: print something more informative	1
// TODO: Could be done with trampolines if we remove "final" flag from	2
// TODO: If the existing continuitions are not move-results, and the	1
// TODO: look at the abstract methods and we can lift the	1
// TODO: throws?	1
// TODO: wrapping this in json wrapper is stupid, can we use raw	1
// TODO: Assert that this never happens. IOPCODE_INIT_CLASS should never make	1
// TODO: Consider making this configurable. It represents the maximum number	1
// TODO: Handle compiled classes. Need to read method bitmap size,	1
// TODO: Remove unused profiled_methods_file option and all build system	1
// TODO:	1
// TODO: Instead of a user-defined limit, or over-approximating by default,	1
// TODO: Any integrity checks should really be in the DexOpcodeData	1
// TODO: relax the constraint that move-result-pseudo must be	1
// TODO: Consider folding the above attempts to reduce the types set	1
// TODO: This is probably the most memory hungry part of the whole building	1
// T58562665: TODO - actually update states for callsites/methodhandles	1
// TODO: We may be able to support dynamic references in overlays	1
// FIXME: this algorithm ignores real infinite loops in the block graph	1
* TODO: Derive this list with static analysis rather than hard-coding	1
* TODO: this probably need to rely on metadata to be correct. Right now it	1
// TODO: annotations?	3
// TODO: Rewrite open/write/close with C/C++ standards. But it works for now.	1
// FIXME: fix cfg's GraphInterface so this const_cast isn't necessary	1
// TODO: rewrite the following code to not require a random-access	1
* TODO: Add useful CFG editing methods	1
// TODO delete these	1
// TODO: Should we allow static finals that are initialized w/ const,	1
// TODO: more fixup here on this crapped up name/signature	2
// TODO: Refactor so that we don't have to list every single possible	1
// TODO: This needs to be adapted to sum up the predecessors.	1
// TODO: Consider using an annotation-based scheme.	1
// TODO: we should be able to relax the check above to be simply	1
* TODO: Consider integrating the full visibility check in	1
// TODO: string usage	2
// TODO: erase fopcode from map	1
// TODO: Isn't that the same as before, just this time correct w/ counting?	1
// TODO: Allow this here, but also insert a check-cast instead of a simple	1
// TODO: Need a better interface.	1
// TODO: Remove classes from scope that are exempt from checking.	1
// TODO: Make the command line -jarpath option like a colon separated	1
// TODO: Track dependencies at more fine-grained level, and use that	1
// FIXME: "Empty" blocks with only `DexPosition`s should be merged	1
# TODO don't know why we get errors like:	1
// TODO: Switch to a proper interprocedural fixpoint analysis.	1
// TODO: Use type inference to figure out required array (component) type.	1
// TODO: deep-inspect array, but likely little impact	1
// TODO: string usage (encoded_value for static fields)	1
// TODO: have load param opcodes store the actual type of the param and	1
// TODO: T31388603 -- Remove unused args for true virtuals.	3
// TODO: Support interface callees.	1
// TODO: Keep around all store dependencies and reuse when possible	1
// TODO: This code uses 0xFFFFFFFF converted to bag_set* as a sentinel value. This is bad practice.	1
// to the parent's active catch handler. TODO: Make this more precise	1
"TODO: Upgrade RuntimeAssertTransform to use cfg.");	1
// TODO: break down signature	2
* TODO: Once we have points-to information, we should expand the callgraph	1
// TODO: T62185151 - better way of applying this on CFGs	1