Support emits of structs; test parsing of varbit fields #634
Conversation
| auto vec = new IR::Vector<IR::StatOrDecl>(); | ||
| auto it = expansionTypes.begin(); | ||
| for (auto e : expansion) { | ||
| auto method = statement->methodCall->method->clone(); |
There was a problem hiding this comment.
Why clone the method if you're not changing it?
There was a problem hiding this comment.
These must be different IR nodes, otherwise all analyses which compute types, def-use, etc. will produce wrong results - the results are usually indexed in a hash-table by IR node.
There was a problem hiding this comment.
A bit worried we may actually need a deeper clone.
There was a problem hiding this comment.
This probably indicates that we should fix those analyses to work properly in the presence of dags -- just storing pointers to IR nodes is not adequate if it also depends on the context of those nodes.
There was a problem hiding this comment.
It is really easiest to have nodes be distinct rather than look at the context.
The question is whether these analyses should establish this invariant by themselves. Is setting visitDagOnce = false enough for an analysis to duplicate all nodes?
* Fix for issue #638 * IRGen fixes - support for autogenerating methods in nested classes - support for friend functions - misc bug fixes * cpplint fixes * Update comment * Minor change in bmv2 backend: set action_const (#652) In the bmv2 JSON, a match-table can indicate that its default action is constant (using 'action_const') and that the default action entry (action + action parameter values) is constant (using 'action_entry_const'). `action_entry_const == true => action_const == true`. The bmv2 backend was setting 'action_entry_const' correctly, but 'action_const' was always set to false. Note that P4_16 does not make a difference between the 2, the presence of `const` in the P4_16 source implies that both attributes should be true. * Fix for issue #629 (#636) * Support emits of structs; test parsing of varbit fields (#634) * Support emits of structs; test parsing of varbit fields * Fixes issue #447 and #630 * Updated reference outputs * Additional tests and bug-fixes for varbits * Fix for issue #638 * Allow empty actions list in tables
This does not solve issue #447 completely, only partially.
More fixes will be necessary.
It should solve issue #630 completely, though.