16:29:35karloszi think separating references from variables at the ast-level is pretty orthogonal to the issue of map-ast having to keep a hash table
16:29:51karloszbut we should probably do it anyway for better policy and source like your old commit said
16:30:27karloszall we really need to do to make map-ast not use the hash table is to make sure we don't visit any child more than once, which we can do by controlling the children method
16:30:53karloszi.e. only function-ast and lexical-bind-ast have lexical-ast/lexical-variable as children
18:05:23karloszi guess its pretty easy to solve the problem of unnecessarily making new conses just by checking whether the car and cdr remain the same
18:09:06karloszBike: could you merge the bignum dumping pr? i have clasp side changes queued up for the ast munging
18:31:22karloszan interesting not about the ast interpreter - we don't actually need a linked list of frames, since the environment has already disappeared by the time we converted to an ast, so just a flat mapping will do
23:31:21Bikeoh, so it's doing something nonconforming
23:32:26Bikei suppose the fact it treats things that can be negative as size_t might cause weird problems
23:33:41Bikeltvc_read_bignum uses compact_read_size_t to read the bignum length. compact_read_size_t returns a size_t, which is never going to be negative. might be an issue.
2:01:52Bikedeleting the build directory was enough
2:03:08karloszi guess a negative vla would be really bad, and that does tmatch the error given
2:03:29karloszi don't know enoguh about C++, but i think in C its okay to go back and forth between unisgned and signed integers
2:03:33Bikeyeah, trying to allocate nine quintillion bytes or whatever might cause a sigbus
2:04:15Bikeevery so often i learn the C conversion rules, and then i forget them because they're weird and boring
2:04:27Bikeand C++ has shit like "glvalues" so that's just a no-go
2:06:48Bikelet's see. "if the target type can represent the value, the value is unchanged. [...] otherwise, if the target type is signed, the behavior is implementation-defined"
2:07:20Bikemaybe someday we can run some strict llvm pass that'll print out six thousand warnings
2:30:33drmeisterDid you get it to work? Another issue with negative bignums?