idetools `c` crashes

[reactormonk]

To reproduce:

cd tests/caas
nimrod serve --server.type:stdin idetools_api.nim

Copy/paste the following line:

c --verbosity:0 --hints:on

Result:

nimrod.nim(91)           nimrod
nimrod.nim(55)           handleCmdLine
main.nim(438)            mainCommand
service.nim(69)          serve
main.nim(309)            mainCommand
main.nim(74)             commandCompileToC
modules.nim(193)         compileProject
modules.nim(186)         compileSystemModule
modules.nim(152)         compileModule
passes.nim(193)          processModule
passes.nim(137)          processTopLevelStmt
sem.nim(410)             myProcess
sem.nim(379)             semStmtAndGenerateGenerics
semstmts.nim(1370)       semStmt
semexprs.nim(861)        semExprNoType
semexprs.nim(2087)       semExpr
semstmts.nim(1195)       semProc
semstmts.nim(1062)       semProcAux
semstmts.nim(831)        semParamList
semtypes.nim(876)        semProcTypeNode
semtypes.nim(843)        semParamType
semtypes.nim(1214)       semTypeNode
semtypes.nim(1055)       semProcTypeWithScope
semtypes.nim(896)        semProcTypeNode
semtypes.nim(763)        liftParamType
semtypes.nim(829)        liftParamType
suggest.nim(330)         markUsed
suggest.nim(323)         suggestSym
SIGSEGV: Illegal storage access. (Attempt to read from nil?)

[reactormonk]

Did some digging, IRC logs:

not sure how to solve it. suggest.nim:325 markUsed calls suggestSym which adds it to a source map in suggetSym if isServing... also when you're asking it to compile
is it a) just create a source map because it can be cached or b) it should not add when compiling
suggest.nim(323)         suggestSym
the last line in the stacktrace
I added an echo statement right in front of it, let's see if that triggers it as well
yup, n.info is apparently the segfault
looks slightly uglier than expected... the `nil' is explicitly passed at semtypes.nim:1055, and it afterwards in liftGenericParam it is markUsed with genericParams as argument, which is always nil called from this direction
this problem also seems to propagate to cmdPretty, because the nil is also passed there
I could just add a nilguard, but I'm not sure this doesn't hide a bigger problem

[dom96]

This should be fixed now as of 79b0848