On Friday, 27 July 2018 at 04:56:01 UTC, Mark wrote:
Hello,
I am building a toy compiler in D, and during this I ran into a
situation.
It involves checking the header of each function, going through
the arguments, and seeing if there is any duplicate identifiers.
I have a python script that feeds a whole bunch of source files
that have one error of a specific type each (with an expected
error).
The only file that causes this problem is called fnDupArgs.nerf
(I called my language nerf).
In my mind, this is a simple check, but for some reason that I
can't identify, it causes the entire computers memory to be
filled up, causing swapping on the hard drive!
Here is my project:
https://github.com/MarkTigchelaar/Nerf-Compiler-2.0
In system_test.py, that test is commented out (un comment it,
line 135)
the source of the problem is in the file analyze_semantics.d,
on line 36:
void add_func_args_to_local_variable_table(Function* func, ref
SymbolTable table) {
string[] func_args = func.arg_names.dup;
string[] arg_types = table.get_function_args(func.name);
if(func_args.length != arg_types.length) {
throw new Exception("Number of function variables is
not the
same as number of types for same
function."); <--compiler error
}
for(int i = 0; i < func_args.length; i++) {
table.add_local_variable(func_args[i], arg_types[i]);
<-- line 36
}
}
this calls the symbol tables method in the file symbol_table.d,
on line 129:
final void add_local_variable(string variable, string type)
{ <-- line 129
import fn_header_syntax_errors: duplicate_fn_args;
if(!is_declared_variable(variable)) {
variable_table[variable] = type;
variables_at_scope_level[scope_level] ~= variable;
} else {
duplicate_fn_args();
}
}
Now I've been staring at this for an hour and a half. If you
leave the one test commented out, everything passes, and all is
well.
But if you run that one extra test, it tries to swallow up all
the available memory on the computer.
I genuinely have 0 clue why this is happening, could someone
help?
If I can't figure this one out, my compiler will have a hole in
it's semantic checking that I don't want.
Thanks!
Have you tried using -profile-gc ?
