#731 redim not working on passed array

[c Rex]

FreeBASIC Compiler - Version 0.90.1 (07-17-2013) for win32
dim as integer a(2)

sub temp(a() as integer)
redim a(3 to 4)
end sub

temp a()
? lbound(a), ubound(a)

[Matthew]

It's to do with a() being a fixed-size array. It's not allocated memory and can't be changed.
It works if you make it dynamic though, e.g. dim a() or redim a(0 to 2).

I don't know what the best behaviour for a fixed array would be. But from brief tests, what happens is that on redim new memory is allocated, overriding the fixed-size data, and is deallocated at the end of the sub (or at least, an array of objects will be constructed and destroyed).

On redim preserve though, it seems to give an "out of memory" error. That would be a bug.

[dkl]

The compiler/runtime currently has no way to know whether the passed array is really dynamic or not. REDIM is only possible in case it's dynamic.

How about adding a flag to the array descriptors which the compiler creates for fixed-size arrays, and letting REDIM trigger a runtime error when given a fixed-size array? ERASE could perhaps also use this to change behaviour accordingly.

[Matthew]

I've just checked in QBasic, and code like this shows a "Duplicate definition" runtime error.
A flag sounds like a good idea. But a corresponding method to test the "fixedness" of an array would be useful too.

[Matthew]

Sorry, just found a related bug: in recent builds, keywords after REDIM [PRESERVE] - e.g. REDIM PRINT foo are causing a segfault.

[dkl]

This should fix the crash issue: [12f1e2]

[Jeff Marshall]

Fixed in fbc 1.08 [91b9f4df11f21d2056316d3836ea4ea00323f71d]

If compiled with '-e' option for run-time error checking, the REDIM will throw an error.

If not compiled with '-e' then then REDIM silently fails and the array is not resized

[Jeff Marshall]

Fixed - See also:
Github PR240 - Check for fixed length arrays when using ERASE & REDIM