Constant E-Lists are not working on a PowerPC use case

[SamuraiCrow]

When trying to run the list hash test in framework, it doesn't produce a pointer to an EList constant.

[SamuraiCrow]

Under new version tested under MorphOS, the list still doesn't show up but the WriteF statement may be the culprit. When breaking up the WriteF to accept one parameter at a time, all but the last parameter now shows up.

[SamuraiCrow]

Getter is returning 0 or NIL to both the listHash and the pointerHash. The problem is either with the compiler or hashBase.

[SamuraiCrow]

I rewrote hashBase module to eliminate a pure virtual constructor. (How did I ever expect that to work?) It still returns a bunch of NILs, however my attempt at writing a minimal test case still consistently works. Is there no way to recreate this bug reliably or eliminate it definitively?

[SamuraiCrow]

This may involve stack corruption. @mness1978 had found similar stack corruption in the CreativE builds in Assembly. This may need more serious consideration.

[dmcoles]

Can you explain more accurately where the issue lies and what the symptoms are? I tried compiling the listhashtest using evo and then ecx targeting os3 and os4 and got similar results with all 3

[SamuraiCrow]

When I declare an E-List inline, I'd sometimes get a NIL value. The test code is from the EEC framework in the hash data structures. https://github.com/EEC-Developers/framework/blob/master/Hash/ListHashTest.e is one I've never gotten to work right.

[dmcoles]

I don't think I can recreate this

[SamuraiCrow]

It may be an artifact of another, already-solved bug. I'll see if I can re-create it myself. If not, I'll just close it. Thanks for looking, @dmcoles .

[dmcoles]

I did find an issue and some inconsistency around how ecx works vs original E/E-VO in relation to lists though. I've sent you a pm on Facebook as it may be easier to discuss. I think it would be good to have consistency between these compilers

[SamuraiCrow]

Thanks for looking again! I'm not often on Facebook anymore but I'll look.

[SamuraiCrow]

Line 7656 in eecmain.e is one place to look because that's where elists are parsed.