Lost Odyssey not Recompiling

[lawlietl4]

Okay so I have been tackling Lost Odyssey as I have a copy of the game and an interest in recomping it, I have run Analyze and am using the jumptable it created, however, whenever I run Recomp on my toml with verified addresses to the gprlr, fpr and vmx functions I get no output and my CPU immediately gets pinned. I will be adding my toml below as well as adding my jumptable file:

LO.toml
`
[main]
file_path = "../Lost Odyssey Disc 1/default.xex"
out_directory_path = "../ppc/"
switch_table_file_path = "../disc1stuff/jumptable.toml"

skip_lr = false
skip_msr = false
ctr_as_local = false
xer_as_local = false
reserved_as_local = false
cr_as_local = false
non_argument_as_local = false
non_volatile_as_local = false

restgprlr_14_address = 0x82B7A700
savegprlr_14_address = 0x82B7A6B0
restfpr_14_address = 0x82B7B17C
savefpr_14_address = 0x82B7B130
restvmx_14_address = 0x82DF2D98
savevmx_14_address = 0x82DF2B00
restvmx_64_address = 0x82DF2E2C
savevmx_64_address = 0x82DF2B94
`
jumptable.toml
https://pastebin.com/9QuzXXPN

[testdriveupgrade]

Use RexGlue, it's more flexible

[rbaucells]

This might be a similar issue to me. It likely is in Function::Analyze, for me it is happening on do while loops. Say it has a blt instruction that jumps backwards to the beginning of the loop body, in the rBase variable it tries to store a negative number (since it goes backwards) but its an unsigned integer so it overflows and then it gets stuck in an infinite loop, and never gets to the actual recompilation logic which is where it spits out the recompiled code.

[lawlietl4]

@rbaucells that's actually a very good possibility, I'm probably going to look through xenon's code and see if I can't fix that

[rbaucells]

I think i may have found a fix/solution. See how the case for it being PPC_OP_B has a if (branchDest < base) check before they do the SearchBlock, i basically copy and pasted it right after

if (lBlock == -1)
{
blocks.emplace_back(lBase, 0).projectedSize = rBase - lBase;
lBlock = blocks.size() - 1;

            // push this first, this gets overriden by the true case as it'd be further away
            DEBUG(blocks[lBlock].parent = blockBase);
            blockStack.emplace_back(lBlock);
        }

and it works fine now. Not sure if this is the correct fix but its the one that worked for me