Fix weird GC issue

[mratsim]

This fixes the your issue.

Debugging process:

My log were different from the one you reported on the forum (I also used GcAssert)

[GCASSERT] decRef: waZctDecRefTraceback (most recent call last)
/home/beta/Programming/scratchspace/nim-help/t_raptor_db.nim(246) t_raptor_db
/home/beta/Programming/scratchspace/nim-help/t_raptor_db.nim(220) main
/home/beta/Programming/scratchspace/nim-help/raptor_db.nim(271) get_length_cutoff
/home/beta/Programming/scratchspace/nim-help/raptor_db.nim(227) load_rdb
/home/beta/Programming/scratchspace/nim-help/raptor_db.nim(168) strlen
/home/beta/.choosenim/toolchains/nim-#devel/lib/system/strmantle.nim(74) nimIntToStr
/home/beta/.choosenim/toolchains/nim-#devel/lib/system/gc.nim(439) newObj

I looked into strlen to check if there wasn't an issue in its usage as it's easy to misuse.
I added an echo addr(a) to check the address and the error changed completely to

GCASSERT] decRef: waZctDecRefTraceback (most recent call last)
/home/beta/Programming/scratchspace/nim-help/t_raptor_db.nim(246) t_raptor_db
/home/beta/Programming/scratchspace/nim-help/t_raptor_db.nim(196) main
/home/beta/Programming/scratchspace/nim-help/raptor_db.nim(273) get_length_cutoff
/home/beta/Programming/scratchspace/nim-help/raptor_db.nim(205) load_rdb
/home/beta/.choosenim/toolchains/nim-#devel/lib/system/gc.nim(435) newObjNoInit

I looked into the C code into what called newObjNoInit, found that it was creating a new string.

But in general a new string even uninitialized works fine in Nim. Looking back into the code,
the only unusual thing related to GC was returning a ref Db (instead of having a type DbRef = ref Db alias) so I tried to remove that and code compiled and ran successfully.

[pb-cdunn]

You're right this line changes the stacktrace. Odd.

[pb-cdunn]

Aha. The problem is with writing into addr result.version, which is nim-string, not cstring.

Thanks!