« first day (564 days earlier)      last day (971 days later) » 

5:48 AM
@ngn concerning ref-counting errors, we discussed that some time again. I assume you know how much memory k has allocated at current time. But do you also know where the memory is (e.g. list of address ranges)?
 
 
10 hours later…
ngn
3:22 PM
@ktye in the memory manager i store linked lists only of free chunks. i don't keep track of chunks while they are in use.
i've already fixed all known refcounting issues, btw
 
 
2 hours later…
5:49 PM
@ngn but if you start with one big-enough region of memory and never grow, it's all k-memory? then you can iterate from the start over each block and see if it's in use. would that work?
 
ngn
@ktye actually when i run out of memory i try to mmap more
@ktye but what if i didn't? what's the benefit of knowing all allocated blocks?
 
@ngn to find errors with ref/unref. after each unit test, delete all variables, traverse all free blocks (mark them if needed), traverse all memory from block to block and see if it's really unused. if you find a used block, you forgot to unref.
 
ngn
@ktye i compare \w before and after each test (except for tests that assign to global variables) - that should detect such leaks
 

« first day (564 days earlier)      last day (971 days later) »