ncdu-diff
ncdu-diff copied to clipboard
Sizes not correctly sorted
Thanks for providing this ncdu-diff tool. It makes it more easy to monitor differences in disk usage.
When we load the json file produced with the ncdu-diff python script with the patched ncdu (1.14), we see this:
20,8 GiB [##########] ****
2,8 GiB [# ] ****
139,2 MiB [ ] ****
19,3 GiB [ ] ****
As you can see this is not correctly sorted on size. Pressing s also does not produce a correctly sorted output.
Pressing g to show the percentages shows a negative value:
20,8 GiB [475,1%] ****
2,8 GiB [ 63,9%] ****
139,2 MiB [ 3,1%] ****
19,3 GiB [-442,1%]****
Any idea why this happens? In the folder with the negative percentage there is new data added
Some of your sizes are out of range of a 32-bit integer. The same problem should happen without my patch, because the relevant code is in the dirlist_sort function which uses int instead of int64_t for sizes.
Thanks for your reply. Actually we don't see this behavior with the normal ncdu. Any suggestions how to can further test/debug this?
I've done some update with regard to integer signs. Could you give the master code a try?