Repository navigation
bsdiff runs very long #2
Description
Activity
- changed the title
[-]compare runs very long and never ends?[/-][+]bsdiff runs very long and never ends?[/+]on Mar 27, 2017 - changed the title
[-]bsdiff runs very long and never ends?[/-][+]bsdiff runs very long[/+]on Mar 27, 2017 Thanks for the report. I will investigate the problem.
Did some research and tests.
It's a well known old problem.
Switching to divsufsort does not help (from Freebsd)Slowdown happens in line 757 / search
Sources:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=409664;msg=14
freebsd/freebsd-src@9c63dcdIIRC in the past we saw a 24 or 48hour diff'ing of a pair large kernel binaries. There's something that can go pathologically wrong on large binary inputs. Colin's paper talk about complexities with a log in them, but it feels like a bug somewhere turns things exponential sometimes.
@phmccarty no, I did some debugging with gdb - it loops in search and not in suffix sort, "scan" variable is slowly increasing.
have switched now from rpm packages (files are rpm db's) to dpkg - now much better
- added 2 commits that reference this issue
on Dec 19, 2017 - added a commit that references this issue
on Feb 23, 2018
I have 2 10MB big files where the bsdiff runs 82 minutes
bsdiff file_a file_b delta.diff
files.zip